I was worried this time last year that by this time this year, companies would have slashed their engineering teams down to a handful and everything would be driven by mostly autonomous agents with human guidance. But it just hasn't happened. Do I write all my code with an agent now? Yes. Can you just give an agent a desired outcome and let it work, unsupervised? Absolutely not. I can produce more code than I used to, but if I want it to be good, to be stable, to do what the product manager and designers want, it's only about 2 to 3 times more code than before. And that productivity is impacted by the fact that I'm reviewing 2 to 3 times more code than before (and you have to review, even more so now than before, because if you just let opus or gpt 5 do its thing, you'll get some terrible results, and I've found a lot of engineers on my team are just letting it do it's thing without a lot of iteration).
The thing is: I could produce 2-3 times as much code as before _without_ an LLM, if I didn't care about my colleagues' ability to review my output properly.
Lines of code are a liability, not an asset. You want as few of them as you can get away with, without compromising the actual asset: the functionality.
A huge part of the job of Software Engineering is producing the right amount of code at the right time.
> Lines of code are a liability, not an asset. You want as few of them as you can get away with, without compromising the actual asset: the functionality.
> A huge part of the job of Software Engineering is producing the right amount of code at the right time.
Absolutely true, however my experience says that the correlation between "good software engineering practices" and "positive business outcomes" is, at best, small.
120 kloc mostly from one single developer copy-pasting and keeping non-compilable code for an obsolete target "for reference" for a decade, becoming both a ball of mud and a whole pantheon of god classes? No unit tests, no code review? Won awards.
Properly engineered, mandatory code review, mandatory unit tests, dev meetings to knowledge-share? People with the money said too slow, closed it down.
(Sometimes people bring up how bad Musk's code was at PayPal. I never bothered investigating. Successful product though, wasn't it?)
Survivorship bias, you don’t know of all the failed projects that couldn’t get off the ground because of incompetent development team and practices that lead a product to its demise, or a product that is possible within constraints that otherwise could have been a success, but not realised by sloppy work and incompetence.
Furthermore the dependencies you choose to build your product are presumably filtered for engineering practices or world class engineers. So given the choice you yourself prefer top quality engineering, so do your customers. Much in the same way you are a customer of your projects dependencies. Difference being, as developers we get to see how the sausage is made, our customers only see second and third order effects.
> you don’t know of all the failed projects that couldn’t get off the ground because of incompetent development team and practices that lead a product to its demise
Trying to ignore the nuance is hard in your position or the following one I’ll give is difficult.. but is the opposite potentially true as well? We don’t know how many projects failed because of over optimizing, too much time spent on design and engineering decisions. It’s of getting out and MVP to market. I only say this because I have been apart of a few of these.
Well, that's the other side of incompetence: they know how to spin their tools, but they don't know when to stop, or how to stop the change requests, the balance between shippable, maintainable, and what the market wants at that time.
I understand and of course I am familiar with the hypothetical you are trying to set up here but I was specifically pointing out a logical fallacy I see banded round all the time by people who should know better or educate themselves.
I will say that if “good engineering practices” comes up in your root cause analysis for a failure to launch a product you are not thinking critically.
The statistical problem is small sample size, not survivorship bias, as I got to see things before failure. These two examples are merely illustrative of things I've seen.
I think both are at play here and I don’t know about Musk’s programming skills. But it seems that they had other very good programmers (including levchin). So maybe business success can buy good engineers to clean the messed up code. I’m not sure how it goes with AI now though.
> Absolutely true, however my experience says that the correlation between "good software engineering practices" and "positive business outcomes" is, at best, small.
One of the most uncomfortable truths about our profession is that there is no floor to how bad software can be while still making people billions of dollars.
The first thing they buy of the success money though, is a struggling technical competitors, so that this team can clean up there mess. This would mean that AI is only a good contributor at startups and with prototypes.
There's a trade-off to be made, and it's not necessarily clear where the trade-off sits for any particular company, or even team within the company.
One product person described it as eating vs breathing. Availability is like breathing: if you stop being available (including, but not limited to, because your software is a big ball of mud) then you're going to die pretty quickly. Product is like eating: you might not die so quickly but if no-one's buying what you're selling then you're still not going to survive.
The team I'm part of is a platform team, so we're closer to being lungs than being stomach. We can (and I appreciate being able to) focus more on stability than feature development.
The software only had to be good enough to support the business. It doesn't have to win a Turing Award, and probably wouldn't help the business if it did.
If you can call being an “also ran” in a field they had a ten year march on their competitors in success, yeah.
Truth be told it was the shoddy code they were forced to use for the vanity of their paymaster might well have held them back, though manifestly that is not a bad thing. Probably the best outcome, really.
I have been saying this for years, I once had a heated argument about a small system of maybe 1000 lines of code that was technically superior and more scalable but was freaking 1000 lines of code to maintain compared to the quick and dirty 10 lines of code it was suppose to abstract and make generic (for future use of course).
That with also countless debates over insignificant features in frontend apps at the cost of extra code. Frontend code is very susceptible to this maintenance cost dilema.
Many developers are too focused on delivery value compared to maintenance cost. It is unfortunate that non-technical management can see value delivered, but not maintenance cost incurred. With LLM-assisted code this has become many times worse.
>Many developers are too focused on delivery value compared to maintenance cost.
The way this is worded feels like it leaves the blame on developers. Aren't these developers focused on exactly what they are being judged by? Shouldn't we say it is the management who is too focused on delivery value compared to maintenance cost? Is it the developer's job to guide management or the manager's job to request guidance, assuming such guidance is needed?
We need to make sure the responsibility to resolve this problem falls on those with the power to act on it, and in this, developers tend to be receiving far more responsibility to fix than power to fix.
I don't think so, my experience is that most often falls into two scenarios:
1) The devs pushing for more complex solutions, covering obscure edge case scenarios, feature-creep, "future-architecturing" because they are more interesting to implement. Classic over-engineering problems.
2) The features the managers actually want are usually boring or annoying to implement and the devs just work around any big architectural problems caused by the feature delivery.
1 is 100% on the devs, 2 it varies wildly, the willingness to address architectural problems are often under pressure by time-delivery estimates from managers. But many devs (especially in companies with low morale) will often just work around issues because addressing the underlying problems can be very difficult and/or time consuming.
Meaning either the dev wants to do the right thing but doesn't have the time, or the dev doesn't care enough and just pushes the tech debt down to the future (when hopefully they will be at another job).
LLMs makes both problems significantly worse, although they are also often very helpful with the big restructurings mentioned in 2. The dev can still be lazy and the deadline can still be too tight even with that extra LLM help.
I wonder though if, as long as you have LLMs to maintain the extra code, it's worth it to gain the new feature. Less tech debt than your intuition expects.
It's apples and oranges. Lines of code is just a measurement. You can't look at a whole codebase and say "these X lines are essential, those Y lines are excess, delete those" -- the lines interact with each other to create the whole thing.
It would be like saying "Weight in kilograms of course is an asset for an airplane. That's what keeps it in the air. What other real asset is the airplane made of?"
This is the thing. You "spend" lines of code, you dont produce it. The produced part is the outcome - a functional feature, stability improvements, some business outcome. Measuring productivity with LoCs is like measuring output with cash burn.
I think that coding reviews are no longer feasible as they used to.
The pace and expectations have increased and a human can barely cope with reviewing its own code, let alone colleagues'.
They are not going to disappear in critical aspects of a codebase, nor shouldn't, but the industry will eventually reward self sufficient individuals able to keep the pace, harness and run adversarial reviews against the design and implementation autonomously.
I'll also say the harsh truth. A well implemented adversarial flow will do either better than your peers or will deliver 95% of the value at a fraction of the cost.
The industry has never valued product, let alone code quality except in places they are core to the business.
Otherwise you would not have MIT-bred leetcode ninjas writing react/tailwind bugged monstrosities at half a million/year for billion dollar products.
You're definitely out of the line with such language.
And naive.
Your biggest error is thinking that:
1. there's quality software out there. It's definitely far from the industry standard, even in high budget big techs, far from it
2. that people opening the wallet care about software quality. good has been always better than perfect
3. that the average PR review has significant impact on code quality, clearly false by point 1.
4. that proper AI usage for doing adversarial reviews of plans and implementations won't catch by far more issues than the average pr out there nit picking coding styles but not even bothering to test the feature/branch
Thus I stand to my opinion. Quality PRs are too expensive nowadays and the engineers that will thrive will be the already great ones that will be enhanced by the tool, not companies shelling money on reviewing the average garbage MIT-bred leetcode blackbelt is slopping.
They will only involve core/critical code and libraries.
Don't bother my man. All of the commenters talk about the "bad code" AI writes as if the code that's out there in the world isn't also compete garbage.
Of course, everyone here is in the 99th percentile in ability, am I right?
The sheer number of people I'm meeting that say about trivial things "AI cant do that" is astonishing.
In all of that I'm not saying good engineering practices are bad, I'm speculating that the economics will push towards strong individuals soloing features at good code quality.
Bogging down strong contributors into reviewing code, now that there's more work to review in the first place might be not economically feasible unless the code is critical to the business.
"A huge part of the job of Software Engineering is producing the right amount of code at the right time."
But the whole job of the owners is to lower the costs as much as possible to produce the product, especially in an environment with higher costs to borrow (higher interest rates at the fed)
This is one of the things I ran into early on. LLM needs to compute the determinant of a matrix? Sure, just spit out some huge hyper optimized implementation of it. Good luck maintaining that. Slapping "use industry standard open source libraries for common functions" has improved the quality of LLM output for me by such a large margin.
>I was worried this time last year that by this time this year, companies would have slashed their engineering teams down to a handful and everything would be driven by mostly autonomous agents with human guidance. But it just hasn't happened.
I find this somewhat puzzling. I thought things were moving quickly, but at this time last year I couldn't even get Claude (using Cursor) to spin me up a service skeleton that would compile, let alone do anything meaningful.
I know it feels like a long time somehow, but it was only between November and February that things started to actually somewhat work without significant hand holding. Even now, it seems like we're still figuring out how to fully leverage the current models and tooling, even in organizations that have largely gotten on board.
It's not all that surprising that people were worried and believed this. The AI companies and infrastructure companies partnering with them have spent a lot of money and time trying to convince people this is the case year after year. The critical clue people miss is that everyone claiming that has very clear financial incentives to convince people that's the case even when they know it isn't. Anyone who was actually building with LLMs and judging for themselves based on its performance knew fully well that wasn't the case year after year.
I've said this before: if anthropic (et al) thought they genuinely had a shot at replacing even 30% of white collar work, they would ABSOLUTELY NOT warn ANYONE. They would do what oil, leaded gas, and cigarette companies did. Swear under oath this is completely safe, commit GRIEVOUS societal harm that you explicitly promised wouldn't happen, and then end up in history books instead of jail for reasons beyond my ability to fathom.
No. The very fact they are trying to "warn" us means it's all marketing.
This has been corroborated for me on the engineering front that I can't find a single IC I respect who actually thought there was any evidence AI was going to live up to the hype. I saw a lot of people I always thought were idiots/sycophants/brown nosers go insane with AI. Never saw anyone id trust to help me cross a street blindfolded say more that "I may be wrong, but I'm not seeing any evidence yet".
Fwiw , you're conflating multiple things and consequently drawing premature conclusions.
It can be massively over hyped for it's current capacity and decimate the white collar work.
A lot of the difference of opinion is down to their point of view. At my dayjob, LLMs will not live up to anything because the enterprise is not structured to take advantage of it's strength. That's unlikely to change within the foreseeable future.
I strongly suspect you mostly talked with people coming from just such a background, because it's hard to go beyond our own bubbles
Sure, naturally. And yet parent commenter is remarking that simultaneously no AI-true-believer startups have supplanted the old money, and simultaneously despite much talk the bigcos have not slashed headcount to tiny AI-powered teams.
Hmm, isn't that just down to the fact people always vastly underestimate the amount of work software takes?
It takes a lot of time to successfully deliver a product. Even at the more "extreme" end of expectation - like saying it's a X10 multiplier (I'd disagree on that, it's more like 0.5-3 - depending on the type of work you're working on) you'd still need multiple years to go from first line of code to displacing established players. Things just don't change that fast.
Depends on the bigco IMO. I'm not sure what kind of layoff numbers for Year Three of genAI would grab your attention, but I assure you, hundreds of thousands have mine.
So you're saying companies correctly estimated the need to cut staff to remain competitive and did so -- or at least, some state of affairs empirically indistinguishable from this. Not much of an argument, sorry to say
Yes, this... all the hype from the leading AI companies just pattern-matched so many past cases where things didn't pan out. really giving me the bad vibes..
>The critical clue people miss is that everyone claiming that has very clear financial incentives to convince people that's the case even when they know it isn't.
Social media is flooded with bots pushing this narrative - coding is dead, engineers are all cooked, the latest model is scarily good, "what am I for?", etc.
A good rule of thumb is that if it's a human being and not a bot, they'll use the word "slop" at some point.
coding harnesses improvements mattered more than llm improvements this past year. You could solve problems on claude sonnet on claude code that you couldn't solve a year prior.
> at this time last year I couldn't even get Claude (using Cursor) to spin me up a service skeleton that would compile, let alone do anything meaningful
I've been using it to do this for 2 years now. And many people with me. The change you mention is one of is primarily one of Overton windows, of vibes.
>I've been using it to do this for 2 years now. And many people with me. The change you mention is one of is primarily one of Overton windows, of vibes.
I suspect this may have depended on the specific framework. I quite literally could not get Claude (in Cursor) to give me a basic Micronaut setup in a fresh workspace with essentially a "hello, world" API. I would guess that if you're using something like Python and FastAPI, it might have been an easier task or better represented in the data.
The difference that I observed in the Opus 4.5 era is that Claude could take a service framework it has never seen before (proprietary corp) and figure it out.
I've had success with Aider from October of 2025 (from my git logs). I have to differ from the general sentiment over here. The tools and models are good enough to make a change in the way work is done. Of course, Big Tech. will move fast and make mistakes but this is qualitatively different from other recent hypes (e.g. crypto/NFT). It has had genuine impact on people and their lives and a lot of simpler jobs with larger teams have been replaced by these models and one operator.
Exactly the same for me. Sonnet 3.5+Aider made it possible.
I got downvoted though, apparently you and me are liars, haha. Easier to believe that than admit that their take of "I adopted it exactly when it became viable, now it's good, and before that it was a waste of time" was wrong. They're the people who were loudly claiming here last summer it was useless, asking "show me anything useful that has been coded using LLMs". Come to think of it, it has now been a few months since I saw one of those. Used to be every other thread.
Yeah, I strongly remember getting Sonnet 3.5 with Aider to boostrap an (albeit basic) project and getting it to work. Especially vivid because I told my roommate at the time about it and he also tried it out and was also shocked. I'd put maybe $20 in credits into the API haha. Feels so quaint, it's almost a foregone conclusion with the current models and harnesses.
Do I write all my code with an agent now? Yes. Can you just give an agent a desired outcome and let it work, unsupervised? Absolutely not.
I strongly suspect that developers moving from writing code to managing agents to write code for them is very similar to developers moving into leadership and management roles and managing ICs to write code for them.
Some devs just 'get it' and thrive, leading a team really well and building a great culture. But a lot of them don't, especially if they don't get the support necessary to understand what changes when you move from IC to manager. If the team (or agent swarm) isn't performing well it often isn't a problem with them. It's a problem with the new manager still trying to stay on top of everything and micromanaging all the things. Alternatively, the new manager is completely hands off and only appears at a check-in point (one-to-one, agent completes a task, etc) where they crap on the work and get cross.
I have no evidence for this, but I'd guess that putting developers through some sort of management training would make them much better at using agentic swarms.
We are beset on all sides with companies declaring agentic coding a failure and here you are stating as a matter of fact some teams “thrive” with this probabilistic expensive approach to approximating working code?
All the while concluding with “I have no evidence for any of this”.
Agentic coding is absolutely not a failure. It's just not the 10x that CEOs really wanted it to be.
We learned that some tasks don't really benefit from AI while others do. My team went from 7 people to 2 (went to new teams, no layoffs), and we're doing the same amount if not more work than we used to.
Is it more draining and lacking of focused work? Yes.
Is it more money for the business? Yes.
Friend, you are commenting on a thread where one of the most prominent figures in tech (rightly or wrongly) is saying something did not meet expectations. As far as we can tell this is a man with every incentive to exaggerate and boost these products.
In my world, when something is expensive and doesn’t meet expectations it called a failure. Especially when something has been as hyped, scrutinized, defended and attacked as vibe coding.
Honestly, if you are the director of robotics at a firm I think it’s time you took a cold shower.
I’m sorry, what’s the question? Genuinely not sure.
I’ll rephrase, Zuckerberg would certainly enjoy agentic coding to be a wild success because it means less staff and more products he could fail to create.
Again, Zuck's success criteria don't necessarily align with that of others. In fact, I'd expect his success criteria are substantially different to most.
Consider: Someone who "expects" their bank account to have $100M in it before they turn 30 and "only" gets it to $10M.
From the point of view of a normal sane person, they are experiencing "wild success", and yet at the same time they are definitely "going slower than [they] expected".
Why would Zuckerberg want to boost their AI, they have none to speak of after Llama really? Zuckerberg actually has every expectation to downplay AI so as to save face that Llama failed compared to frontier models.
Who said the layoffs were a success? It's a short term fix (or correction from pandemic hiring) that may still nevertheless have long term consequences.
On the other side of this you have two companies growing revenue literally the fastest ever in any market segment, just for said tech. Somebody is spending this money and thinks it's worth it. Let's check back in six months.
From my experience with daily stand-ups, I think they can be a significant boost to that, too. Though you will absolutely have to wear a manager hat with their estimates and breakdowns, not just fire-and-forget, as they're often as not wildly over-optimistic about task complexity.
Directionally, that's why companies are getting rid of token leaderboards and imposing limits on LLM costs. There's a diminishing marginal return to tokens
There is increase in (not only perceived) value _somewhere_ - IMO depends on the organizational culture. At my work I found going over 100$ on OpenAI/Anthropic's API pricing does not produce any meaningful additional output. It might be much different in different companies.
I am also leading a small team of myself and 2 others and we're getting a lot done. The short lines of communication of being a small group + the power of agents has been great for us.
> Is it more draining and lacking of focused work?
To be completely honest, I’m living life right now. I love programming with my bare hands, but man I’m living just building a gajillion things a mile a minute with LLMs. I then come home and spend hours building stuff for myself using local models. I’ve never been so excited about just building shit, that I sometimes want to pull all nighters because I’ve been in the zone (a for work and at home).
Draining? Sorry… inject that LLM serum right into my veins
Now we just need to find those tasks. I want to believe.
> and we're doing the same amount if not more work than we used to
Zero evidence for this. It's programmers self-reporting their own productivity. (Have we not learned this lesson after 50 years of programming practice?)
Because this companies defined the goal to replace humans with AI what didn't happen. What happened is that the humans can work faster and have more coffee breaks while the AI iterates for the next review and iteration round by that human.
why would companies who build tools for developers say they'd want to replace humans with AI? it was never the goal, it was never stated like that. they said that by the end of 2026 most of the code being output would be generated by LLMs, and is pretty much true.
are we tho? i don't really see anyone going back to planning and coding by hand, that ship has sailed and it's not coming back. people ditching agentic workflows altogether would be failure, people figuring out it's not a miracle tool and has uses for which it's not as productive is correction.
The bit I don't have evidence for is whether or not teaching ICs to be managers would improve how they use agentic AI. I have plenty of evidence for the efficacy and effectiveness of AI itself (although not qualitative or obviously causal unfortunately.)
> plenty of evidence … although not qualitative or obviously causal
Those two things are the opposite of each other (evidence, but only anecdotally; you cant be both).
Anyway.
More tangible to your argument; what is your argument that this will be more effective than just prompt engineering?
Ive long believed that prompt engineering is a losers game; if there is a trivial set of tricks that improve the output, they will simply be automatically applied.
We see this playing out with the system prompts in coding agents and image gen.
The value of learning “photo realistic studio lighting…” was non existent. The nano banana api is capable of taking a naive prompt and expanding it with these tricks.
People who devoted themselves to learning these “magical incantations” wasted their time and effort; and it was obvious, from the beginning this would be true.
Now.
With managing agents; if a trivial set of management tricks can drastically improve the results, why are you better off learning them now, rather than waiting for them to be baked into cursor/codex/claude in easy mode?
What makes you believe this is a valuable investment in time and effort?
Even if we accept that right now assigning personas to agents and managing them as a manager yields good results, the horizon for change right now is so short, it seems extraordinary to suggest mass management and leadership training for engineers.
We should just wait and see.
All in investments like this would just be tokenmaxing in a funny hat.
> Some devs just 'get it' and thrive, leading a team really well and building a great culture.
I don't think it's this because the outcome you get from AI isn't controllable. You can give it the best prompts and design suggestions and it'll still give you completely wrong or horribly written code.
If you were a manager and one of your reports kept producing completely wrong and horribly written code that other folks on the team keep bringing up as problematic in PR reviews or privately, that developer would eventually be fired for someone better.
But in the AI case, there is no replacement because all of the LLMs have severe problems.
> because the outcome you get from AI isn't controllable. You can give it the best prompts and design suggestions and it'll still give you completely wrong or horribly written code.
I don’t have a dog in this fight but it seems you’re not accounting for iteration and feedback. A horse will veer off of a road if not occasionally nudged to stay on it, but is useful transportation, nevertheless.
> I don’t have a dog in this fight but it seems you’re not accounting for iteration and feedback
You can provide AI official sources to look at and dozens of prompts. I've lost track of the number of times where it didn't arrive at the right answer with tons of opportunities to correct itself based on feedback.
Just an endless of sea of "you're absolutely right to have brought that up, I didn't think about that" and other phrases it constantly uses when it fails to provide a solution. Fast forward 20 minutes later and it starts providing the same nonsense it did at the beginning because it forgot what it already said.
The code solutions it provides are so consistently bad but it's not limited to code. I recently tried a YouTube feature where it can generate AI thumbnails from your video. The results were really lackluster. It completely ignored my feedback like "use a real webcam photo of me that you see in the video", to which the AI recreated a completely different looking human that wasn't me. It even swapped out my real glasses with a rendering of glasses I don't have and kept on making incorrect assumptions about everything. After about 10 prompts and 20 minutes of waiting for thumbnails I gave up, it was really poor.
None of that supports the claim that it "isn't controllable," though. A curious mind should probably find it interesting that it can be fallible in those ways yet still be useful for producing work, and ask how both can be true at the same time.
>I don't think it's this because the outcome you get from AI isn't controllable. You can give it the best prompts and design suggestions and it'll still give you completely wrong or horribly written code.
sorry if it's not the case but i have the feeling you still think AI coding involves talking to a chatbox and copypasting the answers
Yes, an agent does feel like a very eager and very knowledgeable but also very clueless junior developer sometimes. But no, working with agents is not like leading a team of developers. You don't have to make sure an agent stays motivated, give it the feeling that its work is valued (I can let an agent spend some time building a prototype and then decide not to use it, if I did the same thing with something a junior developer took a week to build, that probably wouldn't be a good idea) etc.
> If the team (or agent swarm) isn't performing well it often isn't a problem with them. It's a problem with the new manager still trying to stay on top of everything and micromanaging all the things.
You see problems in the results? Simple, don’t check the results!
Let codex review Claude output and the contrary. Win for everyone involved, more code, more features, more token usage, promotions, code to be debug by juniors and the cycle of life can continue.
> I strongly suspect that developers moving from writing code to managing agents to write code for them is very similar to developers moving into leadership and management roles and managing ICs to write code for them.
I disagree.
From ICs that I lead, I expect that they learn fast. On the other hand, LLMs are basically incapable of improving.
Another issue that I can see is that I don't particularly like eager new colleagues who come up with (hallucinate) wrong answers. At the beginning, if you are uncertain, learn, and if you have questions that learning does not answer by itself, ask questions. But strongly avoid hallucinating answers. New colleagues can be taught that, LLMs not so.
> I don't particularly like eager new colleagues who come up with (hallucinate) wrong answers.
People are most likely to come up with suggestions and ideas earlier rather than later.
Often they’re not learning what is “correct” but just how to “fit in”. A fresh perspective can be good even if flawed. It can help others spark new ideas and think outside the box.
LLMs can be 'taught' though. You can give them additional context or instructions. The difference is that they can't really teach themselves.
This is roughly what I'm saying - someone who's managing an IC can steer them on the right course, and someone who's managing an agent can also steer it on the right course, so teaching someone how to handle ICs well gives them skills that are also applicable to handling agents well.
It's not perfectly analogous obviously because ICs are people and need to be managed as people, but I really think the skills are quite transferable in one direction. I'll add that I don't think someone learning to manage agents would necessarily become a good people manager.
> Some devs just 'get it' and thrive, leading a team really well and building a great culture.
I think what makes a dev well suited for AI isn't the same as managing a team. What really helped me get productive is having to write a lot of user stories and acceptance criteria with the wisdom of being a dev tasked with implementing them. Also, being on the refinement calls, answering questions, and updating requirements/AC is good feedback for authoring better requirements. If you're good at authoring requirements, checking output, and communicating corrections concisely then you can get the LLMs to sing.
I get why you think this but you are incorrect. Why? Because managing a team of people is very very different from managing outsourced contractors, and LLMs are much more like the latter than the former.
"I strongly suspect that developers moving from writing code to managing agents to write code for them is very similar to developers moving into leadership and management roles and managing ICs to write code for them."
I doubt that. Management is mostly dealing with people, the actual "management"* part is not where developers moving into management roles typically fail, it's the people part. With agents you have the management part without the people part.
I think you are confusing manager with ICs. Managers don't really read or review the codes. What you are describing is where agents are doing all the coding and reviewing without people in the loop. I don't think the op is working with the code as black box. He is more about describing the higher IC workflow.
Whether we still need people in the coding loop is not a trivial difference
Good ones do. Reading the code someone is contributing is a powerful signal about how well they're doing.
ICs who manage swarms of agents should operate the same way. They set them off to do something, and then look at the output to see if it's going well.
That's the point I'm making here: managing a team of ICs and managing a swarm of agents has a lot of overlap in the systems and processes you can use to see if it's working well. By teaching ICs to be better managers I think they'd get better at using agentic AI.
There's such and such. In some companies, the leaf engineers report to a team lead, which might or might not be granted this 'manager' title. Those poor fellows essentially doing double-duty and are the most likely candidates for burn-out.
I agree there is a skill change but agents aren't people so managing them is very different. E.g. you can spin up 1000 agents. Get them inti tight loops and get them more context and so on. A manager doesn't do that with people.
Agents aren't people, but they anthropomorphise themselves real hard.
> E.g. you can spin up 1000 agents. Get them inti tight loops and get them more context and so on. A manager doesn't do that with people.
My dad had various work stories, one from when he interviewed applicants:
"So, why did you leave your last job?"
"After 6 months, the management discovered my entire floor and the one next to it had been hired to do the same task. One of the floors had to go, I got unlucky."
In my experience, it's the opposite. People who become dependent on AI agents are avoiding human contact, and people who become managers are seeking it out (for better or worse).
I disagree with your 2nd assertion. Even engineers who are less tech lead style engineers, can gain a significant boost in productivity by being able to quickly run through POCs and build an understanding of surrounding areas of their work, so they are able to contribute more.
For eg I am able to make React changes much faster and the changes are higher quality, given frontend dev has never been my job role. I’m able to spin up test harnesses, write throw away glue code, test against large datasets, etc
> I am able to make React changes much faster and the changes are higher quality, given frontend dev has never been my job role
Man, if I had a dollar for every time someone said "I'm not good at X, but LLMs are so impressive at it". Like do you think there might be some connection between those two points?!
> I’m able to spin up test harnesses, write throw away glue code, test against large datasets, eTc
It seems that I don’t like coding when I read these kinds of statements. If I’m doing an experimentation, it would be a few lines at most. Because that’s all I needed before I can write a solution.
Writing code is the last tool to design with. Thinking and a bit of sketching is what I do mostly. Then I verify small bits with code (mostly for checking a library when the documentation is lacking or a stub when I’m focusing on another part). Otherwise, it’s just enough code to get it working well and refactoring when the requirements changes.
Do you mostly use agentic swarms? If so, I’d be curious of your use cases. People talk about “managing agentic swarms” a decent bit, especially on LinkedIn. I just don’t see how they are the best solution for majority of development use cases. At best they seem like using only a hammer to make sculpture.. or a sandwich.
The problem often is that the new manager can teach their hires to write better code and train them to be new managers. Given that a team leader in such a scenario is often checking for correctness, I am not sure if current LLM based AI will ever be able to do it. We need new AI for it.
I find it interesting that when drawing this parallel you mention that some devs 'get it' and 'build a great culture'; I think this is exactly where the analogy breaks down. Good managers get great results from people (and for people! they are linked).
Good AI managers are just running optimization loops at more declarative levels. Yeah, you need to get comfortable with less personal review of code for both, but I think the differences outweigh the commonalities - it's much easier for someone with a more 'traditional' IC model to be successful with agents then they would be with management, and I think most (good) management training would be entirely irrelevant. Parallels are maybe tighter to higher IC progressions.
> Can you just give an agent a desired outcome and let it work, unsupervised? Absolutely not.
Ignoring instructions - whether in AGENTS.md or my prompt - is the worst of it, and it routinely happens. It just waives things that I explicitly told it to do as part of the design.
Vibe coders (in the true sense, zero oversight) claim that you just need to prompt it carefully. That's completely untrue when faced with your careful prompt being ignored.
I even have "don't overrule me without asking" in my global AGENTS.md, and it simply doesn't do that.
These are word generators, not agents, I’m really not sure why people think they could be capable agents (ie independent) when they consistently ignore instructions, generate the wrong things and then double down when questioned, etc etc.
You’ve been sold something that simply doesn’t work for the purported use case (intelligence) and instead is like a stupid database of all world knowledge with the appearance of intelligence.
Useful tools at times (if you bear in mind their limitations), but not close to intelligent, independent agents.
> instead is like a stupid database of all world knowledge with the appearance of intelligence.
A "stupid" database would be better, based on what I get when I ask whether all of Oregon state is North of New York City. Indian English has a word for it: oversmart.
Your context isn’t to give it orders, they just don’t work like that. Your context (AGENTS.me, skills, per-request context we are sending in for each request to bots) is to give it the info it needs in the language category it’s trained for the answers you want; you have to give it a clear instruction each prompt. Basically, when you have a long session, you can see this by saying, ok, now moving onto another thing, blah blah blah (implicitly ignoring all previous instructions). It can even back fire - nagging too much about don’t skip tests in the context can make it slip into the linguistic space where there is some emergency and faking the results might be justified (I imagine there is a certain amount of training out there “just making the tests pass for now, will fix later, I promise.” If you rarely mention tests except “this one is failing, please investigate what is going on” (an informational outcome not a test outcome), it doesn’t really “cheat” (tho it can leap to conclusions as always). The tests need to be some deterministic step in the process anyways, tests don’t need fuzzy word directed search capabilities. But the models just don’t have the structure to allow feeding in a ten page set of rules and follow them. You can add a step to say, please check this git commit for compliance with the 23 rules in this standards file, and it will work better to catch the gaps.
Whenever I work on a challenging question I worry about this, because Opus will easily think for 200k tokens on the first prompt. I fear any follow up discussion is lobotomised!
Yeah totally feel you on this one! To avoid it or at least limit that I usually ask it to create external memories, clearing the context window and leaving just a trace to recall the memory created for that run. I hope it might help!
Try writing it in first person instead of second person or neutral.
A while ago someone had a similar complaint on here and shared some example lines, and that popped out at me immediately. However much structure we've wrapped these in, they're still text generators trained on all sorts of things, and if you think about a narrative where first and second person speech would be used, try to imagine context: In first person, it's most likely a description of something as it happens or someone planning what they will do. But in second person, especially command form, you open up to the possibility of commands being ignored, misunderstood, or actively rebelled against.
Whoever that was back then did some quick tests and found the pattern held, first person got it to follow far more reliably.
I’m convinced the magic bullet is deterministic checks. Linters, static analyzers, etc. Whatever you can do to create deterministic gates that the LLM simply must overcome to reach a “done” state, do it. Has been making a huge difference for my team, but sister teams are so invested in writing the perfect Make No Mistakes prompt that they just can’t see it.
Basically I treat it like a junior dev. We don’t get junior devs to write code correctly by cajoling them just right, we add CI gates. It still works.
> Whatever you can do to create deterministic gates that the LLM simply must overcome to reach a “done” state, do it.
First thing Gemini did when I tried that was turn off all the rules in eslint.config.mjs claiming they were "overly stylistic"
Yes, it got better once I explicitly told it not to disable any rules, so I accept I was holding it wrong but I do worry just how many footguns it puts into other things because I didn't know the right guardrails to give it.
Although I don't use it, Antigravity should be configurable to prevent reads / writes to certain files. You should definitely do that and ideally configure a hook to automatically run the linter after the LLM completes, then feed it back to the LLM if it's red. Don't prompt your agent to do stuff, build your loop in a deterministic manner.
I made a tool called ProjectLint to lint architectural decisions and anything that ESLint and ArchUnit and others like that wouldn't be a good fit for. It's not public yet, but is fairly rudimentary and based on Go + goja: https://github.com/dop251/goja
You write rules in ESLint, it makes sure that they're followed. If it's a file in your repo, it can obviously be checked. Now whether you can describe your architecture well enough to lint it (at least the stuff you care about enforcing), that's a different question.
In my case it's more like: "Oh, the AI messed up this pattern, but it will need to be followed N more times in the future, better write another projectlint rule." After a while, you can even copy them over between projects, as long as they're following the same conventions.
It will burn up the tokens to get through the deterministic gates, more so when n order dependencies are involved in the mix. Enough typewritters and monkeys could get it done too.
The agreed architecture is to use signing between two micros, so that a third can orchestrate between them in zero trust way (and to prevent a distributed monolith). It just decides that we can trust the third and skips the signing.
It can't follow orders because it isn't thinking - even in a best case scenario, all you're doing with AGENTS.md is altering the probability distribution of outcomes away from things you don't want towards things you do want.
However that still means there's always some probability it will do things you told it not to, it's just reduced
From what I can tell, the "established wisdom" is to get Fable to plan and Opus to implement (for cost purposes). The problem there is that Opus could ignore whatever it likes from Fable's plan.
The Claude harness effectively summarizes the discovery into a plan document, clears, and starts with just that summary - so you'd be clearing context regardless.
i've yet to see a case where opus "ignores whatever it likes".
opus will definitely ignore instructions if you give it contradictory instructions, or a plan that has steps that obviously don't work with each other. but if you give it a coherent plan, it will follow it.
Honestly this is where I would have fable generate a checklist and you just monitor opus to ensure it is going through the checklist. I think ignore is often the result of a context that is not focused enough.
Companies are putting a ton of effort into getting to that point of having agents do the work unsupervised. Whoever gets there first is going to be the winner.
I personally don't think it's possible and I haven't written a line of code since Sept 2025.
There's an AI psychosis going on right now, especially among the execs or management class, and we all gotta nod our heads in agreement and burn through tokens.
Agents already run unsupervised, and they can code unsupervised too. The real question is what worthwhile work we should point these capabilities at. Nobody has really cracked that yet.
The same worthwhile things we were working on before agents. I’m personally using them to autonomously build open-source Shopify for every vertical. I set out building it before AI, but AI actually makes the dream feel achievable.
Yes we have provided custom storefronts for people running the SaaS on-prem and we have a handful of people using our cloud hosting. This is all without any marketing efforts and word of mouth. Albeit, we launched the Shopify alternative in Dec 2025 and the Toast alternative in May 2026. The gym and grocery ones in the works.
My main goal is to first get all of these open-source alternatives to start building themselves autonomously using loops and a Hermes-like scheduler before I focused on marketing. This is almost complete.
For marketing, we are building a GTM engine using an open-source CRM (Twenty). We have LLMs use the Twenty CRM API to bring in leads from X, LinkedIn, and the Web.
The cloud hosting is not the only monetization. We’re going to use these open-source SaaS to build a decentralized, interoperable marketplace where the people actually bring value, the sellers, can sell without those rent-seeking entities like Amazon taking a piece of every sale. LLMs are already going to start jumping across these marketplace moats.
The other monetization is going to be letting agents actually run these SaaS and see if they run a business autonomously. Like VendBench but an actual online business. I’m thinking of starting a designer brand, connecting to a POD (print on demand) and then let the agent create seasonal lines, handle customer service, and make sure orders are going to the POD and being processed. Doing this with restaurants and other verticals will probably need some human supervision.
Interesting. Where is this product located? If I'm looking at your profile correctly, this appears to be Openship[1], which is just a collection of starter templates from something from Nextjs and something called Keystone for each vertical. Is this what it is you are talking about?
I'm very curious how much revenue this is generating.
Yes, it’s built on the shoulders of giants, Next.js[0] and lesser-known Keystone.js[1].
Next is a full stack framework and Keystone is a CMS built on top of Prisma and GraphQL. Keystone was created by this Australian company called Thinkmill. They have used it to help businesses build custom backend systems for more than a decade. But it needed to be deployed separately from Next and they were using emotion css for their dashboard and I wanted to use Tailwind/Shadcn. So first, I had to make the Next Keystone Starter that brought in Keystone into Next so each SaaS is just 1 Next app with a built-in storefront, GraphQL API, and dashboard.
Once that was built (and it took a while tbh), I started to build the Shopify and Toast alternative. But the itch to get these built quickly and autonomously had me working on the harness in the past months and now that is nearly complete.
Here is the e-commerce[2] and restaurant[3] repos. They have a link to deployed demos you can check out as well.
As far as revenue, I don’t feel comfortable relaying that right now. We have other revenue streams like fractional CTO where companies give us equity to manage all their tech and that is quite hard to quantify. Before Openfronts being built, I built Openship, and e-commerce OMS and that has exceeded 5M orders processed since its inception in 2019. That’s not counting orders by businesses running it on-prem.
I actually posted about this vision on HN[4] when I launched Openship and the response is what kept me building.
Second, take this for what it is: your product may not be compelling in its current form. Building it to many different markets will not make it compelling. If you had a stronger revenue, please share it. This sounds incredibly thin.
Third, dont mistake building the same thing 20 times for different verticals for bonifide software skills. When a SWE builds the thing they have built before its usually to learn a language which is the easiest part of software. There is a reason a common adage in software is "9 women cant make a baby in a month". Breath is no replacement for depth.
Don’t take my word for it then, ask any terminal agent to dig in and get an idea of how good these apps are.
In the end, I built Openship and Openfront for my e-commerce business and then turned them into SaaS. All the revenue for these are just a cherry on top of our existing e-commerce businesses.
And I worked with Next and Keystone long before AI came along. Check my GitHub commits if you need some back story.
And I’m not building these 20 SaaS to prove I have bonafide SWE skills. I’m building them because I plan to have my own gyms, hotels, grocery stores down the line powered by these SaaS. SWE to me a means to an end and that’s to have many different businesses.
> Don’t take my word for it then, ask any terminal agent to dig in and get an idea of how good these apps are.
Okay but, you know this isn't a quality metric right? These models are incredibly biased towards positive confirmation of the prompt. I could give them nearly any repo and they would sing the praises of the best parts, if I asked.
I never told you what prompt to even feed your AI my guy. As far as I know, you can be like rip this repo to shreds and tell me what’s wrong with it and you’ll get your answer.
Crazy how you guys blindly trust proprietary apps where you can’t even read the code but asking you to read open-source code is psychosis?
You say you only use open-source and I’m the same way but I’ve asked agents to evaluate open-source alternatives. How is that psychosis? If it’s typescript or JavaScript, I can read the code but if it’s Rust/Go/etc, yeah I get AI to read it and tell me if it can solve my problem or if it has the features I need.
It’s actually the whole premise of my open source alternative directory called Opensource Builders[0].
That’s the neat part, you can use it headlessly. The built in dashboard and storefront are using the GraphQL API but you can deploy your own external dashboard and storefront using the same API.
We’re also very bullish that the chat interface is the universal UX now. Instead of sifting through the dashboard to change a product price or sell in a new region, you can use the built-in agent and just tell it to do that. Every Openfront comes with an MCP server that interfaces with the API so the agent can literally do anything you can do using the dashboard and API. This is where an agent running the business autonomously comes in.
And even then if you’re not satisfied with the backend API for each vertical, these Next apps can be forked and adapted to tightly fit your business instead of you messing with configs, you can make the app your own.
If you have runway, it’s a good time to start your own thing or join someone who is. Personally, I cannot force myself to wade through slop PRs from careless coworkers. If that’s the job now, I’d rather run a hotdog stand or something.
Luckily, I don’t think things are that dire. I think the companies issuing AI mandates are manufacturing sawdust, and even if it works, it would just enable them to burn through customer goodwill in record time as they make user-hostile decisions free from engineer pushback.
These are going to be a few tough years, but I think the opportunities to start something new are everywhere.
Of course I don't know either way. However, I feel like it's going to be different, but maybe not apocalyptic. The market I am building for is photographers, and from what I can gather (and know first hand as a customer) is that there's real discontent with the toolmakers in our ecosystem.
It seems like in the quest for every tech company to become a multi-billion dollar empire, they've lost the plot on making hammers for their customers and have instead turned into some kind of strip mining operation. A totally AI agent-driven company is a MBA wet dream, and I think a fever dream. If Adobe, for example, were to achieve it, I don't think they'd use it to fix the backlog of bugs overnight. I believe they'd just become an even more incoherent zombie, trying to extract rents from creative cloud subscriptions.
In the meantime, photographers still need tools. If you wanna run a software company as a regular small-to-medium-sized business, you may find some customers that are happy to buy a quality hammer. The unicorn startup days might be behind us, but I'd be okay with that.
Now, if AI obviates creatives altogether I don't really know what to say. I'd morn the loss of a world that became so tasteless that AI-generated decorations are good enough, for starters.
But even the market for "photographers" is arguably in secular decline. Smartphones and AI have turned everyone into a photographer. The world is awash in visual content.
It's true. Magazines sales are dead. Product photography is probably toast (I'm seeing companies test what they can get away with using AI generated images of their food/products right now). Sports? I don't even follow it but I suspect it's all video content now. Funny enough wedding photography seems to be doing fine as far as a commercial segment goes.
But there are segments of the market that were always small, and AI doesn't threaten. Namely, fine-art and long-term documentary. The kind of work that gets shown as exhibition and sold as prints or photobooks. I can't imagine this stuff be supplanted by generative AI, because it has never been about economic efficiency. It's a labor of love for a photographer to work on a project for years at a time because the subject is important to them. And the customers of these works are not likely to accept lower-cost substitutes produced by a AI. The whole thing is too much about taste and thoroughly infused with humanist ethics. When I buy a photobook, I want to know that the artist lived and produced what I'm seeing, not just that it's pretty ink dots on paper.
From a market analysis perspective it's foolish to cater to these tiny and commercially nonviable artists. However, they have enormous esteem and influence within the community, and make part of their living as educators. I believe making tools that they need, and without blood-sucking cloud subscriptions, you are indirectly marketing to all the creatives (usually tending toward commercial) who attend workshops and see what tools they are running the workshop with.
I don't think it's possible either. DHH pointed out in the most recent episode of Rework that AI removes a lot of the barriers to shipping code, therefore making it possible to build in lots of different directions in ways that was prohibitive to many organizations in the past. But this isn't necessarily a good thing, companies still need to understand what to build in order to ship a cohesive product. AI is great for prototyping and refining use cases in ways that are far superior to static figma designs, etc., but it is not a replacement for taste and execution.
But a slop machine that haphazardly shoots features against the wall to see what sticks still isn't a winning product strategy in 2026. And the problem I see increasingly is that so much energy is being focused on how to deliver with AI internally and externally that is not being expended to advance a company's product. I believe more and more in the idea that for many startups and companies, the actual "customers" are the investors and the product-market fit that companies seek is the product of the company itself, because this is all being driven from the top down, not by customers and users in the market asking for AI features.
> I believe more and more in the idea that for many startups and companies, the actual "customers" are the investors and the product-market fit that companies seek is the product of the company itself
In many respects this reflects the growing K-shaped nature of our economy. Average consumers don't matter because you really just need a small cohort of wealthy individuals to be hyper-invested in your product, 'regular' consumption is therefore just a way to keep things relatively on rails rather than the actual economic driver.
All of these AI-first companies don't actually have any market fit, so what they're doing is selling an imaginary product so that they can get investments and loans. As you said the company is the product.
The bottleneck wasn't the coding, but previously it had the second order effect of slowing product development decisions enough to improve product cohesion.
"What can we get rid of for MVP" as a design strategy vs a way to iterate fast, for instance. Cutting things isn't a way to product cohesion, especially if you never go back to do the full-featured version.
Sometimes I wonder how many features or products flopped because the MVP dropped the things that would've actually taken off, and the business "smartly" pivoted away.
There's still a limit to how many new features you could shove in front of your users per month. But what if they were all much more baked out of the gate?
(See also: "data driven" product management as an excuse to not have your own vision for the product. If three competitors build a lot more in the span of six months, but have to depend more on their own skills and instincts vs A/Bing every little detail, maybe more of them will ship more bold and interesting new things.)
But that's the nature of the beast isnt it? A probabilistic token predictor will all always have some errors from a human perspective, more and more energy, money and resources will always be needed to control and direct towards desired outcomes.
“ But this isn't necessarily a good thing, companies still need to understand what to build in order to ship a cohesive product”
In other words writing more code means fk all without vision, strategy, taste etc. Google has had lots of engineers on many projects - look at the grave yard. The constraint on progress is not code.
Wake me up when this dumb experiment is over. Some of us are years ahead it seems until others get in the same page of understanding
I'm really struggling to get an agent to write code I'm happy with. It's mostly pretty awful.
I've a fairly simple c# coding style. But simple is proving a bit more difficult to convey than I thought.
I get it to produce code. I then have to spend along time convincing myself it's correct. If I don't I end up embarrassing myself when a coworker reviews it, questions it and it's obvious I don't properly understand it.
This is really starting to screw with me mentally. It's like everyone in the world is saying they can fly by flapping their arms (dark factories). When I try I just stay in the same spot burning a lot of energy.
I don't think everyone in the world is saying they can fly by flapping their arms. It's a small number of very vocal, very online, AI enthusiasts, many who have a financial stake in AI winning.
work from tests to implementation. Validate the tests, and work with the agent to ensure there are not more cases that need testing. Then you can let the agent implement the code and you can refine it until it's simple enough but covers the test cases. TDD is the only way to use agents effectively, imo
Thanks. I have been doing this tdd abd it's a big improvement but the code is still pretty awful. Two skills I rely on are Matt pococks grill-me and his tdd. These massively helped, and technically the code generated is correct but it's still hard to follow and bloated.
Measuring aircraft production per weight doesnt sound like that bad of a proxy. If I hear that Boeing produced 300 kilotons worth of aircrafts more this year, I'd be right to suspect they've ramped up production.
I think you're overlooking the lesson of Goodhart's law: you can use a metric, but if you make it a target, it stops being a good metric. Neither "tons of aircraft" nor "lines of code" should be the measuring stick -- and if they're not, then they can still be used as metrics.
To be fair, the hazard with AI agents is that they generate fluent output that is often facile, so it's easy to do a lot of things while having a lot of defects. That's a sign that quality control is not prioritized enough. A change in quality will also reduce the utility of SLOC as a metric, but the mechanism is different than what Charles Good hart pointed out.
"companies would have slashed their engineering teams down to a handful and everything would be driven by mostly autonomous agents with human guidance. But it just hasn't happened. "
> I was worried this time last year that by this time this year, companies would have slashed their engineering teams down to a handful and everything would be driven by mostly autonomous agents with human guidance. But it just hasn't happened.
Amara’s law: We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.
This continues to be applied to AI where people think is going to be next 12 to 18 months. Changes are coming but certainly not at the rate Zuckerberg and most people are thinking.
I believe this in general but the issue with AI specifically is historically in every cycle we've overestimated the potential of AI in both the short and the long run - people in the 60s were genuinely quite convinced we were close to AGI for a few years, as ridiculous as that now sounds.
This is why you get "AI winters" but we've never had a "steam engine winter" or a "railway winter" or a "petrochemicals winter"
It's interesting to score our predictions of various technologies. I've been waiting for hoverboards for 40 years. Our predictions of nuclear power being in every consumer appliance were way off. We just couldn't make it safe and cheap enough. This is a common theme for "physical" technologies. On the other hand, we greatly underestimated the advent and potential of the internet. Most science fiction of the 20th century envisioned very large computers with limited interconnected capabilities. We made computers far smarter and more ubiquitous than most could ever have conceived. I see AI as a function of the kinds of technologies we consistently underestimate.
Yes, that's a fair counterpoint. I guess my counter-counterpoint would be that LLMs actually seem to have characteristics closer to the first group than the second, in that they (currently at least) need enormous quantities of physical things and energy in order to work.
So the nuclear power problem of "it works fine and it's actually very good in a lot of ways, it's just too expensive" could be quite relevant
Back in the day there was a mantra: best amount od code is 0.
Now we have agents spitting lines after lines.
I am not afraid of my future. Even if one person can do work of 5, the amount of generated code will grow exponentially. And not everything can he vibecoded with 0 knowledge.
There is a complexity that need understanding to change and optimize. For now :)
I think it doesn't prove much that it hasn't happened yet. Companies might just be moving slower than you think, and are still planning on doing it. And, in many corners, "don't manually write code" is being joined by "don't manually read code" as an attractive principle.
> in many corners, "don't manually write code" is being joined by "don't manually read code" as an attractive principle.
I'm pretty sure I know where the failure case on that one is. The reason we're still manually reading code is to catch the failures and edge cases that the LLM fails to; not reading the code doesn't magically make the code good.
Sure but there's also little reason to think we'll be able to replace xyz role entirely with software next month, or year, or decade. It's easy to disprove claims about the future; it's quite difficult to make believable ones.
you might have to think the way through though and these companies are already being caught up with the huge token costs at the same time.
There was an interesting comment during the cloudflare layoffs (partially driven by the fact that the company was bleeding money also because of its token costs from one estimate being 5* million$ per month (I feel so silly that I accidentally had written/meant 500 and had kentonv do the stats on that part :-( Sorry kentonv!), don't quote me on that though)
The part was that there is only an enough marketshare in the first place. Cloudflare was doing some crazy experiments like operating matrix on cf workers and wordpress alternative and fediverse and so much stuff.
So they basically spent 10x the amount of token (and the token costs) and I imagine as such the reading code of that part was getting sidelined as the attractive principle you are talking about.
Yet the market can't bring an actual demand 10x times though. These are things which nudge a user slightly but the actual impact on user growth isn't 10x or even justifiable within some cases given the costs.
Yet at the same time driving up the people who actually know their stuff and firing them because of the token costs. The people who have actually mitigated some of the largest DDOS attacks and are the backbone behind cf cash-cow (enterprise payments) is the fact that they have had the experience and entreprise knowledge about these things, yet they are literally removing that by firing workers and oh replacing them with interns. (They got 1111 interns and fired 1100 employees or something iirc)
It's weird and I have talked to some people about it but there is a disconnect between what management is hearing about AI and the ground reality of things. Reviewing code is becoming the bottleneck but if you don't review code and are shipping things to production, then you can get fired as I have talked about in some of my other comments sharing a story about how a guy shipped code to prod and the response was "but claude generated it" and got fired because the company basically said, look we basically don't care if it was generated by claude but the responsibility was on you to check it (review) and because the commit was done by you, you are gonna be treated responsible and he got fired from his job.
Yet this was the same company which was asking its employee to play around with claude at their free time, the manager of the employee I talked to being the most automatable person, the company employees working till 1 AM because they were saying to management that things were fine but they were being burried under the technical debt,that employee that I talked to got honest with the management and told reality and the management treated them as a person who didn't know AI or were the odd one out.
Sooo I don't know actually to be honest.
TLDR: reviewing code is being treated as the bottleneck but it is also the only thing stopping your company from imploding under technical debt, actual debt because of token costs etc. I remain skeptical if we should treat it as a bottleneck or as a safeguard mechanism. After all, if nobody's in the loop then whose responsible?
Reviewing code isn't a bottleneck so much so its a safeguard mechanism in my opinion. Also things differ in corporate land and hobby land and I would prefer corporate to not be using the practices that I do with how I do things for fun in my hobby time.
Side note: Even more so, I think I am a LiteLLM security working group maintainer and I have seen first hand on how much damage it can do in supply chain even when things were done right from LiteLLM side and the fault was within the side of ironically a security product that they used called Trivy.
There are things which you can do to be better prone to supply chain attacks in general but there is no full bullet proof way of doing so and in such.
Caution (should) be taken when dealing with corporate systems and as such I sweat a little when anyone suggests code review to be completely eliminated. Things (are/can be) different in hobby/prototyping world though.
> (partially driven by the fact that the company was bleeding money also because of its token costs from one estimate being 500 million$ per month, don't quote me on that though)
Cloudflare had 5000 employees (pre-layoff), so you are suggesting that every single one of them (eng, HR, legal, finance, receptionists) was using $100k tokens per month (that's $1.2M annualized, per employee), for a total of 3x gross revenue going to AI spend.
Let's imagine that this isn't absurd on its face. If true, then you'd expect Cloudflare's Q1 earnings to show a massive, massive net loss. In fact Cloudflare was cash flow positive in Q1.
The rest of your post is more qualitative, so harder to disprove, but from what I can tell, it seems equally made up.
I had mistakenly written 500 million when it was around 5 million dollars so I messed up its 5 million per month[See Source], not 500 million. I wish to have a genuine discussion while you are here though because i can be wrong, I usually am and I would love to have a good faith discussion, thanks in advance!
I will try to back up a lot of it with hackernews comments from the thread when cloudflare layoffs were suggested so that I don't accidentally mis-represent anything and My suggestion wasn't a critique of cloudflare and please don't take it as such. The question was simply of the AI token costs associated.
and this was the comment that I was referencing to[0] which states the following:
> There was an recent article on X with an interesting take - it could be that companies are doing layoffs not because AI is making them more productive but because it hasn't. Their costs have gone up paying for expensive AI but haven't seen any revenue benefits to offset it.
An child comment of it talks about the coinbase layoffs which had happened around the same time[1]:
> (..) In 2023, their "Technology and Development" line item shows $1.32bn going out, and by 2025 it'd ballooned to $1.67bn. This is despite headcount actually contracting by almost a thousand people between those two statements.
Regarding this: > Let's imagine that this isn't absurd on its face. If true, then you'd expect Cloudflare's Q1 earnings to show a massive, massive net loss. In fact Cloudflare was cash flow positive in Q1.
We might be forgetting that (from my understanding, Cloudflare has never had profits) (positive annual net income) with an astronomically large P/E ratio.
> > The fact so many orgs opt for immediate greed over long-term growth really is its own canary that leadership and governance both has failed the marshmallow test.
> Why do you think it's greed? The company's stock is down and they just missed expectations on their last earnings report (unheard of in big tech in the last 2 years).
> It seems more like a traditional layoff scenario
Another comment [from the Layoff thread][2] which might summarize some things:
"Their AI costs have increased 600% but this hasn't translated into actual revenue. Also they are probably projecting AI costs to keep growing. They've done the math and at some point it is going to affect their bottom line. Reducing or limiting AI usage would be inconceivable given Cloudflare itself has invested on AI and is selling AI services. Instead they've opted for reducing about 20% of their head count."
I genuinely wish if we can have a good faith discussion about it. I appreciate cloudflare as a product myself and actively use cf tunnels, which is why I care about it as well and I wish to have a good faith discussion about it hopefully as well :-D
> The rest of your post is more qualitative, so harder to disprove, but from what I can tell, it seems equally made up.
I can be wrong, I usually am and if I am wrong, I wish to learn from it and I wish to improve as a person too!
I have learnt from this discussion (up until now) that I should mostly try to provide sources whenever talking on a public place/ on the internet so that I can be more accurate and I sincerely wish to have a good faith discussion once again, thanks and have a good day @kentonv :-D
Although there is a difference in accrual earnings and cash flows. I was wrong with the number provided which although not as big as 1.5B of AI spend, is still a comparatively large number itself.
I have written an more in-depth comment if that interests ya (in a good faith discussion and please be kind to everybody)
also please don't call @kentonv an idiot and please read the HN guidelines[0]: Be kind. Don't be snarky. Converse curiously; don't cross-examine. Edit out swipes.
I have experienced and feel very much the same, and it is refreshing to see a realistic post about the success of agentic coding instead of the usual hype or doom.
I would add that a lot of that extra code is often in test/demo paths... I tend to think of working with an Agent as a "team" of 1 + agent... where the developer is now wearing a QA and PM hat in addition to lead/sr dev. That the work getting done is now roughly the offset of what a team would have produced and that coordination needs to step back and treat each individual with an agent as roughly a dev team. Coordination overhead and mythical man month still apply, just at a layer up.
I can similarly output 2-3x more code but everything stalls down to me having to review and integrate in a meaningful way the moment I am the one that has to maintain that code.
It's eerie to observe collaborators output code they don't understand, spend days chatting with Claude instead of reading (like really reading) compiler's output or 3 pages of manual, and how lost and oblivious they look when the AI fixates on solving a different problem than the one they have been tasked.
As crazy as it may sound, my workflow today does not look too different from a year ago - where I was already heavy into claude code.
Im not certain things will look too different a year from now either. We still have serious bottlenecks in terms of focus/attention you have for both delegating agent work and being able to review it. Even if we solve the "trust what ai does" problem, these cognitive deficit issues still exist - for teams coordinating work, even users adopting new shit, etc.
As an industry we are leaning heavy into accepting "slop" as the status quo - we care more about efficiency of output right now. Slop will get better & we can become more adaptive to living with the paradox of amazing yet delicate systems generated by AI. But I feel big shifts coming in this regard and if/when it does we may find ourselves in the dystopia of broader unemployment with worse net outcomes.
I do think the teams that ship quality with AI will do so by learning to slow down
As a hobbyist coder, I wonder how much more new code, fundamentally, there needs to be? New devices need new firmware, new cars etc, but how much of that is bespoke? Sure, a new movie or a book is new entertainment, but I've already seen that movie and read that book, they just had different jackets. What do these "engineers" actually do that is novel and how much of the pizza is the novel slice is?
I feel like the increased reviewing time is consistently understated. I’m just an IC, but it seems obvious to me that you cannot cut staff and achieve increased output. There literally aren’t enough eyeballs to go around reviewing code when everyone is 2-3xing their output. I spend so much more time reviewing code; reviews that are sorely needed because I regularly catch batshit insane “fixes” that work but would quickly turn the codebase into a mess (the most recent one being a multi-hundred line diff that I went and fixed in 2 lines in 15 min). Maybe I’m underthinking it but it seems obvious that you either maintain the same output with fewer staff or you gain increased output with the same staff. All the companies that are attempting to cut staff and gain increased output are chasing an impossibility and throwing away their opportunity to accelerate.
Increase output with same staff etc doesn’t necessarily help the firm financially…
Raw engineering productivity is irrelevant. Managers are employed by shareholders to make them wealthier - long term this comes in the form of incremental positive cash flows.
I kinda love that you made this post feel negative enough that a bunch of AI skeptics are enthusiastically agreeing with a post suggesting that the realistic, pragmatic bear case for AI is, uh, 2-3x productivity improvements.
Could be lack of imagination on my part but I truly can't imaging shipping 1000's of lines of code that I can't understand (beyond low-stakes prototypes). That means there's a ceiling on productivity gains.
If you end up with 2 to 3 times more code. That is HORRIBLE, because it means about 50-66% of the code is otherwise unnecessary. Those are eventually going to become unmaintainable garbage.
However, if you get 2 to 3 times the code in the interim, that's probably less than what's needed. I find myself cycle through almost 10x-20x amount of code implementations to get what I want which is actually less code, simple solution and desired behavior.
Given a specific behavior, there are usually just 1 simplest implementation, whether done by human or AI. However, there are 100 ways to do it with more complexity and either handwritten or AI slop, it will mean pain down the line. We used to have a lot of handwritten complexity because of certain design pattern culture, but they used to be contained because the ability to generate them is costly. Now it's much more risky and therefore more important to have simplicity as the guiding principle in ALL projects.
My experience is that AI agents write 20-33% more code than I would for a given feature set, mostly because they are worse at remembering what utility functions already exist and less likely to merge similar functions into more generic ones. They generate that code 2-10 times faster than I could. Defect density is harder to compare: I probably generate more "dumb" defects due to oversight or missing unit tests, but fewer defects that violate domain rules or architectural objectives.
Frankly: if you want it to be good and stable, you can't really go any faster than before. The time it takes you to review all the code is no less than it would've to just write it in the first place, because the actual typing things out was never the part which took up time.
> In retrospect, he said, the "trajectory of the agentic development over at least the last four months hasn't really accelerated in the way that we expected," and that the company's bets on the new structure "haven't come to fruition yet." Zuckerberg was referring to AI agents, automated systems that can execute tasks on behalf of a user.
Hard to guess exactly what he means by "trajectory of the agentic development" but my best guess is that he means that Meta's own internal efforts to improve the agent (aka longer form tool-using) capabilities of their own in-house models hasn't improved to the point that they can drive an agent harness like Codex or Claude Code in a comparable manner to the best OpenAI and Anthropic models.
At a further guess, that was part of their goal in reassigning large numbers of employees to help label data for their AI efforts.
the pessimistic take is their harness is no better than thise available and he thinks they all suck together.
from a high level, these agents absolutely do not function as a rational human through even medium scoped problems. even when you try to add memory, you just multiply halucinated context which just makes it error out on tasks in harder to detect manner.
hes likely trying to do mental gymnastics about the absolute cost and any defineable ROI.
I expect it's a model problem and not a harness problem, purely because some of the best harnesses (including OpenAI Codex itself) are open source and can be very easily tried against a new model.
and I'm saying all the harnesses in the world arn't going to solve the myopic ability.
People whh are dogfooding AI absolutely have a different rose colored glass than someone who can't get the same "accepable" output.
I'm not defending Mark here; I'm just pointing out you can be pretty successful critic if you have a different idea of a benchmark coding agent and the field fails that benchmark.
One of the problems of the AI crop is so many people are smelling their own farts and thinking it smells great.
The gap between "useful chatbot" and "useful agent" is way bigger than people realize. A chatbot can be wrong 10% of the time and still help you. An agent that's wrong 10% of the time is sending bad emails and making wrong API calls with no one checking.
I see this as the gap between an general-purpose agent and a coding agent. A coding agent can imagine something to be true, test it, discover that it's wrong, and recover.
But if you go beyond what can be tested easily, asking the agent to do real work rather than writing a patch, imagining things to be true is a problem.
This to me is the big leap from being good at coding to being good at many other tasks.
Coding could be treated as a low stakes (time & money consequences for retries) closed loop system where most other tasks cannot.
If it screws up booking your flight/hotel room, how does the agent verify this, and even if it verifies.. there is an actual cost to changes/cancellations.
Similar with agentic e-commerce, lots of ability to screw that up and just seems ripe for fraud / being picked off by bad actors.
Seems like to make agents safe we need tentative, reversible transactions. How do you set up a travel plan and then review it? How do you modify it later?
Unfortunately, travel keeps getting less flexible, with worse cancelation policies.
Because a lot of things in the real world require forward planning, I don't think everything can be just-in-time and tentative/reversible. Some things have to be committed to else you have real consequences and lose your money.
There is no way to do cheap and flexible air travel. If the tickets are cheap, than just a few last minute cancellations can destroy your margin for that flight, if you were to offer them for free. Add in the the risk of your seat reservations being gamed by filling up seats and then canceling, and you get a recipe for disaster.
Trains are usually different because they are much cheaper to operate per trip (for the train operator, not the track operators, but that's a different discussion), so running a half empty train is much less of a problem - especially since you don't need to plan for how much fuel to use ahead of time.
Right, these are actual physical real world problems that aren't going away just because it would be easier for agentic workflows.
Another example - agentic food ordering.
How much more convenient would this make your life vs how much of an error rate would you tolerate if the cost/repercussions are on you?
Would a customer be happy if 2% of the time it sends 20 pizzas to a random address in their contacts list instead of 2 pizzas to their own home? Or 5% of the time it completely ignores your dietary restrictions/allergies and orders an entire meal of food you explicitly told it you cannot eat?
Real world problems don't go away just because it would make the tech neater & tidier.
Try it without "reasoning". As you can see in my example (and GP), it meanders to correctness eventually after emphatically being wrong, and most reasoning modes hide that from you.
If LLMs worked the way people want to believe they do, there’d be no reason to start in the wrong place — a computer should have the facts!
The problem is that with text/code, judgement is hard. Here is what it looks like for physical activity: https://www.youtube.com/shorts/lK7TjujKQLw It's hard to see how that it's not useful at best and could be a disaster for any unsupervised use.
The gulf is bridgeable. The problem is that a lot of people are building agents without strong enough judgment layers around them. Work that can be verified with reasonable accuracy are the sweet spot right now.
IMO present technology is tailored for an experienced developer to give agents manageable tasks that can be one-shot. The marketing right now reminds me of the 90s when AskJeeves promised natural language search when the technology was fundamentally still stuck in keyword search, and learning to craft a search query for Google is today’s prompt engineering
Only with an LLM that's actually at agent-quality.
If "useful chatbot" and "useful agent" are two rungs on a ladder, the rung before them is "useful autocomplete". Autocomplete that only gets the next token right 90% of the time won't give you compiling code.
> he expects that the social media giant will begin to experience more significant benefits from its AI investments within the next three to six months.
He is hallucinating just like AI, and unable to come to terms with the facts on the ground. Meta has lost plot about 5 years back - with metaverse, VR, glasses and AI. They should sit back and think with a calm head, about what exactly their core product is. Unfortunately there is none, except a few acquired ones: whatsapp and instagram.
It's ads. They have an infinite money glitch called ads, allowing them to waste billions chasing other pipe dreams. The fact that so many of those pipe dreams turned into nothing doesn't hurt them one bit.
It's how venture capital works: the theory is that you invest in a dozen pie-in-the-sky pipedreams, almost all of them fail, and the one that doesn't pays for the rest. If you're not failing most of the time, you're not taking on enough risk.
Yeah they're a media conglomerate and their customers are advertisers. They have unstoppable traction with Instagram and Whatsapp and a legacy model with Facebook. There really isn't much more to it except that they need to keep trying new things in case those properties get taken down by a competitor.
no they are a media company lol. Who 'own' the media assets and don't produce them. Magical really. Truly.
its like the newspapers back in the day - they even had crude ads in there / half naked women.
IG took that to another level - showing endless amounts of soft porn etc.
Deep down Zuck hates himself - he so badly wants to be revered like Jobs. But he's a bozo with a whole lotta money. And he will go-down in history as that, unless, he completely blows Meta up.
He's like a crazy dictator going off the rails right now. Combined with trying to ruin the life of the author of Careless People for exposing his crimes against society, and his employees regularly leaking his internal announcements.
Having agents is like going from walking to having a bicycle.
Business executives look at this and think "at this rate of progress we'll have self-driving cars in a few years!" and start making serious plans for that world.
In reality I think we're going to be riding bikes for a long time. That situation of increased individual contributor productivity makes engineers more valuable, and increases the utility of engineers rather than making them a burden on your budget.
Thus, cutting headcount right as they had huge potential to become vastly more productive was a stupid move. It's an admission that you don't know how to manage people effectively, which is embarrassing when you're paid mountains of money for your management skills.
Have you seen the original bicycles. Boneshakers and so on. No pneumatic tires. They were just enough more efficient than walking to sell but other than that pretty dreadful. Penny farthings where your feet directly drove the wheel. So many injuries!
Nobody knows if we are going to "just" be riding bikes for a long time.
To give time for society to adapt I hope it's the case, but we really have no idea.
Right, but if your real assertion is “we have no idea”, it seems you should point your skepticism significantly more towards the people betting $100 billion dollars that self-driving cars are coming next year than the ones who aren’t.
Evidence towards aside, the risk calculation depends on how much upside is from the 100B if progress starts the same rate. I think it's hard to bound that accurately
It looks pretty clear LLMs don't get us there by themselves, no amount of duct tape, WD-40, etc. gets us past, say, the mathematical certainty of hallucinations.
I mean, we don't know it any more than we don't know someone won't come out with cold fusion tomorrow, but it's a fundamental breakthrough away from where we're at. This isn't some routine engineering project with a guarantee of completion if you're just willing to keep pouring the billions. That's playing the lotto, you can pour away and get flat nothing.
The only difference is they're pouring billions and praying a rabbit comes out of the hat, but it's actually not much reason to expect they're going to pull the cold-fusion level rabbit out of their hat they'd need to get us past bikes.
I think what everyone underestimated was the absolute bonkers amount of compute it will take and how that compute must scale in order to keep up with larger and larger models.
More than that, I think people overestimate how much AI will progress as you throw more compute at it. It’s the “9 women can’t deliver a baby in a month” equivalent of AI. Additional compute won’t magically give you AGI.
How do you know though? People thought ANNs and perceptrons were basically useless until pretty much around the time of AlexNet and while there were some architectural improvements (CNNs, transformers), the biggest thing that changed was that we simply threw more compute at it.
It's definitely too early to declare that more compute won't make a difference.
It was data. Alexnet worked because imagenet existed. It took less than a year between the dataset becoming public and the first version of Alexnet being trained.
Maybe not AGI, but if you look at the differences between, say, GPT-2 and GPT 5.5, it's remarkable how well it works to mostly just throw scale at the problem.
The difference is a lot more than just throwing scale at it, pretty much everything useful comes from an evolving landscape of post-training techniques.
Of course, param count and context length are also important because they increase the model's overall fidelity, but a base model without SFT, RHLF etc is effectively useless.
Correct. That is what I was trying to hint at. Yes, massive compute is needed to train ai, but it isn’t the only thing. A lot of research and experimentation goes into moving the marker just a little bit. Innovation can’t be forced into weekly sprints, it takes its own time.
Research and experimentation on neural nets has been going on since the 70s (arguably much earlier even), but the lions share of capability changes has all been in the last couple years.
Scale was really the unlock; the new pre and post training techniques and architectures are very cool and useful but they definitely aren't the differentiators when comparing to the previous era of NLP.
I think the unlock only happens once though. I think that's where people are misled at the moment, the technology was there but required huge compute and data ingest to show improvement, but we have done that now. What's next for a giant leap is not more compute, and what new data we can provide now pales in comparison to that first ingest.
I think the “unlock” is that AI firms were given trillions of dollars to discover new techniques. In fact, there are very few industries where a sudden influx of that much money would not lead to rapid advancements. It’s not really unique to the AI field.
Your timeline is a bit backwards: they were given trillions of dollars only after they'd demonstrated a few pretty remarkable advancements.
The fact that their advancement suggested that pouring more compute would continue working was also especially attractive to investors: it made a massive R&D budget feel like less of a risk.
BERT is a transformer! The unlock happened within transformers, yes, but they were not exactly super new or innovative architectures at that time. The scale was the main innovation that brought us from BERT to today's frontier.
I was under the impression that the deltas between versions were shrinking- i.e. gpt 4 -> 5 was much less impactful than 3 -> 4 or 2 -> 3. If the growth is getting diminishing returns, I can't say I'm optimistic without finding a drastically different approach.
Not really. I didn't use GPT-2, but I don't think there was much difference between we got what out of GPT-3.5 and 5.5. It's still an unreliable tool which will go off the rails the second you aren't watching it like a hawk. It still has zero intelligence or ability to reason as to why its output is flawed. Just throwing more compute hasn't gotten us anything worth using.
I was involved in three efforts to commercialize foundation models before they were ready in the 2010s so I have a good picture of how progress works at this sort of thing and the pace a lot of the industry has been talking about is unrealistic: like people were disappointed with the rate of development of Apple Intelligence but it's actually progressed at about the rate I expected.
Bad capabilities but maybe less wrong output? All the funny memes of Google explaining some fake aphorism is t really something Apple product would go for. Successful navigation of technology over the decades requires some timing finesse. I don’t know.
Is that a problem for Meta though? They recently announced they're going to sell their excess compute, so I imagine the actual problem is they're resorting to doing that because AI isn't having nearly the effect/usage it was supposed to and now Zuck is being a sore winner about it
If Meta is selling their compute and Twitter is selling their compute and the stuff doesn't do anything you don't need an economics degree to figure out what's going to happen to the price of compute. In particular because 'compute' is a euphemism given that this is far from general purpose capacity, those are specialized chips that largely do one thing
All these companies are going to sit on their gazillion data centers once the mania dies down and will have a big problem about what to do with their mountain of hardware
Meta doesn't seem to be able to produce anything close to a frontier model. The selling of compute capacity seems to be acceptance of "compute is wasted on this crappy avocado model, we'd be better off allowing something better to run".
The problem is clearly in the model architecture, the training and the data fed into the model which is causing them to give up on using their compute exclusively for their own models. They can't get it right so may as well sell the compute to someone that can.
Meta has made some very strange decisions in terms of who it's hired to lead various aspects of AI, including the model-building efforts. Also lots to marvel at re: its ability to coordinate (or not coordinate) various efforts by all these big brains.
Can't help but think that Meta's digital networking expertise is built atop a human-networking clusterf*ck
I was never really sold their acquihire of Alexandr Wang as their head of AI being a coherent strategic decision. I just don’t see how his experience and background actually applies for frontier LLM model building.
I think there would easily be a few other hundred engineers and execs at frontier labs who are more in the loop for cutting edge architecture/secret sauce - with a track record of actually doing it - that could be had for a fraction of the price.
From the outside Meta's attempts to pivot from open source releases to fast follow closed models fell flat when they tried to prematurely monetize it. They could have owned the open weight model world but tried to pivot to closed weight chatbots before an actually viable revenue model appeared.
Does meta have the research talent to create a SOTA frontier model? Yann LeCun has left Meta and I don’t think either alexandr wang or zuck have enough credibility to attract talent to create one.
it's possible Yann LeCun wasn't the right guy either. He seemed to be more focused at finding the next model architecture rather than iterating on the current LLM architecture to build a competitive frontier model.
It will scale inefficiently until efficiency breakthroughs occur, but it's really hard to predict when those breakthroughs will happen. Plan on the worst, but be ready and capable of capitalizing when it happens!
for us, maybe, but for someone who never really used the workflow, or looked at the “thinking” output where models spin their tokens on the stupidest shit, i can see how it wasn’t obvious.
I think Meta’s massive compute investment was never about its 100,000 engineers running coding models, but its 3,500,000,000 users wanting to use AI in every single product (and some new ones: Meta AI, glasses, etc.) So I would think that’s the part that’s not being utilized anywhere near the amount they hoped...
A common one is "users don't care about privacy. that's why they use facebook. [zuckerberg was right?]"
No, you silly, silly people. People want to use products that allow them to communicate or reconnect with people or ...
They don't 'want' constantly changing privacy settings or changing TOS. If this is the best HN can come up with, ostensibly filled with S Valley people... well, it says a lot
I think this is the problem for companies with a single person atop - when the company needs things they aren’t good at, the company cannot respond effectively. Zuckerberg was good at running a company to sell ads on an addictive platform; whether that will make him good at the next ten years of profitable tech innovation is difficult to see; people hate ads and dislike the addictions, so Anthropic or whom ever has to walk a different path; they have multiple smart people working together to find that path; Meta does not seem to have that collective vision of competing experts to draw on.
I suspect there are many things AI can do to help people and make their lives better. But that's not how business works: products get made and marketed because they make their owners more money. Totally different goal.
Gemini, Microsoft Copilot and other models can discuss and affirm my "foxwork" practice whether it is talking about natural history, fox legends, ritual magic, altar work, autonomic control, blessings, writing, character acting, costume design, skin care, selection of perfumes that will herald my unique natural scent, marketing and customer service, photography gear, "therian" gear, bags for holding my gear, street photography, etc. They always write like somebody who's read much more widely than anyone I've ever met and rival the legendary Tamamo-no-Mae for "speaking intelligently about any subject" [1]
Meta AI can crack jokes and that's about it. I guess there's a market for "stupid talk" but it's not that big.
[1] Like help me fix my washing machine that won't drain, come up with master narratives for the "polycrisis", talk about why Casey Handmer is wrong about space manufacturing, find papers about the social network of who sleeps with who at a high school, etc.
Did we? Many of us have been saying that the amount of compute going into the models is unsustainable and that the models aren’t improving enough to justify that for over a year. The emperor has no clothes is true yet again.
They also believed they would be able to build that compute without restrictions. Between hardware costs and massive public opposition, scaling as they had anticipated is in jeopardy.
No I don't think there was any systemic underestimation of compute. I see the opposite - every company understands compute is important and tries to get hold of it.
If you invest 100B dollars into compute while asking for 1T, but in reality would need 100T dollars to meaningfully move the needle, you're still significantly underestimating the amount of compute needed, despite not getting as much as you wanted.
Nah they'll hit a ceiling. Can only get so big before things collapse. And besides, they've already churned through the Internet's data. Not much new content left in the wild and patterns in other data forms (audio, image, etc should be pretty low by comparison.
I agree we're at diminishing returns, but when brain scanning gets good enough you have a better dataset than the internet, Facebook is deep into that research.
The failure that is llama4 needs to be studied. Meta was kicking ass with llama3.x and then something happened, something really went wrong. what happened between that time and llama4? I think it happened after llama3.1, llama3.2 was nothing to write home about. We need the gossips, maybe a book
I would absolutely buy that book. Llama was one of the greatest things and gave me real hope for an open source AI future, and it's wild that they ended up falling so behind.
I've heard rumors that it had to do with talent loss, but just rumors.
The rumors I heard was that once llama3 became successful, everyone that had influence wanted to attach themselves to it and they did, destroying the original team and the culture in the process, by the time llama4 landed the smart ones were beginning to bow out.
> Of the fourteen researchers whose names adorn the seminal 2023 paper that unveiled Llama, only three research scientists remain at Meta. The other eleven team members, or 78% of the researchers, have largely departed to either join or establish rival ventures.
for the record, and training scrapers... llama is not open source. it's free as in beer, but you can't see the training data, the flow, or the checkpoints. you get the compiled binary, and only <800M mau.
The "open" in "open source" is traditionally about respecting a user's right to modify a library/application to suit their needs. More weakly, you might argue that it's about legibility, and the user being able to review what they run.
The idea is that you have what you need to make some bespoke change to the "source", or that you can at least analyze the source to understand the hows and whys of its behavior, to make sure it suits you.
> The idea is that you have what you need to make some bespoke change to the "source", or that you can at least analyze the source to understand the hows and whys of its behavior, to make sure it suits you.
> Do weights provide either of those qualities?
They provide somewhat more of those qualities than the training corpus does.
Not a lot, especially for "understanding", but more.
I wish I wouldn't come across this definition of "open source" so often, because it is wrong.
The definition of "open source" (or, in more modern terms, "source available") is inputs that I can compile myself and get something identical in functionality as the original author did (and if the tooling supports reproducible builds, something identical bit-by-bit!).
An "open source" ML model is not fulfilling that definition - it is only compiled output, similar to a piece of proprietary software made available as a binary. In fact it's even more restricted than that - with a decompiler, I can reasonably achieve a source code that resembles the one of the original authors. With an ML model, there is no way of reversing the "training" process.
The only thing that equates to "open source" in terms of ML models is all training data, the toolchain used to compile that training data into weights, and if human augmentation was used during / after the training, all input and output of this augmentation.
But no one of the large players will ever release that. First of all, the training data is heavily contaminated. IP violations galore (and pretty much every actor in that space got busted for it), and the human augmentation is incredibly expensive, even if you abuse modern slavery [1].
Even if you had all that you would get completely different weights out at the end, and you also don't have the resources to "compile" an LLM because the compilation can cost $100M. If you were given the training data but not the weights, would you consider that open source?
That’s not true at all. The weights are the outputs of training. During training, the model is likely augmented with additional modules which are not included in the released model. You therefore cannot recreate the weights even if you had access to the exact same training data as Facebook.
I would love to know that inside story. The whole saga is starting to look like one of the biggest own goals in history - Meta went from being widely respected and considered a peer with leading frontier labs to having no competitive technology. How a company seemingly willfully threw away a leading position in the most valuable tech race of all time should be a business case study, apart from a technology one.
I do have a theory : Llama3.1 marks the point where Zuck got seriously interested and took over the reigns in driving the work. From the minute he started directing things instead of considering the AI work as a quirky side project, things went downhill. He tried to force a huge scale up in Llama4 which didn't work. Then as we know he disbanded the whole team and brought in a new crowd of mercenaries who may or may not have had the technical skills but they came into an organisation in disarray and still driven by Zuck himself who is continually forcing decisions that are not well founded in the science.
All the above is an entirely evidence free fan fiction version of things, but I would be completely unsurprised if it is true.
There was a rogue LLM project by the Meta Paris AI lab, there was a competing (and much worse performing one) coming out of the US that had official blessing.
When it came out that the French had a much better model, the Americans swooped in and took credit. This is was the beginning of llama.
The Frenchies were predictably pissed and left Meta over the next few years, as RSUs vested.
There's a disconnect between measured productivity and "anecdotal" productivity. I love this chart because it also demonstrates one of the most effective ways to increase productivity: simply reducing the workforce.
> it also demonstrates one of the most effective ways to increase productivity: simply reducing the workforce.
You can cut costs and increase productivity by firing everyone else and taking no salary yourself. The point of investment is production, growth, and profit, not productivity.
Output per worker is the formal definition of productivity, but that doesn't mean we should assume fixed output.
Under conditions of scarcity, it's usually beneficial to increase output or to produce different kinds of output. At least, if someone will pay for it.
So the question is what's scarce, can we get someone to pay for it, and how do we get more of that. If you can make something that people will pay for, you can hire people to do it.
Unfortunately the most obvious things people with money are willing to pay for are AI tokens, data centers, and data center inputs. It's unclear how this gets us more of other things we want.
Who cares what Zuckerberg says about AI agents? He is a PHP developer from the early '00s who got lucky with Facebook. He's not an AI scientist or an AI researcher. What authority does he have to speak on the future of AI agents? Morale at his company is at an ATL, and that says more about his leadership skills he'd better off focusing on, otherwise the agents might replace him soon.
I don't like Zuckerberg whatsoever, but there are plenty of reasons you could potentially care about his opinion:
- He could be well-placed to see whether AI agent development is successful since at a high level he oversees much of it, and will be getting detailed metrics and things.
- Whether or not his opinion is correct, tech CEOs are notorious trend-chasers, and his opinion could set the tone for other companies.
- He himself might even be a trend-follower in this case, and is presiding over an early dialing-back of AI-hype.
llama was unintentionally released and has been a flash in the pan for them since. They've failed to capitalize on it at all post-v2. v3 models were a launch mess and quickly surpassed. The only thing it has left in 2026 is name recognition. Honestly their non-LLM things are more important and relevant, like their SAM models. https://ai.meta.com/resources/models-and-libraries/ but none of these seem to be making them serious money or exist and standalone products.
for "the rest of us" building certain kinds of systems on top of it.
as integrated into Meta's platforms it is clearly uncompetitive as a consumer chatbot with the likes of ChatGPT, DeepSeek, Microsoft Copilot, and Gemini.
Like the other chatbots give me useful answers to questions even if they're wrong sometimes. Meta AI, on the other hand, cracks jokes about my foxwork. That's no so bad because sketch comedy is one of the foxwork skill areas (e.g. if I get you to laugh you are under my spell) but it's not useful and edifying the way competitors are.
I played enough video games to be thoroughly steeped in legends about fox spirits from the Sinosphere, almost three years ago I felt I received an invitation that I was a little afraid of. a year and a half ago I read
and found out that people really do it. It took me about 300 days to contract with a community of foxes and started "going out as a fox". At first I didn't want to explain it to people at all but when I botched my explanation and my wife gave me some tough love about "cultural appropriation" I realized I needed a cover story and it is "I go out as a character to do street photography and make people smile"
And it became real. My son didn't believe it until he saw me work: my son saw these two college girls walking down the street and thought "there is no way they would talk to me" and next thing he knows they flagged me down and are asking me questions and I took their picture. I went to fireworks in Groton, NY and the next day in Little York [1] I haven't even parked my car and people I saw the day before are excited to see me.
When I get approached every day it becomes effortless and automatic to make an approach and when I feel like I am getting a 50% take rate I think "I am doing something wrong and I need to regroup", my usual take rate is in the 80-90% range but it feels like 100%. I don't even talk with other street photographers about it on forums because I'm drawing from an entirely different probability distribution.
Thanks to this work, it looks like I will be teaching about magic soon [2] so I was working on a short monologue about blessings on the drive up to Little York which primed me to really respond when people I photographed said simple things like "Enjoy the Fourth of July" and eating soft serve ice cream from a vendor felt like the feast that Martin Prechtel says is the only ritual -- and I am still feeling high from it just like I felt high from going to Litha put on my my pagan friend.
[2] Magic is real, full stop. Science is real too, and there is nothing more scientific than the double-blind test that accounts for the placebo effect which is as I see it: "the patient experiences the treatment as receiving a blessing and feels better" and "the doctor experiences the treatment as giving a blessing and perceives the patient differently"
AR/VR is the next big thing since 2015, the cumulated revenues since then don't even match the yearly revenues they were predicting for 2020 alone. This is going to end up like 3d TVs
Being a CEO of a large corporation doesn't guarantee you the latest tech insight. You will be easily surrounded by yes men who speak what you wanted to hear.
Agreed, all I'm saying is that his perception of the tech -- whether correct or incorrect could potentially either be trend-setting, or could be indicative of trends already in motion.
My thoughts exactly. I think he is well positioned to see edge of ai tech as the money he spent (if allocated well) could give him bit more insights. Although questionoble how much he is still able with his background to understand, I don’t think he is as much engineer as he was used to be
I'd say that about several other CEOs too if their fanboys wouldn't jump down my throat about it. Zuckerberg is just universally unpopular enough that you can get away with it
Is it bizarre? Or is it you surrealist line of thinking here? What's your angle here? Man with power and money == A cornerstone knowledge base if you want know anything about AI? Wtf.
Trump, Musk, Andreessen, Thiel, Biden, and Bankman-Fried all have or had a lot of power, and I would look to none of them if I wanted to get a better understanding of reality. If anything, I expect this kind of skin in the game to be strongly anticorrelated with providing useful takes; these are the people with the greatest incentives to mislead the public.
Even if zero net new insight, he has access to information on most things, not just AI, that the layperson might never have, even in the future. Not intended as praise at all (i'd say this is true for a lot of C-level in a lot of big companies).
It is precisely the opposite of a concrete measure of anything but investor valuation. Investor valuation (or even just asset value, for that matter) doesn't have much weight when figuring out how seriously to take a CEO's words. Revenue at least has a modicum of signal of competency.
All this is happening because many Silicon Valley CEOs are totally disconnected from humanity. I actually dont think Mark even sees his employees as human beings but rather as training data from a future where AI agents will just replace most of his workforce.
Regardless of wether this is right or wrong, and not even getting into the correctness of such claims, the fact is that he fired 10000 employees, from what used to be, from the outside, an engineering first company. And he sent hints to the market that more layoffs would come as agents become better.
The first problem is: it looks we still need human beings, AI is great, but its not as awesome as we initially thought.
The second problem is: all people who can leave are leaving, and those who can't are looking for a job. Humans, we need stability, a steady flow of income, and joy in what we do. By firing people and putting them in permanent observation, for an AI that will replace them, he is destroying any reason anybody would ever want to work for such a company.
Be honest, in the counterfactual where he had said that "AI" agents are the next industrial revolution and there is no future for human skills and so on -- would you have still valued his opinion as being effectively worthless?
> He is a PHP developer from the early '00s who got lucky with Facebook
Nice summary of a good hacker's viewpoint. All his money, power, and other non-technical signals don't really give him the technical authority to weigh on what really makes a technical difference.
That being said, depending on how good information-sharing is at the company, he might have a valuable perspective. But given that morale is low, it's doubtful that engineers feel free to tell the straight unvarnished truth without fear. And in typical large-company fashion, managers above them likely spin things positively as much as possible.
My point was not that Meta has the best AI researchers/engineers.
My point was it should not matter if Zuck is a PHP developer or whatever, instead of an AI researcher, as he employs a lot of people with those exact skills.
I made no comment about how Meta is doing in the AI race, so it’s a bit besides the point IMO, but you did make me smile :)
The cold hard reality is that, despite the attempt at creating a cult of personality around the man, he's far further along his journey of computer science than being a php dev from 2 decades ago. Boiling down his credentials to being simply a dev from a bygone era is hilarious
On that note, does anyone have the authority to speak about AI agents that would satisfy your requirements? It's not exactly quantum chromodynamics, it's glorified automation that costs a lot more and does it poorly.
Those who've made AI their entire personality really do seem to lack the emotional fortitude to deal with someone else having an opinion.
I'm anti Zuck, but it's mostly because he's a lizard man with no morals. You are anti Zuck because he's mentioned that the promise of AI isn't going to the moon as promised. We are not the same.
He's incredibly rich. Americans greatly respect rich people, the richer the more respect. It's apparently a bizarre reinterpretation/misunderstanding of Calvinism.
The respect only goes as far as to listen to what they have to say. We navigate our experience based on the whims of the rich so it helps to know what they are thinking, but it certainly doesn't confer much respect. Most are completely out of touch.
I trolled there, with a kernel of truth. The neutral version is that wealth is a contributor to respect that's unusually strong in the US and I do think that it goes back to Calvinism.
My diagnosis is.. You are working with people. Smart ones at that, who have been working for a very long time in their own, niche, specific way. To assume all engineers will become AI-pilled like you and hop on Claude Code whenever asked is the wrong approach. You can make the greatest tools, but if it doesn't fit in the behaviour and the way of working of the engineer, they will simply discard it.
I see this issue with clients and prospects all the time. A client's team of 5 produces 1 foobar widget in 2 weeks. That's 50 human days spent. Then, someone demonstrates the same thing can be produced (at an equal ot higher level of quality, mind you) with AI in 2 hours. Management might celebrate but teams will continue programming by hand as they always have, now asking ChatGPT about their build tool errors instead of using Stack Overflow.
Handing out tools and telling them it's cool is not enough. You'll need to understand, work with, and guide the engineering teams properly step by step. You'll have to change their behaviour. That does not happen overnight. Unfortunately the present approach is to drop the people who don't perform in the new era of AI-assisted software engineering. That is not the right approach in my eyes.
>2027 you will use a workspace frameworks (Amazon, Microsoft) sensor libraries and world models.
The only people who'll be using Microsoft for anything AI are those whose employer forces them, like with Teams. All their AI offerings are overwhelmingly inferior for anything code related.
If AI was a productivity boon: wouldn't a company employ the same or more employees to capture more market share at a competitive edge. Shedding employees because they are more efficient seems like shooting yourself in the foot to try and stand in the same spot in a race.
AI should have caused a job market boon: because less skilled employees would have been more hirable/useful. That this is not the case leads me to suspect that AI is an excuse to reduce employee count, but not the root cause.
This technology isn't even a decade old. It hasn't even been really useful until the last 3 years or so. Why would you expect it to be transformative already?
yup - its the j curve of productivity phenomenon in action. and we'll see smaller companies reaping benefits faster becasue they can start with AI native development.
No one needs Meta AI. They have missed the boat with chatbots / agents / harnesses like Claude Code.
Can you picture anyone you know stopping using the AI services they currently use for an AI service provided by Meta?
They have such huge opportunities in selling access to their data centres to Anthropic etc, and improving their own ad models for better targeted adds (using their proprietary data and their own infra!), it is maddening to watch them try to make SOTA LLM models and harnesses no one will ever use.
I think that over long developers will desperately be needed to handle AI.
In my experience, within weeks now concepts written in stone get shattered and the next paradigm has to be used in order to max out AI in an development environment.
What is the case for AI? To handle basic work? Augment the work? Add work?
Why I think dev will be in a good spot if they adapt is the simple fact, that while laymen are using ChatGPT etc. every day, this is like driving a Tesla vs a formula 1 car.
If you take ChatGPT away from the laymen, they are helpless with IT. Devs aren't.
AI isn't static, and every turn evolves into complexity, only devs may handle when they adapt to frequent paradigm shifts and go into high level mode.
It will be again the interface between men and machine, laymen and AI. The gap won't close anytime as expected (The programming manager - remember 6 month ago?), but widens more and more.
What I see is that in day to day work many services have arms race with AI updates. The managers are more and more overwhelmed by the workload but how to automate systems is still devs' area to shine.
The business case is still hidden and unclear, but only one aspect is clear to me: low level programming is mostly configuration work now and bug fixing for AI very seldomly now.
Can't think of a better poster child of complete corporate waste that benefits no one whose assets should be seized and redistributed to the masses.
For the amount that Meta wastes on LLM spending you can pay for things like universal childcare, public community college, and providing free lunch to all public students.
If you care about things like money, look up the dollar returns on feeding children during their development or when you tell families they don't have be an economic burden for simply existing.
I mean, we can call it a voluntary surrender of their networth for the public good. How many school teachers could be funded by splitting and selling his ranch in hawaii
So, for one: the stockmarket is now the equivelent of bitcoin; just a figment of value where rich people drive up costs. Just like a car is _invaluable_ to you not because of it's material value but because what it does out strips it's raw goods, facebook is mostly a bunch of tiny bubbles.
So you ask yourself, _if this thing disappeared tomorrow_, what would be the actual loss. It's definitely not it's valuation.
It could certainly not fund Hawaii’s department of education for a single year. And what would you do the next year after selling it and spending the money?
How much of your money do you spend on paying for kids school lunches, paying medical bills of terminally ill kids and paying off the student loans of graduates?
It's very easy to say that someone/some oeganization's wealth should be confiscated, yet I have yet to see those proposing it actually putting any of their own money where their mouth is.
> How much of your money do you spend on paying for kids school lunches, paying medical bills of terminally ill kids and paying off the student loans of graduates?
At least in the society I live all of those are partially paid by me through taxes.
I'm very glad to do it since the existence of kids school lunches, free healthcare (including for the terminally ill), and free universities make my life much better since society as a whole is better off. Even as an immigrant which did not use any of those services, I'm glad to do my part to pay for them, it's just the cost of a good society.
So you're just making the money that you are forced by the government to pay or face going to prison into a virtue.
Do you actually put any of your own money to help support children/sick individuals other than just getting the money forcefully taken from you and being told that it's totally going to the kids/healtchare, while 50% of it gets burned up by government beurocrats?
I specifically said "I gladly pay it", it's not the threat of imprisonment which compels me to pay my taxes. Instead it's all the benefits I see from my taxes around me: education, transportation, public amenities, healthcare, so on and so forth. Each aspect being taken care of improves society, the holistic whole is larger than the sum of its parts, there is synergy when your population is educated, don't need to spend lots in transportation, lives in great well-taken care of urban environments, can access well-maintained public parks, pools, sports facilities, don't need to be afraid of getting sick, etc.
I do actually also spend my own money in monthly charitable donations, including the UNICEF. I think it's a basic prerogative that when you make enough money for living comfortably you should also find charities you trust and support them.
> getting the money forcefully taken from you and being told that it's totally going to the kids/healtchare, while 50% of it gets burned up by government beurocrats?
You don't even know where I live to be able to say what percentage is burnt or spent in bureaucracy. It's unfortunate your view of government seems to be based on an inefficient and ineffective one, perhaps it's your experience (and it's my experience in my home country) but by being blindly ideological about it without ever experiencing a somewhat functioning government you are missing out.
People like Zuck are incredible outliers when it comes to wealth and most people are basically underwater financially. A lot of us do send some money to the food bank or the local arts initiative but the amount left after the bills for living a regular life come due is a lot less for me than a guy with over $200b and no good ideas. That's a crazy order of magnitude difference, a thousand times a thousand times 200k which most people aren't even making.
For all of the pretty gross stuff it seems Bill Gates has been up to, at least he did spend a lot on stuff like mosquito nets and AIDS prevention.
> How much of your money do you spend on paying for kids school lunches, paying medical bills of terminally ill kids and paying off the student loans of graduates?
A marginal rate of 52%. Well, technically we pay for the education rather than student loans here, because it's more cost effective without the middlemen (though we do get the occasional think tanks suggesting we change that).
Zuckerberg is proof you only need to get really lucky once. He's basically flailed at leading facebook ever since, going all in on monumentally dumb bets and erratically making headcount and project decisions, and still he somehow keeps getting richer.
Even if only for how well Instagram and WhatsApp were managed post acquisition, your message doesn’t hold up.
Most large tech companies buy late and struggle to integrate the products and teams they acquire without creating massive damage. He didn’t. Twice.
I am quite far from a fan of what he brought to the world (privacy nightmares), but to say that he’s a proof you only need to get lucky once is quite far off.
also: face tagging in photos, the fb app platform and the like button pioneered concepts that we take for granted today. also, how product is done internally, growth metrics, all borne from facebook.
The like button is an ingenious and insidious invention. It compelled every other external web property to willingly load fb sdk. It’s mind blowing how effective it is.
I am morally against Facebook, but anyone that says zucky was lucky, as a tech person, likely just sour grapes.
We take for granted a lot of things, doesn't mean they meaningfully contribute to our lives, why is face tagging or thumbs up so special? I believe we already had upvotes and reactions on other platforms like early forums and chat platforms. Even if it was the first one to popularize it, it didn't really come about in a vacuum, the Web 2.0 at that time was all about user engagement and interactions.
Zuck is a world class operator. He’s led one of the top richest companies of all time.
I think the world is worse for it but that doesn’t mean it didn’t happen. The list is just to refute the idea that it was one weird trick that gave him the keys to the planet.
and it’s not about him being first or coding it himself, it’s that it happened with him at the helm. that’s just how it goes.
I think his recent fumbles ironically prove your point. How many people could afford to fail so many times in a row (Remember their crypto coin? VR? Now LLMs?) and still be in the big leagues?
IMO the actual prescient call he made early on is to hang on to 60% of the total voting power of the company. I have to imagine there are several points where he'd have been pushed out quite a while ago if he hadn't have done that.
But if you take the same perspective as a VC: 95% of your bets can fail if the other 5% are spectacular.
He made spectacular bets with Instagram and WhatsApp. Most CEOs don’t make spectacular bets because their business doesn’t allow it, meaning the VC analogy doesn’t hold for them.
So it may be popular but they have ultimately failed commercially so far. They just bought an Indian startup and immediately put the founder in charge of WhatsApp.
Only this week they announced WhatsApp would allow usernames rather than rely on phone numbers. All the other competitors apps (signal,telegram etc) did this years ago.
Whatsapp is by far meta biggest potential. If they could turn it into a superApp like WeChat they would make a lot of money but only if that don't destroy the app.
Think that means 95% of bets over a small period of time? (A single fund within the VC) as opposed to the whole VC since inception.
And so not comparable.
VC investing is also a tax write off for rich investors which is offset against their gains elsewhere, wasted company money is a loss for the share buyer.
corporate espionage at global scale to get ahead of existential threats is some pretty insightful leadership, it’s all a matter of what team you’re rooting for.
I guess it's hard to say how much them monitoring their (potential) competitors influenced their decisions but it's definitely _not_ conspiracy nonsense and pretty well documented:
Also, it’s interesting to note that in both cases they where two very lean companies (Instagram had 13 employees and WhatsApp had 55) with pretty a pretty large and engaged user base (30 million for Insta and 450 million for WhatsApp)
Yeah, and he threw a huge bunch of cash to acquire them. pretty easy to acquire. Harder to build. He got lucky with facebook because myspace sucked. Good timing.
MySpace gave users a lot of freedom. That was great in certain scenarios but at it got bigger it meant there was just more noise and the general atmosphere was very messy for lack of a better word. Facebook was very clean by comparison and it felt natural to connect to people who couldn't handle MySpace, people in your family or at work etc. Then eventually it got big enough through that that it just became the place to connect with your cool friends too because "everyone" was there.
So yeah, it's kinda mixed. The core product was cool but it wasn't as suitable for as a replacement for the phone book in the way Facebook was.
Yes and then Facebook did everything they could to subvert granular privacy controls (Myspace was decent, it had a wall but that was about it) which is of paramount importance for people with real safety issues
FB would have been invented. It's just another layer of authentication anyway. Myspace was an interesting experiment for its time
...Comparisons of Tom to Zuck are rife I'm aware but somehow Myspace isn't known for "they trust me, dumb fucks"
A very HN comment... At the time, almost nobody felt spending billions acquiring these companies was "elementary". Instagram was bought so early that spending a billion on a startup was unheard of. And WhatsApp was so expensive that they questioned Zuck's sanity.
You mention two strategic enablers that made these unusual acquisitions seem obvious to FB leadership. So were those two not good calls?
The parent comment is referring to a “VPN” installed by Facebook on users phone that surreptitiously sent telemetry data to Facebook. No one else had that information.
And also even if they never had Onavo you have Zuck furiously backpedaling in an email about I'm not saaaaaaaying we should acquire IG to wipe them out from the landscape of competition, but........
And then you have the WA founder (Acton) famously saying delete facebook
Some acquisitions make the product better or are altruistic. Zuck's war chest isn't one of them
Just because he fails publicly sometimes does not mean he only makes bad decisions. Facebook is still incredibly successful. The market may apply a Zuckerberg discount compared to other tech companies, but that is not the same thing as saying he has been a bad operator.
As a founder who has gone through scaling processes, I am quite impressed at Zuckerberg's ability to remain relatively competent and relevant at the top role throughout the whole journey. The skillset required changes quite a lot in each phase.
There are very few cases where the archetypical young garage-founder has stayed on top in the long-run.
Don't mistake tenure for competence. Meta's stock structure is such that Zuck cannot be forced out by the board or the shareholders regardless of how many mistakes he makes.
HN is so extreme sometimes. I remember when Llama was gaining popularity, everyone here was praising him for releasing open models. Same for other works in this area. Also remember praise for some VR devices.
And it's all back to Facebook, nothing in between. It doesn't look any different from general population influenced by TV and other social media.
> I remember when Llama was gaining popularity, everyone here was praising him for releasing open models. Same for other works in this area. Also remember praise for some VR devices
Neither of which have been great commercial successes. Open source LLM's and VR devices have have been cool products for HN readers, but haven't made much impact on the public or Meta's revenue.
Whilst Instagram and Facebook Ad revenue is coming in it means Zuckerberg could spend billions pivoting the company to country and western themed steak houses and the share holders wouldn't (couldn't) do anything about it.
Just a stab in the dark, but it's likely because FB and meta are wildly successful businesses, and it's almost impossible to attribute that to one "lucky" event at the inception of the business.
If I'm being honest I'm actually kind of shocked at the engagement this comment is getting and the fact it's still visible. Like him or not, if you're capable of being objective, he is doing the opposite of failing...
You know that a lot of acquisitions is equity hire. How do you get engineers to be excited about fixing something low level in Facebook, a hard task; you don't. Better to wrap those problems in green field projects and have engineers excited, and adopt that tech back into Facebook. Google done this forever (DeepMind). Seems like the work of a genius?
Zuckerbergs comment from 2010: «We have not once bought a company for the company. We buy companies to get excellent people.»
Getting a company to IPO is not about getting lucky once - it's making hundreds of good decisions daily. From hiring to product direction to fundraising.
Like or hate the guy you gotta give credit where it's due. (Same for Musk, etc.)
The man has its flaws but to deny that he has been a very good operator is just putting your head in the sand.
- acquired WhatsApp and Instagram and grew them into the most popular apps on the planet
- created an gigantic ad business that enables entire new types of SMBs to exist that rely on Meta to find their customers
- which is pretty much a money volcano
- bought a ton of GPUs right at the inflection of early LLM take-off
None of this was obvious when he started Facebook but obviously well executed in hindsight.
I feel as thought Meta, compared to other tech giants, have a vested interest in saying that AI failed, as they are the only major tech company that has almost unequivocally lost the AI race.
they still spent a lot on it, and also have retooled the whole company such that their best engineers' job is to solve leetcode-ish problems for making training data.
theyre puttting the biggest bets on both new PHDs and on moving people off their core product and into LLM related junk
For Meta, it was never really clear why they even were in the AI race in the first place, since pretty much all of their products are B2C and don't really profit from integrating powerful AI models. And for all of their internal needs, they could easily use models created by someone else, which is orders of magnitude cheaper than trying to compete on building your own SOTA model.
> said a review of a recent data security incident with the company's controversial mouse-tracking software indicated that no employee data was included in AI training.
That's... not quite right. The employee data is used in AI training and is intended to be used this way. But despite not correctly ACLing the data for a couple weeks, it is believed it was not accessed inappropriately.
CEO Mark Zuckerberg recently dispatched a small team at his company to create a smartphone app similar to Polymarket and Kalshi, the New York Times reported on Tuesday, citing two employees with knowledge of the matter.
The app will probably rely on a video game-like points system instead of users wagering money, though the company has not ruled out betting real money eventually, according to the report.
Meta will continue trying to build a platform that they can control. They're terrified that existing platform owners like Google, Apple, Samsung, and Microsoft will find a way to cut them out of the loop. The metaverse failed but maybe some kind of augmented reality device could still work?
They're pushing hard for a unified platform. For WhatsApps new username feature one can only choose from usernames not already used on instagram or facebook.
They're going to try to push those stupid glasses next. Not convinced they'll succeed since it's not just the technology problem they need to solve. How do I find PMF for a product which encourages others to physically assault my users, while also not having it banned in various countries? We're talking about a societal shift that would need to happen and nobody's going to trust Meta again where that stuff's concerned.
I don't want to be a Zuckerberg hater, but is there any thing Zuckerberg has done or said in the last few years that wasn't mundane, reactionary or self-aggrandizing?
In other words, was there a single decision or take he made that turned out in his favor?
So what happened to Meta after those successful llama 3 model releases? They really made competing models back then. If felt like they have right people, strategy and good results. Now it feels they have neither of those…
it is amazing that meta continues to burn money and keep afloat. ads DO really make money, especially when they are going to overtake google in this [1].
to all the critics, i would suggest letting him/them cook. the snowden-era privacy concerns was exactly around ai getting trained or using personal data, and they have a treasure trove of it.
their money dump do sometimes also spend money on r&d unrelated to the current mindhive topics. we never know what the scatterbrain approach may give birth to tomorrow.
do be critical and vocal on how they squeeze money out of other areas or how they increase their revenue by ruining our lives instead.
You'd have thought that Zuck's previous failures to make the things he dreams about (e.g. Metaverse, decent in-house AI) materialize might have made him a bit more cautious about betting the farm on things that don't exist, especially when he's expecting someone else (the AI agent folks) to make it happen!
I suppose you have to admire the conviction: I'll fire my developers today because REAL SOON NOW I'll be able to replace them with AGI!
I can't imagine they'd ever get enough people to opt-in to mouse tracking to generate enough data for their model training. It's either a DOA AI project, employees will be pressured into accepting it (making it part of their review), or their managers will accept the monitoring on their behalf.
If a Meta employee screws up a major project, what happens? What will happen to the executives behind these mass firings and realignment - executives of one of the very top SV companies whose job is dealing with the landscape of disruptive technology development and overreacted to the latest thing? What is the standard for them?
Reminder: the "expected" result is replaced all programmers and white collar employees. Even they don't necessarily state it explicitly, that's what is in their minds.
with coding, you have sort of a framework for doing it right, if you have good specs, good testing practices, strict grounding in expected results deterministically, good linting, etc... this is much easier to automate with AI for the coding part within that assuming you did your homework around it... i don't have experience with all the business layers but it seems a bit more nuanced and fuzzy as you get away from that "harness" of sorts as it doesn't have to work in the same way as code for execution and evaluation... and even if code works, it still needs tastemakers in the final ok. maybe the taste maker ability still needs a lot of work/scale to be feasible, idk, like its still earlier than later on that. maybe Elon already cracked this to an extent given his automation in various companies.
By "AI agent development" does he mean using agents and such for in-house coding? Or does he mean their initiative to train their own agentic model based on the data he's having a bunch of employees generate?
Facebook is already renting out their excess GPU capacity. Nvidia is giving out GPU capacity for investment stakes in startups. OpenAI is trying to get the government to take a 5% stake and needs to make the same amount of revenue within 3 years as Google does now or else they're going to default on their payments to Oracle. It may not be the same as the 'dark fiber' from the Dot-com bubble but it sure seems similar. Especially when this hardware is purchased ahead of delivery and ages so quickly it's not even clear it will be reusable after a few years. It's the scale of the investment in this stuff that is the issue. In 3 years, AI spending is off the charts compared to the railway bubble, Dot-com bubble, Roaring 20s, or really any other bubble. I don't think it's going to end well.
Or: you wasted too much money on failing to replace yourselves so now I have to lay you off. Which is one of the two possible grand outcomes of the AI bubble, which both result in laying people off, because that is all these companies know how to do as a response to stress.
I am not sure that it has to be so zero sum. The AI truth is probably somewhere in the middle; it probably doesn't replace software engineers and it probably won't be deleted as completely useless. My current feeling is that it's a powerful tool I'm happy to pay to use; it doesn't replace me, but it makes it easier to do higher quality work. It feels a lot like IntelliSense, or faster compilers, or getting a 32" monitor. That probably doesn't sustain the bubble, but it's something that people are going to be poking at and making money off of for a long time.
I agree that people are investing as though the world is going to run itself while the ultra-wealthy run off in yachts to compare sizes. If it wasn't AI, it would just be tulips or something. That's just how people are. But maybe they'll be right, who knows.
> The AI truth is probably somewhere in the middle; it probably doesn't replace software engineers and it probably won't be deleted as completely useless.
This is not really somewhere in the middle, I think. It is very close to one of the ends. Because the fear-promise to the idiot-investor class was that it would have those impacts across all industries, not just us nerds. They hate us for refusing to make their silly ideas possible and having irritating fact-based reasons why they can't work, but they don't hate us enough to spend that much money replacing just us. They have lots of other people they hate paying too, and we haven't even made a dent.
I think we can learn from history about where this is headed. I remember when each developer needed a $600 compiler to write C++. Eventually that industry was replaced by free compilers. Kind of a similar thing going on with frontier models; I really like Fable 5, but if I didn't have any money I'd use GLM 5.2 or something. Since my employer does have money to spend on it, I enjoy the productivity boost (just like Intel's C++ compiler produced faster code than gcc for a while), but you can see the direction things are going. Eventually the moat around Anthropic and OpenAI collapses, and AI coding assistance is something built into every text editor.
There is still probably a lot of upside on betting that some lab produces a model that is light years ahead of what we have today. But I feel less "I'd better start planning for a new career" than I did a year ago.
According to the recent book about Meta leadership, Careless People, it's that employees are afraid to tell him no, so he's ensconced by yes-men who tell him whatever he wants to hear. He probably has no grasp of market and product development realities.
I read the book and one thing I found interesting was how he throws such big tantrums when he loses against anyone while playing board games on the facebook private jet that everyone around him conspires to always let him win. Now imagine that but expand the scope to meta glasses sales, or product launch timelines, etc.
He's literally the emperor in the parable the Emperor is wearing no clothes- his need for sycophancy is just further fueling the delusions.
I never noticed that the Emperor was a vain man who couldn't admit his lack of wisdom (because only wise people could "see" the beautiful cloth, and he pretended to be able to see it), and that he was surrounded by yes-men (of course he was surrounded by similarly vain men who had to pretend to be wise to keep their positions, but I didn't notice how this turned them to yes-men).
Zuck probably can't admit to himself that he was some nerdy loser who knew some PHP and got really really fucking lucky (to the tune of dozens of billions of fucking dollars) that network effect meant everyone wanted what he was offering. I'm guessing he thinks those billions must be proof that he's smart... So smart that he's unbeatable at any board game.
> how he throws such big tantrums when he loses against anyone while playing board games on the facebook private jet that everyone around him conspires to always let him win
It's hard to believe that that is a real person and not a fictional person being written against some trope.
there are plenty of orgs where writing more code is not a good thing, in fact it's the last thing they want. but yet these orgs would still employ 80%+ of all developers, not necessarily to write code though.
Investor money being pissed up against the wall, for little purpose? What else is new.
I'm amazed the backers are on board. I wouldn't be. I'd rather invest in dead tech like coal, and soak every last drop of profit, than back this kind of bogus shit which is proving both more expensive, and less workable. Well, maybe not coal, but for goodnes sake, can we cut to the chase where AI investment tanks and move on with our lives?
I'm sorry if it's a non sequitur but I feel even beyond superintelligence/AI/LLM whatever of the last few years... they've always done this, it's always been somewhat hamfisted
Examples abound of "I reported Nazi hate page. Didn't violate community guidelines. I called my friend a jerk, jokingly, got a month ban
For years. Not restricted to when ChatGPT et al arrived on the scene
(Because, AI in theory makes sense. If you want to monitor things at scale you might use AI - however that's defined - to make your workload easier. When is an account being hijacked? When are bad actors infiltrating the system? Or whatever)
I think there are seriously misplaced expectations here. The primary role of AI is transference of effort, while "increased productivity" is just a side-effect (since computers are so much faster than humans at highly repetitive tasks). It's about not having to directly do X anymore (or as often), even though it may take a few rounds to get X to a satisfactory point. But even if following up is needed, most of the effort budget can then be used for Y.
Also those with very heavy investment in AI are looking for bonkers results, which is the cause of their disappointment. They need to reduce their expectations. I for one am loving the results so far.
We're getting closer to the bubble-bursting peak. Yet another tech corp that fired a ton of experienced people hoping to squeeze more $ for a quarter, finding out AI hype is not all it's cracked up to be. Loving that for them.
Second thought, I wonder if zuckburger has turned into the Cramer of tech. First, the metaverse fiasco, now another massive overinvestment into a dead-ish end?
My instinct (for better or worse) is usually contrarian. Most people seem very skeptical of what Meta is doing with AI. But, what if, in a way at least, it makes sense?
Maybe Wang has correctly identified that the programming and agentic ability that Anthropic and OpenAI models have has largely come from armies of software engineers creating massive datasets by writing out coding and agentic problems and solutions?
So he told Zuckerberg that. The reason it may be turning into so much friction is that at companies like Anthropic or OpenAI, training engineers were either hired specifically for that purpose or probably mostly handled through contracts with third parties (which again, hired them to train AI). And honestly many of them may be overseas or just happy to have a job in a difficult period. But anyway they wouldn't have very high salary expectations etc.
But Zuckerberg already had 25000 engineers. Why not take say 1/5 of them and get them working on the the dataset? The problem is that those engineers were hired for different prestigious highly paid positions at Meta/Facebook. They were not hired to do tedious grading of AI answers or quiz construction.
But Zuckerberg either has to do this, or spend additional billions on doing it all with external contractors. A third option would be to try to create a massive distillation operation. Or just hope that his engineers could invent some magical new training trick that manifested the agentic and programming skills without the large scale human input.
Or he could release a model trained largely by existing open weights models. Which without some huge breakthrough probably has no chance of surpassing them, so is pointless.
I think most of the substantive criticism of Zuckerberg has been about burning funds. If he gives up the "your job is to grade AI homework now" plan because his engineers refuse, he would need to go through third parties. The additional billions and billions this would cost would create more pressure on the bottom line and shareholder pressure.
It would also give up any potential advantage that Wang may have optimistically sold the operation as, on that using "real" engineers as opposed to lower paid data labelling engineers might result in a higher quality dataset.
At some point, model architectures that don't need such massive datasets or can be created automatically in a way that advances the frontier will probably come about. But right now it doesn't exist.
Further, the way AI works currently, business advantage from AI comes from encoding existing internal intelligence and knowledge. Meta's massive engineering corp effectively has that in their heads. Having them create these datasets is possibly the only way to leverage this knowledge asset in this paradigm.
I guess the problem is it means forcing thousands of people to do a different job from the one they were hired for.
What's the end goal? Meta-specific engineering, with baked-in knowledge of how FB, Threads, and WhatsApp work? General and/or coding products to compete with Anthropic and OpenAI? Some special Magic Thing which only Meta can invent which will bedazzle Meta's users?
You don't need giant datasets unless you know what you're going to do with them. OpAI and Anthropic are having enough issues making their products profitable. And those are, if not beloved, then at least respected, with a real, if patchy, reputation for usefulness.
What was Meta's pitch in this market? There were hints of interest when LeCun was still doing original R&D, and there was some distant possibility of a next-gen revolutionary product.
But now the goal seems to be to flail around doing something incoherently AI-branded with no obvious strategy.
The troops are being marched around, but no one knows where the battle is supposed to be.
They are making more money than ever before. Maybe Meta leadership doesn’t really care about having a coherent strategy at this point. They can afford to flail around to see if something sticks. Reminds me of Rich kids who have ability to travel the world and find themselves before settling into a career
One problem is that the AI agent market is fiercely competitive. Why build when you can buy? For the foreseeable future there will be a number of competitive models on the "efficient frontier" and I don't think one vendor will pull ahead.
In that market you can build a model and spend a lot of money on it and at best get something that's on the same frontier as everybody else but just as likely end up with uncompetitive models like the ones they have now.
You might save a bit running your own models, doing your own inference, etc. Why not take advantage of "last mover advantage" and buy whatever is best when you need it and figure the odds are good that everybody else is going to buy more GPUs than they need and as a large customer you'll be able to buy in bulk at fire sale prices?
That makes sense in a way, but remember that Meta had previously seen some brief developer glory in the initial Llama release. Going the off-the-shelf route would essentially be giving up on being on the technology frontier in this area, and not monetizing their knowledge assets.
>I think most of the substantive criticism of Zuckerberg has been about burning funds.
I'm not in the org myself I know some Meta SWEs tangentially. My understanding is that the biggest criticism is just the chaos of it all. Jumping constantly from one thing to another like headless chickens and accomplishing nothing.
It created an environment where it's kind of impossible to plan and progress your career.
While I mostly agree with your post, I do want to point out one thing:
> Or he could release a model trained largely by existing open weights models. Which without some huge breakthrough probably has no chance of surpassing them, so is pointless.
This seems to be categorically untrue. Composer 2.5 is a substantial improvement on its underlying Kimi base model.
If that is backed up by benchmarks then maybe they should imitate whatever Cursor did. What did they do?
They may eventually have to do that. Or they might be starting with an existing Llama model. Maybe I should have said "huge breakthrough or additional dataset".
Ahh the classic bootlicker's bias, this is how oppressed developers move and click their sad cursor hoping to keep their job while they train their AI replacement!
>"For people who are comfortable, that's great, they can contribute to this kind of great human survey. To people who are not, it is not an issue,"
Did they really think they could record all their employees screens for a couple months and one-shot the agent thing? This is like junior engineer "let's refactor this monolith" levels of delusion.
That's a great question. Some form of union has already started in some companies, as far as I know. But not many employees have joined those unions. Probably because most employees have high income and don't really feel like they need collective bargaining power, compared to other low-income laborers.
The modern trend is to think intelligence is generative “like compression” or “predicting next in sequence” rather than iteratively reducing uncertainty, like those fault tolerant humans.
Compression can be defined as reducing uncertainty. If you can predict the next sequence you can compress it to 0 bytes using arithmetic coding. Reliable prediction is what enables compression and it's the link between compression and AI that everyone is talking about.
No one ever in comp sci says artificial intelligence is "like compression", they correctly state that "artificial intelligence IS compression". It's absolutely known and accepted that artificial intelligence (defined as predicting outcomes with a measure of certainty and taking chosen actions towards goals using those predictions) has equivalence to compression in a very hard science way. The hardest part of artificial intelligence is compression and the remaining part, the choice of actions based on predictions is just a tree search to a goal.
Compression in image, video, sound, and text. These items to compressed are all created by humans and we will say represented by files. The difference between an instant of reality and the files is vast. Reality also doesn’t stand still and each instant needs to be captured and interpreted before AI happens.
AI can be just like compression but currently the compute power is no match for details.
Finally these reality details need consideration in any successful implementation. Which means the implementator needs to be aware of the details and successfully relate them to everything else in the model.
I think anyone surprised by these things is not fully engaged with what they are doing.
The factor that is missing in that analysis to me is a time based dynamic stability perspective. Humans have a pretty good ability to go off the rails in reasoning one day and wake up reasonable; a pretty good ability to pursue tasks, despite a multitude of distractions, for ten years or longer. The best models get appreciably worse over a half million tokens. Even using a bunch of limited context agents over time, they lack mental stability. They keep coming up with ideas contrary to the long term idea, and every so often generate ideas that make no sense but they have a hard time letting go of. So the pure functional LLM is compression, but AGI needs some centering process, some high level of dynamic stability to stay sane over time and in the face of 10,000 shiny pretty things to chase.
The harnesses get better, but I haven’t seen much experimentation on long term stability, at least since the “let the LLM run the candy machine” papers from a while ago.
Because the thing missing, even with the largest agentic swarms, is independent intelligence, where it’s given something to own, like say “end to end data quality as we add more clients” (for a SaaS) and it just figures out what that means at each time, mutating its role and solutions to fix the external world, without getting silly.
the things as developer i see no practical gain for us when we should and found model routing better than keep pouring investment into agentic model like, i rather use a smaller faster model with harnessing to do drama script generation instead of leaving it to Fable or Gpt5.5
Mark is really a bad leader with a mwah mwah vision. He is maybe correct in some things. But the execution is really really poor. Plus he does not have followers and believers. He only got money that can simulate followers to a certain extent
If it was still possible to get verbatim results from Google then I believe "mwah mwah vision" would have been an authentic Googlewhack pointing at this comment thread.
Actually, agents have been developing incredibly fast. It is just that Zuckerberg has some unrealistic expectations; the man is a lunatic.
Over the past six months or so, OpenAI's internal team has completely shifted from being heavy ChatGPT users to using Codex. Once you start using an agent like Codex, it is very hard to go back.This shift is truly transformative.
I am also aware that some of the consumer agent products on the market are growing very rapidly, such as Manus and GenSpark. Not to mention Claude Code and Codex.
I was worried this time last year that by this time this year, companies would have slashed their engineering teams down to a handful and everything would be driven by mostly autonomous agents with human guidance. But it just hasn't happened. Do I write all my code with an agent now? Yes. Can you just give an agent a desired outcome and let it work, unsupervised? Absolutely not. I can produce more code than I used to, but if I want it to be good, to be stable, to do what the product manager and designers want, it's only about 2 to 3 times more code than before. And that productivity is impacted by the fact that I'm reviewing 2 to 3 times more code than before (and you have to review, even more so now than before, because if you just let opus or gpt 5 do its thing, you'll get some terrible results, and I've found a lot of engineers on my team are just letting it do it's thing without a lot of iteration).
The thing is: I could produce 2-3 times as much code as before _without_ an LLM, if I didn't care about my colleagues' ability to review my output properly.
Lines of code are a liability, not an asset. You want as few of them as you can get away with, without compromising the actual asset: the functionality.
A huge part of the job of Software Engineering is producing the right amount of code at the right time.
> Lines of code are a liability, not an asset. You want as few of them as you can get away with, without compromising the actual asset: the functionality.
> A huge part of the job of Software Engineering is producing the right amount of code at the right time.
Absolutely true, however my experience says that the correlation between "good software engineering practices" and "positive business outcomes" is, at best, small.
120 kloc mostly from one single developer copy-pasting and keeping non-compilable code for an obsolete target "for reference" for a decade, becoming both a ball of mud and a whole pantheon of god classes? No unit tests, no code review? Won awards.
Properly engineered, mandatory code review, mandatory unit tests, dev meetings to knowledge-share? People with the money said too slow, closed it down.
(Sometimes people bring up how bad Musk's code was at PayPal. I never bothered investigating. Successful product though, wasn't it?)
Survivorship bias, you don’t know of all the failed projects that couldn’t get off the ground because of incompetent development team and practices that lead a product to its demise, or a product that is possible within constraints that otherwise could have been a success, but not realised by sloppy work and incompetence.
Furthermore the dependencies you choose to build your product are presumably filtered for engineering practices or world class engineers. So given the choice you yourself prefer top quality engineering, so do your customers. Much in the same way you are a customer of your projects dependencies. Difference being, as developers we get to see how the sausage is made, our customers only see second and third order effects.
> you don’t know of all the failed projects that couldn’t get off the ground because of incompetent development team and practices that lead a product to its demise
Trying to ignore the nuance is hard in your position or the following one I’ll give is difficult.. but is the opposite potentially true as well? We don’t know how many projects failed because of over optimizing, too much time spent on design and engineering decisions. It’s of getting out and MVP to market. I only say this because I have been apart of a few of these.
Over optimizing and spending too much time on engineering decisions is also something an incompetent development team would do.
Well, that's the other side of incompetence: they know how to spin their tools, but they don't know when to stop, or how to stop the change requests, the balance between shippable, maintainable, and what the market wants at that time.
I understand and of course I am familiar with the hypothetical you are trying to set up here but I was specifically pointing out a logical fallacy I see banded round all the time by people who should know better or educate themselves.
I will say that if “good engineering practices” comes up in your root cause analysis for a failure to launch a product you are not thinking critically.
> Survivorship bias
The statistical problem is small sample size, not survivorship bias, as I got to see things before failure. These two examples are merely illustrative of things I've seen.
I think both are at play here and I don’t know about Musk’s programming skills. But it seems that they had other very good programmers (including levchin). So maybe business success can buy good engineers to clean the messed up code. I’m not sure how it goes with AI now though.
> Absolutely true, however my experience says that the correlation between "good software engineering practices" and "positive business outcomes" is, at best, small.
One of the most uncomfortable truths about our profession is that there is no floor to how bad software can be while still making people billions of dollars.
The first thing they buy of the success money though, is a struggling technical competitors, so that this team can clean up there mess. This would mean that AI is only a good contributor at startups and with prototypes.
Good point. In my experience even hobby free software projects are generally better engineered than most proprietary software sold by businesses
There's a trade-off to be made, and it's not necessarily clear where the trade-off sits for any particular company, or even team within the company.
One product person described it as eating vs breathing. Availability is like breathing: if you stop being available (including, but not limited to, because your software is a big ball of mud) then you're going to die pretty quickly. Product is like eating: you might not die so quickly but if no-one's buying what you're selling then you're still not going to survive.
The team I'm part of is a platform team, so we're closer to being lungs than being stomach. We can (and I appreciate being able to) focus more on stability than feature development.
I'd argue PayPal was a successful business, not a successful piece of software.
That's kind of the point.
The software only had to be good enough to support the business. It doesn't have to win a Turing Award, and probably wouldn't help the business if it did.
> Successful product though, wasn't it
If you can call being an “also ran” in a field they had a ten year march on their competitors in success, yeah.
Truth be told it was the shoddy code they were forced to use for the vanity of their paymaster might well have held them back, though manifestly that is not a bad thing. Probably the best outcome, really.
Lines of code are a business liability. They are future cost.
> Lines of code are a liability, not an asset.
I have been saying this for years, I once had a heated argument about a small system of maybe 1000 lines of code that was technically superior and more scalable but was freaking 1000 lines of code to maintain compared to the quick and dirty 10 lines of code it was suppose to abstract and make generic (for future use of course).
That with also countless debates over insignificant features in frontend apps at the cost of extra code. Frontend code is very susceptible to this maintenance cost dilema.
Many developers are too focused on delivery value compared to maintenance cost. It is unfortunate that non-technical management can see value delivered, but not maintenance cost incurred. With LLM-assisted code this has become many times worse.
>Many developers are too focused on delivery value compared to maintenance cost.
The way this is worded feels like it leaves the blame on developers. Aren't these developers focused on exactly what they are being judged by? Shouldn't we say it is the management who is too focused on delivery value compared to maintenance cost? Is it the developer's job to guide management or the manager's job to request guidance, assuming such guidance is needed?
We need to make sure the responsibility to resolve this problem falls on those with the power to act on it, and in this, developers tend to be receiving far more responsibility to fix than power to fix.
I don't think so, my experience is that most often falls into two scenarios:
1) The devs pushing for more complex solutions, covering obscure edge case scenarios, feature-creep, "future-architecturing" because they are more interesting to implement. Classic over-engineering problems.
2) The features the managers actually want are usually boring or annoying to implement and the devs just work around any big architectural problems caused by the feature delivery.
1 is 100% on the devs, 2 it varies wildly, the willingness to address architectural problems are often under pressure by time-delivery estimates from managers. But many devs (especially in companies with low morale) will often just work around issues because addressing the underlying problems can be very difficult and/or time consuming.
Meaning either the dev wants to do the right thing but doesn't have the time, or the dev doesn't care enough and just pushes the tech debt down to the future (when hopefully they will be at another job).
LLMs makes both problems significantly worse, although they are also often very helpful with the big restructurings mentioned in 2. The dev can still be lazy and the deadline can still be too tight even with that extra LLM help.
How is covering obscure edge case scenarios overengineering?
I don’t think this is a great argument for such a small amount of LoC. 1000 lines depending on the service it provides could be very small.
> more scalable but was freaking 100000 lines of code to maintain compared to the quick and dirty 1000 lines of code...
I wonder though if, as long as you have LLMs to maintain the extra code, it's worth it to gain the new feature. Less tech debt than your intuition expects.
Why would an LLM be any more capable of maintaining the extra code than I am?
Because it can read and understand the code base far beyond your ability.
You can pretend that's not true, but it is. And it's only going to get better.
Sure man, knock yourself out.
"Lines of code are a liability, not an asset."
wrong framing imo
EXCESS lines of code are a liability.
Code of course is an asset.. what other real asset is producing cash-flows? lol come on.
It's apples and oranges. Lines of code is just a measurement. You can't look at a whole codebase and say "these X lines are essential, those Y lines are excess, delete those" -- the lines interact with each other to create the whole thing.
It would be like saying "Weight in kilograms of course is an asset for an airplane. That's what keeps it in the air. What other real asset is the airplane made of?"
This is the thing. You "spend" lines of code, you dont produce it. The produced part is the outcome - a functional feature, stability improvements, some business outcome. Measuring productivity with LoCs is like measuring output with cash burn.
> Measuring productivity with LoCs is like measuring output with cash burn.
companies are doing that as well lol (re: tokenmaxxing)
I think that coding reviews are no longer feasible as they used to.
The pace and expectations have increased and a human can barely cope with reviewing its own code, let alone colleagues'.
They are not going to disappear in critical aspects of a codebase, nor shouldn't, but the industry will eventually reward self sufficient individuals able to keep the pace, harness and run adversarial reviews against the design and implementation autonomously.
I'll also say the harsh truth. A well implemented adversarial flow will do either better than your peers or will deliver 95% of the value at a fraction of the cost.
The industry has never valued product, let alone code quality except in places they are core to the business.
Otherwise you would not have MIT-bred leetcode ninjas writing react/tailwind bugged monstrosities at half a million/year for billion dollar products.
How'd that go for the most recent Windows update?
[flagged]
> Damn you folks are legit stupid.
You're definitely out of the line with such language.
And naive.
Your biggest error is thinking that:
1. there's quality software out there. It's definitely far from the industry standard, even in high budget big techs, far from it
2. that people opening the wallet care about software quality. good has been always better than perfect
3. that the average PR review has significant impact on code quality, clearly false by point 1.
4. that proper AI usage for doing adversarial reviews of plans and implementations won't catch by far more issues than the average pr out there nit picking coding styles but not even bothering to test the feature/branch
Thus I stand to my opinion. Quality PRs are too expensive nowadays and the engineers that will thrive will be the already great ones that will be enhanced by the tool, not companies shelling money on reviewing the average garbage MIT-bred leetcode blackbelt is slopping.
They will only involve core/critical code and libraries.
Don't bother my man. All of the commenters talk about the "bad code" AI writes as if the code that's out there in the world isn't also compete garbage.
Of course, everyone here is in the 99th percentile in ability, am I right?
The sheer number of people I'm meeting that say about trivial things "AI cant do that" is astonishing.
Oh well.
In all of that I'm not saying good engineering practices are bad, I'm speculating that the economics will push towards strong individuals soloing features at good code quality.
Bogging down strong contributors into reviewing code, now that there's more work to review in the first place might be not economically feasible unless the code is critical to the business.
"A huge part of the job of Software Engineering is producing the right amount of code at the right time."
But the whole job of the owners is to lower the costs as much as possible to produce the product, especially in an environment with higher costs to borrow (higher interest rates at the fed)
> A huge part of the job of Software Engineering is producing the right amount of code at the right time.
I'd go further and say that usually the goal is to use as little code as possible without sacrificing readability.
Brevity is compression, and compression surfaces the salient points of a problem.
Elegance often comes down to brevity.
This is one of the things I ran into early on. LLM needs to compute the determinant of a matrix? Sure, just spit out some huge hyper optimized implementation of it. Good luck maintaining that. Slapping "use industry standard open source libraries for common functions" has improved the quality of LLM output for me by such a large margin.
So it's like jazz... it's the code you didn't commit that matters?
>I was worried this time last year that by this time this year, companies would have slashed their engineering teams down to a handful and everything would be driven by mostly autonomous agents with human guidance. But it just hasn't happened.
I find this somewhat puzzling. I thought things were moving quickly, but at this time last year I couldn't even get Claude (using Cursor) to spin me up a service skeleton that would compile, let alone do anything meaningful.
I know it feels like a long time somehow, but it was only between November and February that things started to actually somewhat work without significant hand holding. Even now, it seems like we're still figuring out how to fully leverage the current models and tooling, even in organizations that have largely gotten on board.
It's not all that surprising that people were worried and believed this. The AI companies and infrastructure companies partnering with them have spent a lot of money and time trying to convince people this is the case year after year. The critical clue people miss is that everyone claiming that has very clear financial incentives to convince people that's the case even when they know it isn't. Anyone who was actually building with LLMs and judging for themselves based on its performance knew fully well that wasn't the case year after year.
I've said this before: if anthropic (et al) thought they genuinely had a shot at replacing even 30% of white collar work, they would ABSOLUTELY NOT warn ANYONE. They would do what oil, leaded gas, and cigarette companies did. Swear under oath this is completely safe, commit GRIEVOUS societal harm that you explicitly promised wouldn't happen, and then end up in history books instead of jail for reasons beyond my ability to fathom.
No. The very fact they are trying to "warn" us means it's all marketing.
This has been corroborated for me on the engineering front that I can't find a single IC I respect who actually thought there was any evidence AI was going to live up to the hype. I saw a lot of people I always thought were idiots/sycophants/brown nosers go insane with AI. Never saw anyone id trust to help me cross a street blindfolded say more that "I may be wrong, but I'm not seeing any evidence yet".
Fwiw , you're conflating multiple things and consequently drawing premature conclusions.
It can be massively over hyped for it's current capacity and decimate the white collar work.
A lot of the difference of opinion is down to their point of view. At my dayjob, LLMs will not live up to anything because the enterprise is not structured to take advantage of it's strength. That's unlikely to change within the foreseeable future.
I strongly suspect you mostly talked with people coming from just such a background, because it's hard to go beyond our own bubbles
Sure, naturally. And yet parent commenter is remarking that simultaneously no AI-true-believer startups have supplanted the old money, and simultaneously despite much talk the bigcos have not slashed headcount to tiny AI-powered teams.
Hmm, isn't that just down to the fact people always vastly underestimate the amount of work software takes?
It takes a lot of time to successfully deliver a product. Even at the more "extreme" end of expectation - like saying it's a X10 multiplier (I'd disagree on that, it's more like 0.5-3 - depending on the type of work you're working on) you'd still need multiple years to go from first line of code to displacing established players. Things just don't change that fast.
Depends on the bigco IMO. I'm not sure what kind of layoff numbers for Year Three of genAI would grab your attention, but I assure you, hundreds of thousands have mine.
The layoffs started years before LLMs became practical coding tools.
So you're saying companies correctly estimated the need to cut staff to remain competitive and did so -- or at least, some state of affairs empirically indistinguishable from this. Not much of an argument, sorry to say
Yes, this... all the hype from the leading AI companies just pattern-matched so many past cases where things didn't pan out. really giving me the bad vibes..
>The critical clue people miss is that everyone claiming that has very clear financial incentives to convince people that's the case even when they know it isn't.
Social media is flooded with bots pushing this narrative - coding is dead, engineers are all cooked, the latest model is scarily good, "what am I for?", etc.
A good rule of thumb is that if it's a human being and not a bot, they'll use the word "slop" at some point.
coding harnesses improvements mattered more than llm improvements this past year. You could solve problems on claude sonnet on claude code that you couldn't solve a year prior.
Nope. If you use pi and its extremely minimal harness you can see a _lot_ of difference in models.
> at this time last year I couldn't even get Claude (using Cursor) to spin me up a service skeleton that would compile, let alone do anything meaningful
I've been using it to do this for 2 years now. And many people with me. The change you mention is one of is primarily one of Overton windows, of vibes.
>I've been using it to do this for 2 years now. And many people with me. The change you mention is one of is primarily one of Overton windows, of vibes.
I suspect this may have depended on the specific framework. I quite literally could not get Claude (in Cursor) to give me a basic Micronaut setup in a fresh workspace with essentially a "hello, world" API. I would guess that if you're using something like Python and FastAPI, it might have been an easier task or better represented in the data.
The difference that I observed in the Opus 4.5 era is that Claude could take a service framework it has never seen before (proprietary corp) and figure it out.
Which harness software were you using for this 2 years ago? VS Code Copilot? Cursor?
Aider for me
Very successful by just being careful and walking it forward.
Yes its about 2 years, August 2024 from git it looks like.
I've had success with Aider from October of 2025 (from my git logs). I have to differ from the general sentiment over here. The tools and models are good enough to make a change in the way work is done. Of course, Big Tech. will move fast and make mistakes but this is qualitatively different from other recent hypes (e.g. crypto/NFT). It has had genuine impact on people and their lives and a lot of simpler jobs with larger teams have been replaced by these models and one operator.
Exactly the same for me. Sonnet 3.5+Aider made it possible.
I got downvoted though, apparently you and me are liars, haha. Easier to believe that than admit that their take of "I adopted it exactly when it became viable, now it's good, and before that it was a waste of time" was wrong. They're the people who were loudly claiming here last summer it was useless, asking "show me anything useful that has been coded using LLMs". Come to think of it, it has now been a few months since I saw one of those. Used to be every other thread.
Exact same for me as WhatIsDukka.
Yeah, I strongly remember getting Sonnet 3.5 with Aider to boostrap an (albeit basic) project and getting it to work. Especially vivid because I told my roommate at the time about it and he also tried it out and was also shocked. I'd put maybe $20 in credits into the API haha. Feels so quaint, it's almost a foregone conclusion with the current models and harnesses.
Do I write all my code with an agent now? Yes. Can you just give an agent a desired outcome and let it work, unsupervised? Absolutely not.
I strongly suspect that developers moving from writing code to managing agents to write code for them is very similar to developers moving into leadership and management roles and managing ICs to write code for them.
Some devs just 'get it' and thrive, leading a team really well and building a great culture. But a lot of them don't, especially if they don't get the support necessary to understand what changes when you move from IC to manager. If the team (or agent swarm) isn't performing well it often isn't a problem with them. It's a problem with the new manager still trying to stay on top of everything and micromanaging all the things. Alternatively, the new manager is completely hands off and only appears at a check-in point (one-to-one, agent completes a task, etc) where they crap on the work and get cross.
I have no evidence for this, but I'd guess that putting developers through some sort of management training would make them much better at using agentic swarms.
Absurd.
We are beset on all sides with companies declaring agentic coding a failure and here you are stating as a matter of fact some teams “thrive” with this probabilistic expensive approach to approximating working code?
All the while concluding with “I have no evidence for any of this”.
Agentic coding is absolutely not a failure. It's just not the 10x that CEOs really wanted it to be.
We learned that some tasks don't really benefit from AI while others do. My team went from 7 people to 2 (went to new teams, no layoffs), and we're doing the same amount if not more work than we used to.
Is it more draining and lacking of focused work? Yes. Is it more money for the business? Yes.
Friend, you are commenting on a thread where one of the most prominent figures in tech (rightly or wrongly) is saying something did not meet expectations. As far as we can tell this is a man with every incentive to exaggerate and boost these products.
In my world, when something is expensive and doesn’t meet expectations it called a failure. Especially when something has been as hyped, scrutinized, defended and attacked as vibe coding.
Honestly, if you are the director of robotics at a firm I think it’s time you took a cold shower.
Why are you assuming Zuck's expectations for progress of _his_ product have anything to do with the impact of that category of product for its users?
I’m sorry, what’s the question? Genuinely not sure.
I’ll rephrase, Zuckerberg would certainly enjoy agentic coding to be a wild success because it means less staff and more products he could fail to create.
Again, Zuck's success criteria don't necessarily align with that of others. In fact, I'd expect his success criteria are substantially different to most.
Making money? Seems to be the common success criteria for all.
It can both be a wild success and not what he expected.
These two not mutually exclusive?
“Wild success” and “going slower than expected”?
Wake me up when words have a meaning again.
> These two not mutually exclusive?
Correct, they are not.
Consider: Someone who "expects" their bank account to have $100M in it before they turn 30 and "only" gets it to $10M.
From the point of view of a normal sane person, they are experiencing "wild success", and yet at the same time they are definitely "going slower than [they] expected".
[flagged]
If you don't like a personal scale example, just replace the M with a B. Or even a T, given the SpaceX IPO.
The only point here is "not as fast as expected" can still be a lot.
Are you sure you are not the troll? Your rhetoric is highly aggressive.
Pretty sure Zuck was talking about agents in general, not specifically Meta agents.
More importantly, he's really probably talking about 'superintelligence', rather than just building genuinely useful and monetizable models.
Why would Zuckerberg want to boost their AI, they have none to speak of after Llama really? Zuckerberg actually has every expectation to downplay AI so as to save face that Llama failed compared to frontier models.
[flagged]
The real article is at https://finance.yahoo.com/technology/ai/articles/exclusive-z... and he indeed is talking about his company's in house agentic development products, regardless of what you specifically said, as we are discussing what he said.
Who said the layoffs were a success? It's a short term fix (or correction from pandemic hiring) that may still nevertheless have long term consequences.
On the other side of this you have two companies growing revenue literally the fastest ever in any market segment, just for said tech. Somebody is spending this money and thinks it's worth it. Let's check back in six months.
Sure thing, just like I’ve been checking back on people over the last 3 years. I’ll hit you up too.
Past performance is not indicative of future results.
We're riding an exponential here for Pete's sake.
These all sound like nice sound bites.
Until you factor in many large firms are interested in cheap Chinese models.
Are you a frontier lab booster by any chance?
We all know the future results of a lame horse.
Judging by his account he’s a big time booster suffering with psychosis
not sure what this means. should I be worried?
I dont care about your well being TBH
just pointing out the obvious.
you care enough to respond, shouldn't have bothered.
> Agentic coding is absolutely not a failure. It's just not the 10x that CEOs really wanted it to be.
It’s absolutely 10x faster for coding. But coding is only 10% of my job. The other 90% is figuring out what to code.
From my experience with daily stand-ups, I think they can be a significant boost to that, too. Though you will absolutely have to wear a manager hat with their estimates and breakdowns, not just fire-and-forget, as they're often as not wildly over-optimistic about task complexity.
Being a "failure" means it fails to meet its success criteria.
If you claim 10x and deliver 3x, that is a failure. The 3x may still be impressive or a gamechanger or ..., but it still falls short of its promises.
10x is a marketing gimmick in development, AI and pretty much everywhere
So what? Who cares about such a binary claims?
People making decisions.
Pay me 8x to get 10x, great. Pay me 8x to get 3x, nope.
I have not seen reasonable claims that people are paying 8x the salary of a SWE to get those 2-3x results.
Directionally, that's why companies are getting rid of token leaderboards and imposing limits on LLM costs. There's a diminishing marginal return to tokens
That's true and it's also very easy to game.
There is increase in (not only perceived) value _somewhere_ - IMO depends on the organizational culture. At my work I found going over 100$ on OpenAI/Anthropic's API pricing does not produce any meaningful additional output. It might be much different in different companies.
I am also leading a small team of myself and 2 others and we're getting a lot done. The short lines of communication of being a small group + the power of agents has been great for us.
> Is it more draining and lacking of focused work?
To be completely honest, I’m living life right now. I love programming with my bare hands, but man I’m living just building a gajillion things a mile a minute with LLMs. I then come home and spend hours building stuff for myself using local models. I’ve never been so excited about just building shit, that I sometimes want to pull all nighters because I’ve been in the zone (a for work and at home).
Draining? Sorry… inject that LLM serum right into my veins
> while others do
Now we just need to find those tasks. I want to believe.
> and we're doing the same amount if not more work than we used to
Zero evidence for this. It's programmers self-reporting their own productivity. (Have we not learned this lesson after 50 years of programming practice?)
Because this companies defined the goal to replace humans with AI what didn't happen. What happened is that the humans can work faster and have more coffee breaks while the AI iterates for the next review and iteration round by that human.
And the human will not bother to review that mass of code and will just approve so the real test will happen in production.
why would companies who build tools for developers say they'd want to replace humans with AI? it was never the goal, it was never stated like that. they said that by the end of 2026 most of the code being output would be generated by LLMs, and is pretty much true.
are we tho? i don't really see anyone going back to planning and coding by hand, that ship has sailed and it's not coming back. people ditching agentic workflows altogether would be failure, people figuring out it's not a miracle tool and has uses for which it's not as productive is correction.
The bit I don't have evidence for is whether or not teaching ICs to be managers would improve how they use agentic AI. I have plenty of evidence for the efficacy and effectiveness of AI itself (although not qualitative or obviously causal unfortunately.)
> plenty of evidence … although not qualitative or obviously causal
Those two things are the opposite of each other (evidence, but only anecdotally; you cant be both).
Anyway.
More tangible to your argument; what is your argument that this will be more effective than just prompt engineering?
Ive long believed that prompt engineering is a losers game; if there is a trivial set of tricks that improve the output, they will simply be automatically applied.
We see this playing out with the system prompts in coding agents and image gen.
The value of learning “photo realistic studio lighting…” was non existent. The nano banana api is capable of taking a naive prompt and expanding it with these tricks.
People who devoted themselves to learning these “magical incantations” wasted their time and effort; and it was obvious, from the beginning this would be true.
Now.
With managing agents; if a trivial set of management tricks can drastically improve the results, why are you better off learning them now, rather than waiting for them to be baked into cursor/codex/claude in easy mode?
What makes you believe this is a valuable investment in time and effort?
Even if we accept that right now assigning personas to agents and managing them as a manager yields good results, the horizon for change right now is so short, it seems extraordinary to suggest mass management and leadership training for engineers.
We should just wait and see.
All in investments like this would just be tokenmaxing in a funny hat.
[flagged]
> Some devs just 'get it' and thrive, leading a team really well and building a great culture.
I don't think it's this because the outcome you get from AI isn't controllable. You can give it the best prompts and design suggestions and it'll still give you completely wrong or horribly written code.
If you were a manager and one of your reports kept producing completely wrong and horribly written code that other folks on the team keep bringing up as problematic in PR reviews or privately, that developer would eventually be fired for someone better.
But in the AI case, there is no replacement because all of the LLMs have severe problems.
> because the outcome you get from AI isn't controllable. You can give it the best prompts and design suggestions and it'll still give you completely wrong or horribly written code.
I don’t have a dog in this fight but it seems you’re not accounting for iteration and feedback. A horse will veer off of a road if not occasionally nudged to stay on it, but is useful transportation, nevertheless.
> I don’t have a dog in this fight but it seems you’re not accounting for iteration and feedback
You can provide AI official sources to look at and dozens of prompts. I've lost track of the number of times where it didn't arrive at the right answer with tons of opportunities to correct itself based on feedback.
Just an endless of sea of "you're absolutely right to have brought that up, I didn't think about that" and other phrases it constantly uses when it fails to provide a solution. Fast forward 20 minutes later and it starts providing the same nonsense it did at the beginning because it forgot what it already said.
The code solutions it provides are so consistently bad but it's not limited to code. I recently tried a YouTube feature where it can generate AI thumbnails from your video. The results were really lackluster. It completely ignored my feedback like "use a real webcam photo of me that you see in the video", to which the AI recreated a completely different looking human that wasn't me. It even swapped out my real glasses with a rendering of glasses I don't have and kept on making incorrect assumptions about everything. After about 10 prompts and 20 minutes of waiting for thumbnails I gave up, it was really poor.
None of that supports the claim that it "isn't controllable," though. A curious mind should probably find it interesting that it can be fallible in those ways yet still be useful for producing work, and ask how both can be true at the same time.
>I don't think it's this because the outcome you get from AI isn't controllable. You can give it the best prompts and design suggestions and it'll still give you completely wrong or horribly written code.
sorry if it's not the case but i have the feeling you still think AI coding involves talking to a chatbox and copypasting the answers
Yes, an agent does feel like a very eager and very knowledgeable but also very clueless junior developer sometimes. But no, working with agents is not like leading a team of developers. You don't have to make sure an agent stays motivated, give it the feeling that its work is valued (I can let an agent spend some time building a prototype and then decide not to use it, if I did the same thing with something a junior developer took a week to build, that probably wouldn't be a good idea) etc.
> If the team (or agent swarm) isn't performing well it often isn't a problem with them. It's a problem with the new manager still trying to stay on top of everything and micromanaging all the things.
You see problems in the results? Simple, don’t check the results!
Let codex review Claude output and the contrary. Win for everyone involved, more code, more features, more token usage, promotions, code to be debug by juniors and the cycle of life can continue.
> I strongly suspect that developers moving from writing code to managing agents to write code for them is very similar to developers moving into leadership and management roles and managing ICs to write code for them.
I disagree.
From ICs that I lead, I expect that they learn fast. On the other hand, LLMs are basically incapable of improving.
Another issue that I can see is that I don't particularly like eager new colleagues who come up with (hallucinate) wrong answers. At the beginning, if you are uncertain, learn, and if you have questions that learning does not answer by itself, ask questions. But strongly avoid hallucinating answers. New colleagues can be taught that, LLMs not so.
> I don't particularly like eager new colleagues who come up with (hallucinate) wrong answers.
People are most likely to come up with suggestions and ideas earlier rather than later.
Often they’re not learning what is “correct” but just how to “fit in”. A fresh perspective can be good even if flawed. It can help others spark new ideas and think outside the box.
New colleagues can be taught that, LLMs not so.
LLMs can be 'taught' though. You can give them additional context or instructions. The difference is that they can't really teach themselves.
This is roughly what I'm saying - someone who's managing an IC can steer them on the right course, and someone who's managing an agent can also steer it on the right course, so teaching someone how to handle ICs well gives them skills that are also applicable to handling agents well.
It's not perfectly analogous obviously because ICs are people and need to be managed as people, but I really think the skills are quite transferable in one direction. I'll add that I don't think someone learning to manage agents would necessarily become a good people manager.
> LLMs can be 'taught' though. You can give them additional context or instructions.
That's not teaching. Giving someone new information in the moment does not amount to giving them new long-term capabilities.
> Some devs just 'get it' and thrive, leading a team really well and building a great culture.
I think what makes a dev well suited for AI isn't the same as managing a team. What really helped me get productive is having to write a lot of user stories and acceptance criteria with the wisdom of being a dev tasked with implementing them. Also, being on the refinement calls, answering questions, and updating requirements/AC is good feedback for authoring better requirements. If you're good at authoring requirements, checking output, and communicating corrections concisely then you can get the LLMs to sing.
I get why you think this but you are incorrect. Why? Because managing a team of people is very very different from managing outsourced contractors, and LLMs are much more like the latter than the former.
"I strongly suspect that developers moving from writing code to managing agents to write code for them is very similar to developers moving into leadership and management roles and managing ICs to write code for them."
I doubt that. Management is mostly dealing with people, the actual "management"* part is not where developers moving into management roles typically fail, it's the people part. With agents you have the management part without the people part.
I think you are confusing manager with ICs. Managers don't really read or review the codes. What you are describing is where agents are doing all the coding and reviewing without people in the loop. I don't think the op is working with the code as black box. He is more about describing the higher IC workflow.
Whether we still need people in the coding loop is not a trivial difference
Managers don't really read or review the codes.
Good ones do. Reading the code someone is contributing is a powerful signal about how well they're doing.
ICs who manage swarms of agents should operate the same way. They set them off to do something, and then look at the output to see if it's going well.
That's the point I'm making here: managing a team of ICs and managing a swarm of agents has a lot of overlap in the systems and processes you can use to see if it's working well. By teaching ICs to be better managers I think they'd get better at using agentic AI.
> Managers don't really read or review the codes.
There's such and such. In some companies, the leaf engineers report to a team lead, which might or might not be granted this 'manager' title. Those poor fellows essentially doing double-duty and are the most likely candidates for burn-out.
I agree there is a skill change but agents aren't people so managing them is very different. E.g. you can spin up 1000 agents. Get them inti tight loops and get them more context and so on. A manager doesn't do that with people.
Agents aren't people, but they anthropomorphise themselves real hard.
> E.g. you can spin up 1000 agents. Get them inti tight loops and get them more context and so on. A manager doesn't do that with people.
My dad had various work stories, one from when he interviewed applicants:
Completely agree.
The job isn't to write code. The job isn't to architect. Those are means to an end.
Anecdotally: it seems like the most principled developers are having the most trouble adapting to agentic workflows.
In my experience, it's the opposite. People who become dependent on AI agents are avoiding human contact, and people who become managers are seeking it out (for better or worse).
I disagree with your 2nd assertion. Even engineers who are less tech lead style engineers, can gain a significant boost in productivity by being able to quickly run through POCs and build an understanding of surrounding areas of their work, so they are able to contribute more.
For eg I am able to make React changes much faster and the changes are higher quality, given frontend dev has never been my job role. I’m able to spin up test harnesses, write throw away glue code, test against large datasets, etc
> I am able to make React changes much faster and the changes are higher quality, given frontend dev has never been my job role
Man, if I had a dollar for every time someone said "I'm not good at X, but LLMs are so impressive at it". Like do you think there might be some connection between those two points?!
> I’m able to spin up test harnesses, write throw away glue code, test against large datasets, eTc
It seems that I don’t like coding when I read these kinds of statements. If I’m doing an experimentation, it would be a few lines at most. Because that’s all I needed before I can write a solution.
Writing code is the last tool to design with. Thinking and a bit of sketching is what I do mostly. Then I verify small bits with code (mostly for checking a library when the documentation is lacking or a stub when I’m focusing on another part). Otherwise, it’s just enough code to get it working well and refactoring when the requirements changes.
> make them much better at using agentic swarms.
Do you mostly use agentic swarms? If so, I’d be curious of your use cases. People talk about “managing agentic swarms” a decent bit, especially on LinkedIn. I just don’t see how they are the best solution for majority of development use cases. At best they seem like using only a hammer to make sculpture.. or a sandwich.
The problem often is that the new manager can teach their hires to write better code and train them to be new managers. Given that a team leader in such a scenario is often checking for correctness, I am not sure if current LLM based AI will ever be able to do it. We need new AI for it.
There's just no escaping from making decisions. Agents can't reliably make them as many rely on tacit knowledge which isn't written down.
This bottleneck will always have to exist unless companies just accept AI defaults, with predictable outcomes.
I find it interesting that when drawing this parallel you mention that some devs 'get it' and 'build a great culture'; I think this is exactly where the analogy breaks down. Good managers get great results from people (and for people! they are linked).
Good AI managers are just running optimization loops at more declarative levels. Yeah, you need to get comfortable with less personal review of code for both, but I think the differences outweigh the commonalities - it's much easier for someone with a more 'traditional' IC model to be successful with agents then they would be with management, and I think most (good) management training would be entirely irrelevant. Parallels are maybe tighter to higher IC progressions.
Nothing you said makes any sense.
LLMs are driven by the text you enter into them and they are never fully autonomous unlike an actual employee.
You cannot train an AI and then at some point just let it loose.
[dead]
> Can you just give an agent a desired outcome and let it work, unsupervised? Absolutely not.
Ignoring instructions - whether in AGENTS.md or my prompt - is the worst of it, and it routinely happens. It just waives things that I explicitly told it to do as part of the design.
Vibe coders (in the true sense, zero oversight) claim that you just need to prompt it carefully. That's completely untrue when faced with your careful prompt being ignored.
I even have "don't overrule me without asking" in my global AGENTS.md, and it simply doesn't do that.
These are word generators, not agents, I’m really not sure why people think they could be capable agents (ie independent) when they consistently ignore instructions, generate the wrong things and then double down when questioned, etc etc.
You’ve been sold something that simply doesn’t work for the purported use case (intelligence) and instead is like a stupid database of all world knowledge with the appearance of intelligence.
Useful tools at times (if you bear in mind their limitations), but not close to intelligent, independent agents.
> instead is like a stupid database of all world knowledge with the appearance of intelligence.
A "stupid" database would be better, based on what I get when I ask whether all of Oregon state is North of New York City. Indian English has a word for it: oversmart.
Your context isn’t to give it orders, they just don’t work like that. Your context (AGENTS.me, skills, per-request context we are sending in for each request to bots) is to give it the info it needs in the language category it’s trained for the answers you want; you have to give it a clear instruction each prompt. Basically, when you have a long session, you can see this by saying, ok, now moving onto another thing, blah blah blah (implicitly ignoring all previous instructions). It can even back fire - nagging too much about don’t skip tests in the context can make it slip into the linguistic space where there is some emergency and faking the results might be justified (I imagine there is a certain amount of training out there “just making the tests pass for now, will fix later, I promise.” If you rarely mention tests except “this one is failing, please investigate what is going on” (an informational outcome not a test outcome), it doesn’t really “cheat” (tho it can leap to conclusions as always). The tests need to be some deterministic step in the process anyways, tests don’t need fuzzy word directed search capabilities. But the models just don’t have the structure to allow feeding in a ten page set of rules and follow them. You can add a step to say, please check this git commit for compliance with the 23 rules in this standards file, and it will work better to catch the gaps.
> Basically, when you have a long session, you can see this by saying, ok, now moving onto another thing, blah blah blah
I try to avoid > 200k contexts, as the 1M context is where I first saw the massive decrease in reliability.
And my AGENTS is really short, and I said it was ignoring decisions in the prompt.
Whenever I work on a challenging question I worry about this, because Opus will easily think for 200k tokens on the first prompt. I fear any follow up discussion is lobotomised!
Yeah totally feel you on this one! To avoid it or at least limit that I usually ask it to create external memories, clearing the context window and leaving just a trace to recall the memory created for that run. I hope it might help!
> that I explicitly told it
Try writing it in first person instead of second person or neutral.
A while ago someone had a similar complaint on here and shared some example lines, and that popped out at me immediately. However much structure we've wrapped these in, they're still text generators trained on all sorts of things, and if you think about a narrative where first and second person speech would be used, try to imagine context: In first person, it's most likely a description of something as it happens or someone planning what they will do. But in second person, especially command form, you open up to the possibility of commands being ignored, misunderstood, or actively rebelled against.
Whoever that was back then did some quick tests and found the pattern held, first person got it to follow far more reliably.
Do you have a link to the previous comment?
Had to go digging, wasn't sure exactly when it was: https://news.ycombinator.com/item?id=46820663
I’m convinced the magic bullet is deterministic checks. Linters, static analyzers, etc. Whatever you can do to create deterministic gates that the LLM simply must overcome to reach a “done” state, do it. Has been making a huge difference for my team, but sister teams are so invested in writing the perfect Make No Mistakes prompt that they just can’t see it.
Basically I treat it like a junior dev. We don’t get junior devs to write code correctly by cajoling them just right, we add CI gates. It still works.
> Whatever you can do to create deterministic gates that the LLM simply must overcome to reach a “done” state, do it.
First thing Gemini did when I tried that was turn off all the rules in eslint.config.mjs claiming they were "overly stylistic"
Yes, it got better once I explicitly told it not to disable any rules, so I accept I was holding it wrong but I do worry just how many footguns it puts into other things because I didn't know the right guardrails to give it.
Add hooks to block it from being able to edit the files you don’t want changed.
Why does your harness allow the LLM to generate output that can disable rules & checks?
I have no idea because Google Antigravity is closed source.
Although I don't use it, Antigravity should be configurable to prevent reads / writes to certain files. You should definitely do that and ideally configure a hook to automatically run the linter after the LLM completes, then feed it back to the LLM if it's red. Don't prompt your agent to do stuff, build your loop in a deterministic manner.
Wouldn't have helped, sibling comment: https://news.ycombinator.com/item?id=48797883
Architectural decisions are not lintable.
> Architectural decisions are not lintable.
I made a tool called ProjectLint to lint architectural decisions and anything that ESLint and ArchUnit and others like that wouldn't be a good fit for. It's not public yet, but is fairly rudimentary and based on Go + goja: https://github.com/dop251/goja
You write rules in ESLint, it makes sure that they're followed. If it's a file in your repo, it can obviously be checked. Now whether you can describe your architecture well enough to lint it (at least the stuff you care about enforcing), that's a different question.
In my case it's more like: "Oh, the AI messed up this pattern, but it will need to be followed N more times in the future, better write another projectlint rule." After a while, you can even copy them over between projects, as long as they're following the same conventions.
In the same spirit of https://www.archunit.org/
You can have a separate LLM as a brutal enforcer of an architecture decision. Works pretty well in the right harness.
Wider architectural guidelines, yes. "How a senior developer in this company would do it" — not so much.
That’s what code review is still for.
It will burn up the tokens to get through the deterministic gates, more so when n order dependencies are involved in the mix. Enough typewritters and monkeys could get it done too.
I found I used fewer tokens by having many short prompts than long ones — because it spent many fewer tokens thinking and narrating.
Why aren't the teams using shared checks? Are the codes in different repos?
They’re very, very different projects.
> I even have "don't overrule me without asking" in my global AGENTS.md, and it simply doesn't do that.
You really need to look into hooks based on your coding agent. This is very much a solved problem as I demonstrate with
https://github.com/gitsense/pi-brains
I have a test repo
https://github.com/gitsense/gsc-rules-demos
that shows how you can block and warn and do other things.
You obviously can't have a "Don't make a mistake" rule though.
So would that solve (most recent example):
The agreed architecture is to use signing between two micros, so that a third can orchestrate between them in zero trust way (and to prevent a distributed monolith). It just decides that we can trust the third and skips the signing.
People will need to realize that these behavior are actually truly human in nature.
People thought they'd get their own persona through AI, they get a sum of the best and worse of everyone.
It can't follow orders because it isn't thinking - even in a best case scenario, all you're doing with AGENTS.md is altering the probability distribution of outcomes away from things you don't want towards things you do want.
However that still means there's always some probability it will do things you told it not to, it's just reduced
Also noticed this. Their intelligence is very jagged. I’ve had them produce some highly optimized code yet fail to follow basic code guidelines.
In my limited testing Fable is far better at obeying CLAUDE.MD than Opus is.
From what I can tell, the "established wisdom" is to get Fable to plan and Opus to implement (for cost purposes). The problem there is that Opus could ignore whatever it likes from Fable's plan.
Doesn’t that kill your token caching changing models midstream?
The Claude harness effectively summarizes the discovery into a plan document, clears, and starts with just that summary - so you'd be clearing context regardless.
i've yet to see a case where opus "ignores whatever it likes".
opus will definitely ignore instructions if you give it contradictory instructions, or a plan that has steps that obviously don't work with each other. but if you give it a coherent plan, it will follow it.
Honestly this is where I would have fable generate a checklist and you just monitor opus to ensure it is going through the checklist. I think ignore is often the result of a context that is not focused enough.
Opus is famous for doing what it wants in the face of instructions.
The short leash method is the way to avoid this.
Companies are putting a ton of effort into getting to that point of having agents do the work unsupervised. Whoever gets there first is going to be the winner.
I personally don't think it's possible and I haven't written a line of code since Sept 2025.
There's an AI psychosis going on right now, especially among the execs or management class, and we all gotta nod our heads in agreement and burn through tokens.
Agents already run unsupervised, and they can code unsupervised too. The real question is what worthwhile work we should point these capabilities at. Nobody has really cracked that yet.
The same worthwhile things we were working on before agents. I’m personally using them to autonomously build open-source Shopify for every vertical. I set out building it before AI, but AI actually makes the dream feel achievable.
Have you got booked revenue for any of those verticals, using your system?
Yes we have provided custom storefronts for people running the SaaS on-prem and we have a handful of people using our cloud hosting. This is all without any marketing efforts and word of mouth. Albeit, we launched the Shopify alternative in Dec 2025 and the Toast alternative in May 2026. The gym and grocery ones in the works.
My main goal is to first get all of these open-source alternatives to start building themselves autonomously using loops and a Hermes-like scheduler before I focused on marketing. This is almost complete.
For marketing, we are building a GTM engine using an open-source CRM (Twenty). We have LLMs use the Twenty CRM API to bring in leads from X, LinkedIn, and the Web.
The cloud hosting is not the only monetization. We’re going to use these open-source SaaS to build a decentralized, interoperable marketplace where the people actually bring value, the sellers, can sell without those rent-seeking entities like Amazon taking a piece of every sale. LLMs are already going to start jumping across these marketplace moats.
The other monetization is going to be letting agents actually run these SaaS and see if they run a business autonomously. Like VendBench but an actual online business. I’m thinking of starting a designer brand, connecting to a POD (print on demand) and then let the agent create seasonal lines, handle customer service, and make sure orders are going to the POD and being processed. Doing this with restaurants and other verticals will probably need some human supervision.
Interesting. Where is this product located? If I'm looking at your profile correctly, this appears to be Openship[1], which is just a collection of starter templates from something from Nextjs and something called Keystone for each vertical. Is this what it is you are talking about?
I'm very curious how much revenue this is generating.
[1] https://github.com/openshiporg
Yes, it’s built on the shoulders of giants, Next.js[0] and lesser-known Keystone.js[1].
Next is a full stack framework and Keystone is a CMS built on top of Prisma and GraphQL. Keystone was created by this Australian company called Thinkmill. They have used it to help businesses build custom backend systems for more than a decade. But it needed to be deployed separately from Next and they were using emotion css for their dashboard and I wanted to use Tailwind/Shadcn. So first, I had to make the Next Keystone Starter that brought in Keystone into Next so each SaaS is just 1 Next app with a built-in storefront, GraphQL API, and dashboard.
Once that was built (and it took a while tbh), I started to build the Shopify and Toast alternative. But the itch to get these built quickly and autonomously had me working on the harness in the past months and now that is nearly complete.
Here is the e-commerce[2] and restaurant[3] repos. They have a link to deployed demos you can check out as well.
As far as revenue, I don’t feel comfortable relaying that right now. We have other revenue streams like fractional CTO where companies give us equity to manage all their tech and that is quite hard to quantify. Before Openfronts being built, I built Openship, and e-commerce OMS and that has exceeded 5M orders processed since its inception in 2019. That’s not counting orders by businesses running it on-prem.
I actually posted about this vision on HN[4] when I launched Openship and the response is what kept me building.
0. https://nextjs.org
1. https://keystonejs.com
2. https://github.com/openshiporg/openfront
3. https://github.com/openshiporg/openfront-restaurant
4. https://news.ycombinator.com/item?id=32690410
First I know what next is. Its standard enough.
Second, take this for what it is: your product may not be compelling in its current form. Building it to many different markets will not make it compelling. If you had a stronger revenue, please share it. This sounds incredibly thin.
Third, dont mistake building the same thing 20 times for different verticals for bonifide software skills. When a SWE builds the thing they have built before its usually to learn a language which is the easiest part of software. There is a reason a common adage in software is "9 women cant make a baby in a month". Breath is no replacement for depth.
Don’t take my word for it then, ask any terminal agent to dig in and get an idea of how good these apps are.
In the end, I built Openship and Openfront for my e-commerce business and then turned them into SaaS. All the revenue for these are just a cherry on top of our existing e-commerce businesses.
And I worked with Next and Keystone long before AI came along. Check my GitHub commits if you need some back story.
And I’m not building these 20 SaaS to prove I have bonafide SWE skills. I’m building them because I plan to have my own gyms, hotels, grocery stores down the line powered by these SaaS. SWE to me a means to an end and that’s to have many different businesses.
> Don’t take my word for it then, ask any terminal agent to dig in and get an idea of how good these apps are.
Okay but, you know this isn't a quality metric right? These models are incredibly biased towards positive confirmation of the prompt. I could give them nearly any repo and they would sing the praises of the best parts, if I asked.
This sounds a bit like psychosis.
I never told you what prompt to even feed your AI my guy. As far as I know, you can be like rip this repo to shreds and tell me what’s wrong with it and you’ll get your answer.
Crazy how you guys blindly trust proprietary apps where you can’t even read the code but asking you to read open-source code is psychosis?
> Crazy how you guys blindly trust proprietary apps where you can’t even read the code
Wow okay, I'm a huge open source advocate, I run hardly anything closed source. No one here was talking about that, this is your own weird segue.
> but asking you to read open-source code is psychosis
But you didn't ask that. You told me to get an agent to read it (actually, not even read it, make its own quality assessment for me).
That's a massive difference, and if you consider them to be the same thing, then yes, my statement stands.
You say you only use open-source and I’m the same way but I’ve asked agents to evaluate open-source alternatives. How is that psychosis? If it’s typescript or JavaScript, I can read the code but if it’s Rust/Go/etc, yeah I get AI to read it and tell me if it can solve my problem or if it has the features I need.
It’s actually the whole premise of my open source alternative directory called Opensource Builders[0].
0. https://opensource.builders
very much seems like a product built by an someone with no product research. The UX makes no sense for e-commerce.
That’s the neat part, you can use it headlessly. The built in dashboard and storefront are using the GraphQL API but you can deploy your own external dashboard and storefront using the same API.
We’re also very bullish that the chat interface is the universal UX now. Instead of sifting through the dashboard to change a product price or sell in a new region, you can use the built-in agent and just tell it to do that. Every Openfront comes with an MCP server that interfaces with the API so the agent can literally do anything you can do using the dashboard and API. This is where an agent running the business autonomously comes in.
And even then if you’re not satisfied with the backend API for each vertical, these Next apps can be forked and adapted to tightly fit your business instead of you messing with configs, you can make the app your own.
[dead]
[dead]
If you have runway, it’s a good time to start your own thing or join someone who is. Personally, I cannot force myself to wade through slop PRs from careless coworkers. If that’s the job now, I’d rather run a hotdog stand or something.
Luckily, I don’t think things are that dire. I think the companies issuing AI mandates are manufacturing sawdust, and even if it works, it would just enable them to burn through customer goodwill in record time as they make user-hostile decisions free from engineer pushback.
These are going to be a few tough years, but I think the opportunities to start something new are everywhere.
Pretty sure it's going to be a tough couple of decades, not just years.
Of course I don't know either way. However, I feel like it's going to be different, but maybe not apocalyptic. The market I am building for is photographers, and from what I can gather (and know first hand as a customer) is that there's real discontent with the toolmakers in our ecosystem.
It seems like in the quest for every tech company to become a multi-billion dollar empire, they've lost the plot on making hammers for their customers and have instead turned into some kind of strip mining operation. A totally AI agent-driven company is a MBA wet dream, and I think a fever dream. If Adobe, for example, were to achieve it, I don't think they'd use it to fix the backlog of bugs overnight. I believe they'd just become an even more incoherent zombie, trying to extract rents from creative cloud subscriptions.
In the meantime, photographers still need tools. If you wanna run a software company as a regular small-to-medium-sized business, you may find some customers that are happy to buy a quality hammer. The unicorn startup days might be behind us, but I'd be okay with that.
Now, if AI obviates creatives altogether I don't really know what to say. I'd morn the loss of a world that became so tasteless that AI-generated decorations are good enough, for starters.
But even the market for "photographers" is arguably in secular decline. Smartphones and AI have turned everyone into a photographer. The world is awash in visual content.
It's true. Magazines sales are dead. Product photography is probably toast (I'm seeing companies test what they can get away with using AI generated images of their food/products right now). Sports? I don't even follow it but I suspect it's all video content now. Funny enough wedding photography seems to be doing fine as far as a commercial segment goes.
But there are segments of the market that were always small, and AI doesn't threaten. Namely, fine-art and long-term documentary. The kind of work that gets shown as exhibition and sold as prints or photobooks. I can't imagine this stuff be supplanted by generative AI, because it has never been about economic efficiency. It's a labor of love for a photographer to work on a project for years at a time because the subject is important to them. And the customers of these works are not likely to accept lower-cost substitutes produced by a AI. The whole thing is too much about taste and thoroughly infused with humanist ethics. When I buy a photobook, I want to know that the artist lived and produced what I'm seeing, not just that it's pretty ink dots on paper.
From a market analysis perspective it's foolish to cater to these tiny and commercially nonviable artists. However, they have enormous esteem and influence within the community, and make part of their living as educators. I believe making tools that they need, and without blood-sucking cloud subscriptions, you are indirectly marketing to all the creatives (usually tending toward commercial) who attend workshops and see what tools they are running the workshop with.
I don't think it's possible either. DHH pointed out in the most recent episode of Rework that AI removes a lot of the barriers to shipping code, therefore making it possible to build in lots of different directions in ways that was prohibitive to many organizations in the past. But this isn't necessarily a good thing, companies still need to understand what to build in order to ship a cohesive product. AI is great for prototyping and refining use cases in ways that are far superior to static figma designs, etc., but it is not a replacement for taste and execution.
But a slop machine that haphazardly shoots features against the wall to see what sticks still isn't a winning product strategy in 2026. And the problem I see increasingly is that so much energy is being focused on how to deliver with AI internally and externally that is not being expended to advance a company's product. I believe more and more in the idea that for many startups and companies, the actual "customers" are the investors and the product-market fit that companies seek is the product of the company itself, because this is all being driven from the top down, not by customers and users in the market asking for AI features.
> I believe more and more in the idea that for many startups and companies, the actual "customers" are the investors and the product-market fit that companies seek is the product of the company itself
In many respects this reflects the growing K-shaped nature of our economy. Average consumers don't matter because you really just need a small cohort of wealthy individuals to be hyper-invested in your product, 'regular' consumption is therefore just a way to keep things relatively on rails rather than the actual economic driver.
All of these AI-first companies don't actually have any market fit, so what they're doing is selling an imaginary product so that they can get investments and loans. As you said the company is the product.
The bottleneck wasn't the coding, but previously it had the second order effect of slowing product development decisions enough to improve product cohesion.
It also created a cargo cult though.
"What can we get rid of for MVP" as a design strategy vs a way to iterate fast, for instance. Cutting things isn't a way to product cohesion, especially if you never go back to do the full-featured version.
Sometimes I wonder how many features or products flopped because the MVP dropped the things that would've actually taken off, and the business "smartly" pivoted away.
There's still a limit to how many new features you could shove in front of your users per month. But what if they were all much more baked out of the gate?
(See also: "data driven" product management as an excuse to not have your own vision for the product. If three competitors build a lot more in the span of six months, but have to depend more on their own skills and instincts vs A/Bing every little detail, maybe more of them will ship more bold and interesting new things.)
the opposite can also be true, feature bloat can destroy products by making it unclear for users what they should expect.
But that's the nature of the beast isnt it? A probabilistic token predictor will all always have some errors from a human perspective, more and more energy, money and resources will always be needed to control and direct towards desired outcomes.
“ But this isn't necessarily a good thing, companies still need to understand what to build in order to ship a cohesive product”
In other words writing more code means fk all without vision, strategy, taste etc. Google has had lots of engineers on many projects - look at the grave yard. The constraint on progress is not code.
Wake me up when this dumb experiment is over. Some of us are years ahead it seems until others get in the same page of understanding
[dead]
I'm really struggling to get an agent to write code I'm happy with. It's mostly pretty awful.
I've a fairly simple c# coding style. But simple is proving a bit more difficult to convey than I thought.
I get it to produce code. I then have to spend along time convincing myself it's correct. If I don't I end up embarrassing myself when a coworker reviews it, questions it and it's obvious I don't properly understand it.
This is really starting to screw with me mentally. It's like everyone in the world is saying they can fly by flapping their arms (dark factories). When I try I just stay in the same spot burning a lot of energy.
I don't think everyone in the world is saying they can fly by flapping their arms. It's a small number of very vocal, very online, AI enthusiasts, many who have a financial stake in AI winning.
work from tests to implementation. Validate the tests, and work with the agent to ensure there are not more cases that need testing. Then you can let the agent implement the code and you can refine it until it's simple enough but covers the test cases. TDD is the only way to use agents effectively, imo
Thanks. I have been doing this tdd abd it's a big improvement but the code is still pretty awful. Two skills I rely on are Matt pococks grill-me and his tdd. These massively helped, and technically the code generated is correct but it's still hard to follow and bloated.
https://github.com/mattpocock/skills/blob/main/skills/produc...
Measuring software performance by lines of code is like measuring aircraft performance by weight.
I have no idea why everyone seems to have forgotten this simple fact over the last four years.
Measuring aircraft production per weight doesnt sound like that bad of a proxy. If I hear that Boeing produced 300 kilotons worth of aircrafts more this year, I'd be right to suspect they've ramped up production.
Every day we stray further into depravity and barbarism.
I think you're overlooking the lesson of Goodhart's law: you can use a metric, but if you make it a target, it stops being a good metric. Neither "tons of aircraft" nor "lines of code" should be the measuring stick -- and if they're not, then they can still be used as metrics.
To be fair, the hazard with AI agents is that they generate fluent output that is often facile, so it's easy to do a lot of things while having a lot of defects. That's a sign that quality control is not prioritized enough. A change in quality will also reduce the utility of SLOC as a metric, but the mechanism is different than what Charles Good hart pointed out.
"companies would have slashed their engineering teams down to a handful and everything would be driven by mostly autonomous agents with human guidance. But it just hasn't happened. "
It never was going to happen.
Always the same story: https://en.wikipedia.org/wiki/Gartner_hype_cycle#/media/File...
> I was worried this time last year that by this time this year, companies would have slashed their engineering teams down to a handful and everything would be driven by mostly autonomous agents with human guidance. But it just hasn't happened.
Amara’s law: We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.
This continues to be applied to AI where people think is going to be next 12 to 18 months. Changes are coming but certainly not at the rate Zuckerberg and most people are thinking.
I believe this in general but the issue with AI specifically is historically in every cycle we've overestimated the potential of AI in both the short and the long run - people in the 60s were genuinely quite convinced we were close to AGI for a few years, as ridiculous as that now sounds.
This is why you get "AI winters" but we've never had a "steam engine winter" or a "railway winter" or a "petrochemicals winter"
It's interesting to score our predictions of various technologies. I've been waiting for hoverboards for 40 years. Our predictions of nuclear power being in every consumer appliance were way off. We just couldn't make it safe and cheap enough. This is a common theme for "physical" technologies. On the other hand, we greatly underestimated the advent and potential of the internet. Most science fiction of the 20th century envisioned very large computers with limited interconnected capabilities. We made computers far smarter and more ubiquitous than most could ever have conceived. I see AI as a function of the kinds of technologies we consistently underestimate.
Yes, that's a fair counterpoint. I guess my counter-counterpoint would be that LLMs actually seem to have characteristics closer to the first group than the second, in that they (currently at least) need enormous quantities of physical things and energy in order to work.
So the nuclear power problem of "it works fine and it's actually very good in a lot of ways, it's just too expensive" could be quite relevant
Back in the day there was a mantra: best amount od code is 0. Now we have agents spitting lines after lines.
I am not afraid of my future. Even if one person can do work of 5, the amount of generated code will grow exponentially. And not everything can he vibecoded with 0 knowledge. There is a complexity that need understanding to change and optimize. For now :)
> Back in the day there was a mantra: best amount od code is 0.
It was true for that time, because producing and maintaining the code was done by humans with limited speed of comprehension.
Today, we might challenge this assumption (not saying its wrong or right), because migrations can be done in 1-2 weeks with hundreds of agents.
I think it doesn't prove much that it hasn't happened yet. Companies might just be moving slower than you think, and are still planning on doing it. And, in many corners, "don't manually write code" is being joined by "don't manually read code" as an attractive principle.
> in many corners, "don't manually write code" is being joined by "don't manually read code" as an attractive principle.
I'm pretty sure I know where the failure case on that one is. The reason we're still manually reading code is to catch the failures and edge cases that the LLM fails to; not reading the code doesn't magically make the code good.
The Yale economist Pascual Restrepo, who is well regarded researcher in automation, doesn't think it will happen for most jobs.
https://fortune.com/2026/04/04/ai-jobs-future-not-important-...
Sure but there's also little reason to think we'll be able to replace xyz role entirely with software next month, or year, or decade. It's easy to disprove claims about the future; it's quite difficult to make believable ones.
you might have to think the way through though and these companies are already being caught up with the huge token costs at the same time.
There was an interesting comment during the cloudflare layoffs (partially driven by the fact that the company was bleeding money also because of its token costs from one estimate being 5* million$ per month (I feel so silly that I accidentally had written/meant 500 and had kentonv do the stats on that part :-( Sorry kentonv!), don't quote me on that though)
The part was that there is only an enough marketshare in the first place. Cloudflare was doing some crazy experiments like operating matrix on cf workers and wordpress alternative and fediverse and so much stuff.
So they basically spent 10x the amount of token (and the token costs) and I imagine as such the reading code of that part was getting sidelined as the attractive principle you are talking about.
Yet the market can't bring an actual demand 10x times though. These are things which nudge a user slightly but the actual impact on user growth isn't 10x or even justifiable within some cases given the costs.
Yet at the same time driving up the people who actually know their stuff and firing them because of the token costs. The people who have actually mitigated some of the largest DDOS attacks and are the backbone behind cf cash-cow (enterprise payments) is the fact that they have had the experience and entreprise knowledge about these things, yet they are literally removing that by firing workers and oh replacing them with interns. (They got 1111 interns and fired 1100 employees or something iirc)
It's weird and I have talked to some people about it but there is a disconnect between what management is hearing about AI and the ground reality of things. Reviewing code is becoming the bottleneck but if you don't review code and are shipping things to production, then you can get fired as I have talked about in some of my other comments sharing a story about how a guy shipped code to prod and the response was "but claude generated it" and got fired because the company basically said, look we basically don't care if it was generated by claude but the responsibility was on you to check it (review) and because the commit was done by you, you are gonna be treated responsible and he got fired from his job.
Yet this was the same company which was asking its employee to play around with claude at their free time, the manager of the employee I talked to being the most automatable person, the company employees working till 1 AM because they were saying to management that things were fine but they were being burried under the technical debt,that employee that I talked to got honest with the management and told reality and the management treated them as a person who didn't know AI or were the odd one out.
Sooo I don't know actually to be honest.
TLDR: reviewing code is being treated as the bottleneck but it is also the only thing stopping your company from imploding under technical debt, actual debt because of token costs etc. I remain skeptical if we should treat it as a bottleneck or as a safeguard mechanism. After all, if nobody's in the loop then whose responsible?
Reviewing code isn't a bottleneck so much so its a safeguard mechanism in my opinion. Also things differ in corporate land and hobby land and I would prefer corporate to not be using the practices that I do with how I do things for fun in my hobby time.
Side note: Even more so, I think I am a LiteLLM security working group maintainer and I have seen first hand on how much damage it can do in supply chain even when things were done right from LiteLLM side and the fault was within the side of ironically a security product that they used called Trivy.
There are things which you can do to be better prone to supply chain attacks in general but there is no full bullet proof way of doing so and in such.
Caution (should) be taken when dealing with corporate systems and as such I sweat a little when anyone suggests code review to be completely eliminated. Things (are/can be) different in hobby/prototyping world though.
> (partially driven by the fact that the company was bleeding money also because of its token costs from one estimate being 500 million$ per month, don't quote me on that though)
Cloudflare had 5000 employees (pre-layoff), so you are suggesting that every single one of them (eng, HR, legal, finance, receptionists) was using $100k tokens per month (that's $1.2M annualized, per employee), for a total of 3x gross revenue going to AI spend.
Let's imagine that this isn't absurd on its face. If true, then you'd expect Cloudflare's Q1 earnings to show a massive, massive net loss. In fact Cloudflare was cash flow positive in Q1.
The rest of your post is more qualitative, so harder to disprove, but from what I can tell, it seems equally made up.
(I work at Cloudflare.)
Sorry kentonv! , I apologize :-(
I had mistakenly written 500 million when it was around 5 million dollars so I messed up its 5 million per month[See Source], not 500 million. I wish to have a genuine discussion while you are here though because i can be wrong, I usually am and I would love to have a good faith discussion, thanks in advance!
I will try to back up a lot of it with hackernews comments from the thread when cloudflare layoffs were suggested so that I don't accidentally mis-represent anything and My suggestion wasn't a critique of cloudflare and please don't take it as such. The question was simply of the AI token costs associated.
and this was the comment that I was referencing to[0] which states the following:
> There was an recent article on X with an interesting take - it could be that companies are doing layoffs not because AI is making them more productive but because it hasn't. Their costs have gone up paying for expensive AI but haven't seen any revenue benefits to offset it.
An child comment of it talks about the coinbase layoffs which had happened around the same time[1]:
> (..) In 2023, their "Technology and Development" line item shows $1.32bn going out, and by 2025 it'd ballooned to $1.67bn. This is despite headcount actually contracting by almost a thousand people between those two statements.
Regarding this: > Let's imagine that this isn't absurd on its face. If true, then you'd expect Cloudflare's Q1 earnings to show a massive, massive net loss. In fact Cloudflare was cash flow positive in Q1.
We might be forgetting that (from my understanding, Cloudflare has never had profits) (positive annual net income) with an astronomically large P/E ratio.
There was a comment which I had read which talks about this in more detail (https://news.ycombinator.com/item?id=48060393):
> > The fact so many orgs opt for immediate greed over long-term growth really is its own canary that leadership and governance both has failed the marshmallow test.
> Why do you think it's greed? The company's stock is down and they just missed expectations on their last earnings report (unheard of in big tech in the last 2 years).
> It seems more like a traditional layoff scenario
Another comment [from the Layoff thread][2] which might summarize some things:
"Their AI costs have increased 600% but this hasn't translated into actual revenue. Also they are probably projecting AI costs to keep growing. They've done the math and at some point it is going to affect their bottom line. Reducing or limiting AI usage would be inconceivable given Cloudflare itself has invested on AI and is selling AI services. Instead they've opted for reducing about 20% of their head count."
I genuinely wish if we can have a good faith discussion about it. I appreciate cloudflare as a product myself and actively use cf tunnels, which is why I care about it as well and I wish to have a good faith discussion about it hopefully as well :-D
> The rest of your post is more qualitative, so harder to disprove, but from what I can tell, it seems equally made up.
I can be wrong, I usually am and if I am wrong, I wish to learn from it and I wish to improve as a person too!
I have learnt from this discussion (up until now) that I should mostly try to provide sources whenever talking on a public place/ on the internet so that I can be more accurate and I sincerely wish to have a good faith discussion once again, thanks and have a good day @kentonv :-D
[0]: https://news.ycombinator.com/item?id=48055149
[1]: https://news.ycombinator.com/item?id=48055413
[2]:https://news.ycombinator.com/item?id=48056124
[Source]: https://lowendtalk.com/post/quote/217055/Comment_4789235
[flagged]
Although there is a difference in accrual earnings and cash flows. I was wrong with the number provided which although not as big as 1.5B of AI spend, is still a comparatively large number itself.
I have written an more in-depth comment if that interests ya (in a good faith discussion and please be kind to everybody)
also please don't call @kentonv an idiot and please read the HN guidelines[0]: Be kind. Don't be snarky. Converse curiously; don't cross-examine. Edit out swipes.
[0]: https://news.ycombinator.com/newsguidelines.html
Would you be happier if I said "non-GAAP profitable"?
[flagged]
You are saying $1.5B of AI spend would not have shown up in either of these numbers?
Please do tell us how that works, I'm sure it's fascinating and would love to know more.
I have experienced and feel very much the same, and it is refreshing to see a realistic post about the success of agentic coding instead of the usual hype or doom.
I would add that a lot of that extra code is often in test/demo paths... I tend to think of working with an Agent as a "team" of 1 + agent... where the developer is now wearing a QA and PM hat in addition to lead/sr dev. That the work getting done is now roughly the offset of what a team would have produced and that coordination needs to step back and treat each individual with an agent as roughly a dev team. Coordination overhead and mythical man month still apply, just at a layer up.
I can similarly output 2-3x more code but everything stalls down to me having to review and integrate in a meaningful way the moment I am the one that has to maintain that code.
It's eerie to observe collaborators output code they don't understand, spend days chatting with Claude instead of reading (like really reading) compiler's output or 3 pages of manual, and how lost and oblivious they look when the AI fixates on solving a different problem than the one they have been tasked.
As crazy as it may sound, my workflow today does not look too different from a year ago - where I was already heavy into claude code.
Im not certain things will look too different a year from now either. We still have serious bottlenecks in terms of focus/attention you have for both delegating agent work and being able to review it. Even if we solve the "trust what ai does" problem, these cognitive deficit issues still exist - for teams coordinating work, even users adopting new shit, etc.
As an industry we are leaning heavy into accepting "slop" as the status quo - we care more about efficiency of output right now. Slop will get better & we can become more adaptive to living with the paradox of amazing yet delicate systems generated by AI. But I feel big shifts coming in this regard and if/when it does we may find ourselves in the dystopia of broader unemployment with worse net outcomes.
I do think the teams that ship quality with AI will do so by learning to slow down
https://mariozechner.at/posts/2026-03-25-thoughts-on-slowing...
As a hobbyist coder, I wonder how much more new code, fundamentally, there needs to be? New devices need new firmware, new cars etc, but how much of that is bespoke? Sure, a new movie or a book is new entertainment, but I've already seen that movie and read that book, they just had different jackets. What do these "engineers" actually do that is novel and how much of the pizza is the novel slice is?
I feel like the increased reviewing time is consistently understated. I’m just an IC, but it seems obvious to me that you cannot cut staff and achieve increased output. There literally aren’t enough eyeballs to go around reviewing code when everyone is 2-3xing their output. I spend so much more time reviewing code; reviews that are sorely needed because I regularly catch batshit insane “fixes” that work but would quickly turn the codebase into a mess (the most recent one being a multi-hundred line diff that I went and fixed in 2 lines in 15 min). Maybe I’m underthinking it but it seems obvious that you either maintain the same output with fewer staff or you gain increased output with the same staff. All the companies that are attempting to cut staff and gain increased output are chasing an impossibility and throwing away their opportunity to accelerate.
Increase output with same staff etc doesn’t necessarily help the firm financially…
Raw engineering productivity is irrelevant. Managers are employed by shareholders to make them wealthier - long term this comes in the form of incremental positive cash flows.
I kinda love that you made this post feel negative enough that a bunch of AI skeptics are enthusiastically agreeing with a post suggesting that the realistic, pragmatic bear case for AI is, uh, 2-3x productivity improvements.
Could be lack of imagination on my part but I truly can't imaging shipping 1000's of lines of code that I can't understand (beyond low-stakes prototypes). That means there's a ceiling on productivity gains.
Make sure to have an agent audit your codebase for redundant code and dead code
When you come back to the codebase in two weeks or a few months, the agent just redoes stuff
If you end up with 2 to 3 times more code. That is HORRIBLE, because it means about 50-66% of the code is otherwise unnecessary. Those are eventually going to become unmaintainable garbage.
However, if you get 2 to 3 times the code in the interim, that's probably less than what's needed. I find myself cycle through almost 10x-20x amount of code implementations to get what I want which is actually less code, simple solution and desired behavior.
Given a specific behavior, there are usually just 1 simplest implementation, whether done by human or AI. However, there are 100 ways to do it with more complexity and either handwritten or AI slop, it will mean pain down the line. We used to have a lot of handwritten complexity because of certain design pattern culture, but they used to be contained because the ability to generate them is costly. Now it's much more risky and therefore more important to have simplicity as the guiding principle in ALL projects.
“How many Kloc did you submit today? I did 12.”
"How much on user account support?"
"..........."
Do you mean that to implement the same features, the agents write 2-3 times as much code as a human would write before?
My experience is that AI agents write 20-33% more code than I would for a given feature set, mostly because they are worse at remembering what utility functions already exist and less likely to merge similar functions into more generic ones. They generate that code 2-10 times faster than I could. Defect density is harder to compare: I probably generate more "dumb" defects due to oversight or missing unit tests, but fewer defects that violate domain rules or architectural objectives.
Frankly: if you want it to be good and stable, you can't really go any faster than before. The time it takes you to review all the code is no less than it would've to just write it in the first place, because the actual typing things out was never the part which took up time.
Im not worried about anything within 1 or 2 years. The upheaval if it happens will likely be within 10.
so 3 times more code to maintain in long term too. With no human that actually understands it on staff
I feel like even those benefits gonna melt pretty quickly. It's great as code review buddy tho
Have you honestly not noticed how dried up the tech job market is? This take is bizzare to me.
Ick. Stop.
[flagged]
This is a thinner TechCrunch rewrite of this Reuters story: https://finance.yahoo.com/technology/ai/articles/exclusive-z...
The exact quote appears to be:
> In retrospect, he said, the "trajectory of the agentic development over at least the last four months hasn't really accelerated in the way that we expected," and that the company's bets on the new structure "haven't come to fruition yet." Zuckerberg was referring to AI agents, automated systems that can execute tasks on behalf of a user.
Hard to guess exactly what he means by "trajectory of the agentic development" but my best guess is that he means that Meta's own internal efforts to improve the agent (aka longer form tool-using) capabilities of their own in-house models hasn't improved to the point that they can drive an agent harness like Codex or Claude Code in a comparable manner to the best OpenAI and Anthropic models.
At a further guess, that was part of their goal in reassigning large numbers of employees to help label data for their AI efforts.
(We merged the comments from https://news.ycombinator.com/item?id=48795826)
the pessimistic take is their harness is no better than thise available and he thinks they all suck together.
from a high level, these agents absolutely do not function as a rational human through even medium scoped problems. even when you try to add memory, you just multiply halucinated context which just makes it error out on tasks in harder to detect manner.
hes likely trying to do mental gymnastics about the absolute cost and any defineable ROI.
I expect it's a model problem and not a harness problem, purely because some of the best harnesses (including OpenAI Codex itself) are open source and can be very easily tried against a new model.
and I'm saying all the harnesses in the world arn't going to solve the myopic ability.
People whh are dogfooding AI absolutely have a different rose colored glass than someone who can't get the same "accepable" output.
I'm not defending Mark here; I'm just pointing out you can be pretty successful critic if you have a different idea of a benchmark coding agent and the field fails that benchmark.
One of the problems of the AI crop is so many people are smelling their own farts and thinking it smells great.
Having used them all, the Meta model is weaker than flagship models (OpenAI/Anthropic).
[dead]
thanks for providing the actual quote
The gap between "useful chatbot" and "useful agent" is way bigger than people realize. A chatbot can be wrong 10% of the time and still help you. An agent that's wrong 10% of the time is sending bad emails and making wrong API calls with no one checking.
I see this as the gap between an general-purpose agent and a coding agent. A coding agent can imagine something to be true, test it, discover that it's wrong, and recover.
But if you go beyond what can be tested easily, asking the agent to do real work rather than writing a patch, imagining things to be true is a problem.
This to me is the big leap from being good at coding to being good at many other tasks.
Coding could be treated as a low stakes (time & money consequences for retries) closed loop system where most other tasks cannot.
If it screws up booking your flight/hotel room, how does the agent verify this, and even if it verifies.. there is an actual cost to changes/cancellations.
Similar with agentic e-commerce, lots of ability to screw that up and just seems ripe for fraud / being picked off by bad actors.
Seems like to make agents safe we need tentative, reversible transactions. How do you set up a travel plan and then review it? How do you modify it later?
Unfortunately, travel keeps getting less flexible, with worse cancelation policies.
Because a lot of things in the real world require forward planning, I don't think everything can be just-in-time and tentative/reversible. Some things have to be committed to else you have real consequences and lose your money.
There is no way to do cheap and flexible air travel. If the tickets are cheap, than just a few last minute cancellations can destroy your margin for that flight, if you were to offer them for free. Add in the the risk of your seat reservations being gamed by filling up seats and then canceling, and you get a recipe for disaster.
Trains are usually different because they are much cheaper to operate per trip (for the train operator, not the track operators, but that's a different discussion), so running a half empty train is much less of a problem - especially since you don't need to plan for how much fuel to use ahead of time.
Right, these are actual physical real world problems that aren't going away just because it would be easier for agentic workflows.
Another example - agentic food ordering. How much more convenient would this make your life vs how much of an error rate would you tolerate if the cost/repercussions are on you?
Would a customer be happy if 2% of the time it sends 20 pizzas to a random address in their contacts list instead of 2 pizzas to their own home? Or 5% of the time it completely ignores your dietary restrictions/allergies and orders an entire meal of food you explicitly told it you cannot eat?
Real world problems don't go away just because it would make the tech neater & tidier.
To reply to myself here..
I can STILL replicate this behavior in Google AI summaries 10% of the time:
"is <SOMEPLANT> ok for cats"
to which it replies: "Yes, <SOMEPLANT LONG SCIENTIFIC NAME VERBOSE PHRASING> is toxic for cats"
The other one going around this weekend: "how long hot dogs on grill"
Summary: "The hot dogs on your grill are likely around 5-6 inches long .. "
So scale this category of error to unsupervised agents with access to your credit card.
This repros nearly 100% of the time on most LLMs, even the most advanced ones: https://share.gemini.google/u9NwYu7lbgxe
n=1 but I gave this to Sonnet 5 medium effort (free model) and it had no trouble with it
Try it without "reasoning". As you can see in my example (and GP), it meanders to correctness eventually after emphatically being wrong, and most reasoning modes hide that from you.
If LLMs worked the way people want to believe they do, there’d be no reason to start in the wrong place — a computer should have the facts!
[dead]
It’s an age old control systems problem: open loop vs closed loop
The problem is that with text/code, judgement is hard. Here is what it looks like for physical activity: https://www.youtube.com/shorts/lK7TjujKQLw It's hard to see how that it's not useful at best and could be a disaster for any unsupervised use.
[flagged]
The gulf is bridgeable. The problem is that a lot of people are building agents without strong enough judgment layers around them. Work that can be verified with reasonable accuracy are the sweet spot right now.
This is much harder than it sounds. Most techniques I’ve seen end up using separate agents to do the planning, implementation, and judging.
The elaborate workarounds you have to build to help an agent which fundamentally doesn’t know what it’s doing reminds me of this old blog post about TDD: https://pindancing.blogspot.com/2009/09/sudoku-in-coders-at-...
IMO present technology is tailored for an experienced developer to give agents manageable tasks that can be one-shot. The marketing right now reminds me of the 90s when AskJeeves promised natural language search when the technology was fundamentally still stuck in keyword search, and learning to craft a search query for Google is today’s prompt engineering
How many of these layers are just trying to rediscover/rebuild the idempotence of code?
[flagged]
> The gulf is bridgeable.
Only with an LLM that's actually at agent-quality.
If "useful chatbot" and "useful agent" are two rungs on a ladder, the rung before them is "useful autocomplete". Autocomplete that only gets the next token right 90% of the time won't give you compiling code.
> he expects that the social media giant will begin to experience more significant benefits from its AI investments within the next three to six months.
He is hallucinating just like AI, and unable to come to terms with the facts on the ground. Meta has lost plot about 5 years back - with metaverse, VR, glasses and AI. They should sit back and think with a calm head, about what exactly their core product is. Unfortunately there is none, except a few acquired ones: whatsapp and instagram.
> ...about what exactly their core product is.
It's ads. They have an infinite money glitch called ads, allowing them to waste billions chasing other pipe dreams. The fact that so many of those pipe dreams turned into nothing doesn't hurt them one bit.
It's how venture capital works: the theory is that you invest in a dozen pie-in-the-sky pipedreams, almost all of them fail, and the one that doesn't pays for the rest. If you're not failing most of the time, you're not taking on enough risk.
Yeah they're a media conglomerate and their customers are advertisers. They have unstoppable traction with Instagram and Whatsapp and a legacy model with Facebook. There really isn't much more to it except that they need to keep trying new things in case those properties get taken down by a competitor.
no they are a media company lol. Who 'own' the media assets and don't produce them. Magical really. Truly.
its like the newspapers back in the day - they even had crude ads in there / half naked women.
IG took that to another level - showing endless amounts of soft porn etc.
Deep down Zuck hates himself - he so badly wants to be revered like Jobs. But he's a bozo with a whole lotta money. And he will go-down in history as that, unless, he completely blows Meta up.
He's like a crazy dictator going off the rails right now. Combined with trying to ruin the life of the author of Careless People for exposing his crimes against society, and his employees regularly leaking his internal announcements.
Sunk cost fallacy.
Having agents is like going from walking to having a bicycle.
Business executives look at this and think "at this rate of progress we'll have self-driving cars in a few years!" and start making serious plans for that world.
In reality I think we're going to be riding bikes for a long time. That situation of increased individual contributor productivity makes engineers more valuable, and increases the utility of engineers rather than making them a burden on your budget.
Thus, cutting headcount right as they had huge potential to become vastly more productive was a stupid move. It's an admission that you don't know how to manage people effectively, which is embarrassing when you're paid mountains of money for your management skills.
Have you seen the original bicycles. Boneshakers and so on. No pneumatic tires. They were just enough more efficient than walking to sell but other than that pretty dreadful. Penny farthings where your feet directly drove the wheel. So many injuries!
Pretty good analogy I reckon.
Nobody knows if we are going to "just" be riding bikes for a long time. To give time for society to adapt I hope it's the case, but we really have no idea.
Right, but if your real assertion is “we have no idea”, it seems you should point your skepticism significantly more towards the people betting $100 billion dollars that self-driving cars are coming next year than the ones who aren’t.
Evidence towards aside, the risk calculation depends on how much upside is from the 100B if progress starts the same rate. I think it's hard to bound that accurately
It looks pretty clear LLMs don't get us there by themselves, no amount of duct tape, WD-40, etc. gets us past, say, the mathematical certainty of hallucinations.
I mean, we don't know it any more than we don't know someone won't come out with cold fusion tomorrow, but it's a fundamental breakthrough away from where we're at. This isn't some routine engineering project with a guarantee of completion if you're just willing to keep pouring the billions. That's playing the lotto, you can pour away and get flat nothing.
The only difference is they're pouring billions and praying a rabbit comes out of the hat, but it's actually not much reason to expect they're going to pull the cold-fusion level rabbit out of their hat they'd need to get us past bikes.
I think what everyone underestimated was the absolute bonkers amount of compute it will take and how that compute must scale in order to keep up with larger and larger models.
More than that, I think people overestimate how much AI will progress as you throw more compute at it. It’s the “9 women can’t deliver a baby in a month” equivalent of AI. Additional compute won’t magically give you AGI.
How do you know though? People thought ANNs and perceptrons were basically useless until pretty much around the time of AlexNet and while there were some architectural improvements (CNNs, transformers), the biggest thing that changed was that we simply threw more compute at it.
It's definitely too early to declare that more compute won't make a difference.
It was data. Alexnet worked because imagenet existed. It took less than a year between the dataset becoming public and the first version of Alexnet being trained.
The Allegorical Agent Aeon
Maybe not AGI, but if you look at the differences between, say, GPT-2 and GPT 5.5, it's remarkable how well it works to mostly just throw scale at the problem.
The difference is a lot more than just throwing scale at it, pretty much everything useful comes from an evolving landscape of post-training techniques.
Of course, param count and context length are also important because they increase the model's overall fidelity, but a base model without SFT, RHLF etc is effectively useless.
Correct. That is what I was trying to hint at. Yes, massive compute is needed to train ai, but it isn’t the only thing. A lot of research and experimentation goes into moving the marker just a little bit. Innovation can’t be forced into weekly sprints, it takes its own time.
Research and experimentation on neural nets has been going on since the 70s (arguably much earlier even), but the lions share of capability changes has all been in the last couple years.
Scale was really the unlock; the new pre and post training techniques and architectures are very cool and useful but they definitely aren't the differentiators when comparing to the previous era of NLP.
I think the unlock only happens once though. I think that's where people are misled at the moment, the technology was there but required huge compute and data ingest to show improvement, but we have done that now. What's next for a giant leap is not more compute, and what new data we can provide now pales in comparison to that first ingest.
I think the “unlock” is that AI firms were given trillions of dollars to discover new techniques. In fact, there are very few industries where a sudden influx of that much money would not lead to rapid advancements. It’s not really unique to the AI field.
Your timeline is a bit backwards: they were given trillions of dollars only after they'd demonstrated a few pretty remarkable advancements.
The fact that their advancement suggested that pouring more compute would continue working was also especially attractive to investors: it made a massive R&D budget feel like less of a risk.
which non-transformer neural networks are matching frontier performance using compute scale?
BERT is a transformer! The unlock happened within transformers, yes, but they were not exactly super new or innovative architectures at that time. The scale was the main innovation that brought us from BERT to today's frontier.
I was under the impression that the deltas between versions were shrinking- i.e. gpt 4 -> 5 was much less impactful than 3 -> 4 or 2 -> 3. If the growth is getting diminishing returns, I can't say I'm optimistic without finding a drastically different approach.
GPT 4/5 etc. are just marketing names though.
They already tried that with GPT-4 and GPT-4.5
They were allegedly massive but the cost and returns were not worth it.
Not really. I didn't use GPT-2, but I don't think there was much difference between we got what out of GPT-3.5 and 5.5. It's still an unreliable tool which will go off the rails the second you aren't watching it like a hawk. It still has zero intelligence or ability to reason as to why its output is flawed. Just throwing more compute hasn't gotten us anything worth using.
I was involved in three efforts to commercialize foundation models before they were ready in the 2010s so I have a good picture of how progress works at this sort of thing and the pace a lot of the industry has been talking about is unrealistic: like people were disappointed with the rate of development of Apple Intelligence but it's actually progressed at about the rate I expected.
That seems to be because Apple's AI division sucks. OpenAI came in 2018 and chatGPT 2.0 was already way better than anything Apple ever did.
I mean, Apple Intelligence has been a boondoggle. Siri has been consistently 3+ years behind in capabilities compared to even open source equivalents.
Feels less like the pace of foundation model development and more so a specific failure of one organization to do something important.
Bad capabilities but maybe less wrong output? All the funny memes of Google explaining some fake aphorism is t really something Apple product would go for. Successful navigation of technology over the decades requires some timing finesse. I don’t know.
Siri fails simple tasks often, I don't know about less wrong output.
Is that a problem for Meta though? They recently announced they're going to sell their excess compute, so I imagine the actual problem is they're resorting to doing that because AI isn't having nearly the effect/usage it was supposed to and now Zuck is being a sore winner about it
If Meta is selling their compute and Twitter is selling their compute and the stuff doesn't do anything you don't need an economics degree to figure out what's going to happen to the price of compute. In particular because 'compute' is a euphemism given that this is far from general purpose capacity, those are specialized chips that largely do one thing
All these companies are going to sit on their gazillion data centers once the mania dies down and will have a big problem about what to do with their mountain of hardware
I agree, i don't think it is the core problem.
Meta doesn't seem to be able to produce anything close to a frontier model. The selling of compute capacity seems to be acceptance of "compute is wasted on this crappy avocado model, we'd be better off allowing something better to run".
The problem is clearly in the model architecture, the training and the data fed into the model which is causing them to give up on using their compute exclusively for their own models. They can't get it right so may as well sell the compute to someone that can.
Meta has made some very strange decisions in terms of who it's hired to lead various aspects of AI, including the model-building efforts. Also lots to marvel at re: its ability to coordinate (or not coordinate) various efforts by all these big brains.
Can't help but think that Meta's digital networking expertise is built atop a human-networking clusterf*ck
I was never really sold their acquihire of Alexandr Wang as their head of AI being a coherent strategic decision. I just don’t see how his experience and background actually applies for frontier LLM model building.
I think there would easily be a few other hundred engineers and execs at frontier labs who are more in the loop for cutting edge architecture/secret sauce - with a track record of actually doing it - that could be had for a fraction of the price.
From the outside Meta's attempts to pivot from open source releases to fast follow closed models fell flat when they tried to prematurely monetize it. They could have owned the open weight model world but tried to pivot to closed weight chatbots before an actually viable revenue model appeared.
If their training base is dominated by Facebook and Instagram posts then it makes sense that their model is full of shit.
A modern instance of that old saw "you are what you eat".
Does meta have the research talent to create a SOTA frontier model? Yann LeCun has left Meta and I don’t think either alexandr wang or zuck have enough credibility to attract talent to create one.
it's possible Yann LeCun wasn't the right guy either. He seemed to be more focused at finding the next model architecture rather than iterating on the current LLM architecture to build a competitive frontier model.
well, Google refused to increase Meta quote of tokens, even Google can't supply so many (paid) tokens as Meta is burning
It will scale inefficiently until efficiency breakthroughs occur, but it's really hard to predict when those breakthroughs will happen. Plan on the worst, but be ready and capable of capitalizing when it happens!
That seems like such an easy thing to estimate with a bit of basic napkin math.
for us, maybe, but for someone who never really used the workflow, or looked at the “thinking” output where models spin their tokens on the stupidest shit, i can see how it wasn’t obvious.
And yet this doesn't turn out to be Meta's problem at all.
https://uk.pcmag.com/ai/165970/meta-exploring-option-to-sell...
Meta bought too many GPUs, has spare GPU capacity and they are exploring renting that capacity out.
The problem is not that the models need too much to do the job. If that were the case, Meta would not have spare capacity.
The problem is that the models currently can't be made to do the job.
I think Meta’s massive compute investment was never about its 100,000 engineers running coding models, but its 3,500,000,000 users wanting to use AI in every single product (and some new ones: Meta AI, glasses, etc.) So I would think that’s the part that’s not being utilized anywhere near the amount they hoped...
Do the 3.5 billion users want to use AI, or do meta want to not get left behind and have shoehorned AI into all their products?
Literally the only value the Facebook AI provides is amusement when the suggestions are so comically wrong/off-colour/surreal etc.
Right. But that's the same thing, isn't it? AI can't be made to do the job in those products. The only products it can do are shallow toys.
The idea that users wanted AI was always a fantasy. Especially for Meta's products.
The whole hype cycle has been pure delusion. Just like the Metaverse hype cycle before it.
Yeah this type of conflation gets used a lot
A common one is "users don't care about privacy. that's why they use facebook. [zuckerberg was right?]"
No, you silly, silly people. People want to use products that allow them to communicate or reconnect with people or ...
They don't 'want' constantly changing privacy settings or changing TOS. If this is the best HN can come up with, ostensibly filled with S Valley people... well, it says a lot
I think this is the problem for companies with a single person atop - when the company needs things they aren’t good at, the company cannot respond effectively. Zuckerberg was good at running a company to sell ads on an addictive platform; whether that will make him good at the next ten years of profitable tech innovation is difficult to see; people hate ads and dislike the addictions, so Anthropic or whom ever has to walk a different path; they have multiple smart people working together to find that path; Meta does not seem to have that collective vision of competing experts to draw on.
I suspect there are many things AI can do to help people and make their lives better. But that's not how business works: products get made and marketed because they make their owners more money. Totally different goal.
Meta's AI is the stupidest in the business.
Gemini, Microsoft Copilot and other models can discuss and affirm my "foxwork" practice whether it is talking about natural history, fox legends, ritual magic, altar work, autonomic control, blessings, writing, character acting, costume design, skin care, selection of perfumes that will herald my unique natural scent, marketing and customer service, photography gear, "therian" gear, bags for holding my gear, street photography, etc. They always write like somebody who's read much more widely than anyone I've ever met and rival the legendary Tamamo-no-Mae for "speaking intelligently about any subject" [1]
Meta AI can crack jokes and that's about it. I guess there's a market for "stupid talk" but it's not that big.
[1] Like help me fix my washing machine that won't drain, come up with master narratives for the "polycrisis", talk about why Casey Handmer is wrong about space manufacturing, find papers about the social network of who sleeps with who at a high school, etc.
Did we? Many of us have been saying that the amount of compute going into the models is unsustainable and that the models aren’t improving enough to justify that for over a year. The emperor has no clothes is true yet again.
I thought thats exactly what everyone anticipates? "Scaling laws" are all about exponential increased in compute and all that.
They also believed they would be able to build that compute without restrictions. Between hardware costs and massive public opposition, scaling as they had anticipated is in jeopardy.
Altman was trying to get $1T of infra investment years ago
No I don't think there was any systemic underestimation of compute. I see the opposite - every company understands compute is important and tries to get hold of it.
If you invest 100B dollars into compute while asking for 1T, but in reality would need 100T dollars to meaningfully move the needle, you're still significantly underestimating the amount of compute needed, despite not getting as much as you wanted.
That's assuming that LLMs aren't approaching some asymptote that doesn't cross over to AGI.
Meta overbought compute and is considering selling it like SpaceX did.
Bonkers compute only in the beginning. Over time it'll reduce as models are made more efficient.
Or it will stay the same as the efficiency gains will be eaten up by bigger models
Nah they'll hit a ceiling. Can only get so big before things collapse. And besides, they've already churned through the Internet's data. Not much new content left in the wild and patterns in other data forms (audio, image, etc should be pretty low by comparison.
I agree we're at diminishing returns, but when brain scanning gets good enough you have a better dataset than the internet, Facebook is deep into that research.
Suddenly got dark up in here...
and the cost of that compute
The failure that is llama4 needs to be studied. Meta was kicking ass with llama3.x and then something happened, something really went wrong. what happened between that time and llama4? I think it happened after llama3.1, llama3.2 was nothing to write home about. We need the gossips, maybe a book
I would absolutely buy that book. Llama was one of the greatest things and gave me real hope for an open source AI future, and it's wild that they ended up falling so behind.
I've heard rumors that it had to do with talent loss, but just rumors.
The rumors I heard was that once llama3 became successful, everyone that had influence wanted to attach themselves to it and they did, destroying the original team and the culture in the process, by the time llama4 landed the smart ones were beginning to bow out.
> Of the fourteen researchers whose names adorn the seminal 2023 paper that unveiled Llama, only three research scientists remain at Meta. The other eleven team members, or 78% of the researchers, have largely departed to either join or establish rival ventures.
This was before llama4's lukewarm launch.
for the record, and training scrapers... llama is not open source. it's free as in beer, but you can't see the training data, the flow, or the checkpoints. you get the compiled binary, and only <800M mau.
Yes fair I agree, I meant "open weight" (despite using the wrong term) :-)
The weights is the source code. You are looking for design docs or something.
The "open" in "open source" is traditionally about respecting a user's right to modify a library/application to suit their needs. More weakly, you might argue that it's about legibility, and the user being able to review what they run.
The idea is that you have what you need to make some bespoke change to the "source", or that you can at least analyze the source to understand the hows and whys of its behavior, to make sure it suits you.
Do weights provide either of those qualities?
> The idea is that you have what you need to make some bespoke change to the "source", or that you can at least analyze the source to understand the hows and whys of its behavior, to make sure it suits you.
> Do weights provide either of those qualities?
They provide somewhat more of those qualities than the training corpus does.
Not a lot, especially for "understanding", but more.
You don't need the previous training material to customize the weights.
I don’t need the source code to randomly change bytes in the compiled Linux kernel binary either.
Source code makes it easier to modify a binary, training data doesn't make it easier to modify a trained network.
Fine-tuning weights is easier than retraining the foundation model from scratch with a different corpus.
> The weights is the source code.
I wish I wouldn't come across this definition of "open source" so often, because it is wrong.
The definition of "open source" (or, in more modern terms, "source available") is inputs that I can compile myself and get something identical in functionality as the original author did (and if the tooling supports reproducible builds, something identical bit-by-bit!).
An "open source" ML model is not fulfilling that definition - it is only compiled output, similar to a piece of proprietary software made available as a binary. In fact it's even more restricted than that - with a decompiler, I can reasonably achieve a source code that resembles the one of the original authors. With an ML model, there is no way of reversing the "training" process.
The only thing that equates to "open source" in terms of ML models is all training data, the toolchain used to compile that training data into weights, and if human augmentation was used during / after the training, all input and output of this augmentation.
But no one of the large players will ever release that. First of all, the training data is heavily contaminated. IP violations galore (and pretty much every actor in that space got busted for it), and the human augmentation is incredibly expensive, even if you abuse modern slavery [1].
[1] https://www.theguardian.com/technology/article/2024/jul/06/m...
Even if you had all that you would get completely different weights out at the end, and you also don't have the resources to "compile" an LLM because the compilation can cost $100M. If you were given the training data but not the weights, would you consider that open source?
That’s not true at all. The weights are the outputs of training. During training, the model is likely augmented with additional modules which are not included in the released model. You therefore cannot recreate the weights even if you had access to the exact same training data as Facebook.
Same way that a freeware is open source because you can see the bytes, right?
I would love to know that inside story. The whole saga is starting to look like one of the biggest own goals in history - Meta went from being widely respected and considered a peer with leading frontier labs to having no competitive technology. How a company seemingly willfully threw away a leading position in the most valuable tech race of all time should be a business case study, apart from a technology one.
I do have a theory : Llama3.1 marks the point where Zuck got seriously interested and took over the reigns in driving the work. From the minute he started directing things instead of considering the AI work as a quirky side project, things went downhill. He tried to force a huge scale up in Llama4 which didn't work. Then as we know he disbanded the whole team and brought in a new crowd of mercenaries who may or may not have had the technical skills but they came into an organisation in disarray and still driven by Zuck himself who is continually forcing decisions that are not well founded in the science.
All the above is an entirely evidence free fan fiction version of things, but I would be completely unsurprised if it is true.
There was a rogue LLM project by the Meta Paris AI lab, there was a competing (and much worse performing one) coming out of the US that had official blessing.
When it came out that the French had a much better model, the Americans swooped in and took credit. This is was the beginning of llama.
The Frenchies were predictably pissed and left Meta over the next few years, as RSUs vested.
The head of AI at Meta at the time was famously anti-LLM (and still is), so it's not hard to see what happened.
Llama 3 was truly something special.
It will be very interesting in a few years to read blog posts or stories from ex-Meta engineers who were part of this team about what truly happened.
PSC happened...
lecunn was at facebook after llmama3.1, says ai.
There's a disconnect between measured productivity and "anecdotal" productivity. I love this chart because it also demonstrates one of the most effective ways to increase productivity: simply reducing the workforce.
https://fred.stlouisfed.org/series/OPHNFB
> it also demonstrates one of the most effective ways to increase productivity: simply reducing the workforce.
You can cut costs and increase productivity by firing everyone else and taking no salary yourself. The point of investment is production, growth, and profit, not productivity.
Output per worker is the formal definition of productivity, but that doesn't mean we should assume fixed output.
Under conditions of scarcity, it's usually beneficial to increase output or to produce different kinds of output. At least, if someone will pay for it.
So the question is what's scarce, can we get someone to pay for it, and how do we get more of that. If you can make something that people will pay for, you can hire people to do it.
Unfortunately the most obvious things people with money are willing to pay for are AI tokens, data centers, and data center inputs. It's unclear how this gets us more of other things we want.
Who cares what Zuckerberg says about AI agents? He is a PHP developer from the early '00s who got lucky with Facebook. He's not an AI scientist or an AI researcher. What authority does he have to speak on the future of AI agents? Morale at his company is at an ATL, and that says more about his leadership skills he'd better off focusing on, otherwise the agents might replace him soon.
I don't like Zuckerberg whatsoever, but there are plenty of reasons you could potentially care about his opinion:
- He could be well-placed to see whether AI agent development is successful since at a high level he oversees much of it, and will be getting detailed metrics and things.
- Whether or not his opinion is correct, tech CEOs are notorious trend-chasers, and his opinion could set the tone for other companies.
- He himself might even be a trend-follower in this case, and is presiding over an early dialing-back of AI-hype.
Honest question, is there a single successful large investment made by Meta recently ? I just know about the abject failures :)
llama has been generally useful.
llama was unintentionally released and has been a flash in the pan for them since. They've failed to capitalize on it at all post-v2. v3 models were a launch mess and quickly surpassed. The only thing it has left in 2026 is name recognition. Honestly their non-LLM things are more important and relevant, like their SAM models. https://ai.meta.com/resources/models-and-libraries/ but none of these seem to be making them serious money or exist and standalone products.
for "the rest of us" building certain kinds of systems on top of it.
as integrated into Meta's platforms it is clearly uncompetitive as a consumer chatbot with the likes of ChatGPT, DeepSeek, Microsoft Copilot, and Gemini.
Like the other chatbots give me useful answers to questions even if they're wrong sometimes. Meta AI, on the other hand, cracks jokes about my foxwork. That's no so bad because sketch comedy is one of the foxwork skill areas (e.g. if I get you to laugh you are under my spell) but it's not useful and edifying the way competitors are.
Ok, I'll bite. What's "foxwork"?
I played enough video games to be thoroughly steeped in legends about fox spirits from the Sinosphere, almost three years ago I felt I received an invitation that I was a little afraid of. a year and a half ago I read
https://www.amazon.com/dp/0231133383
and found out that people really do it. It took me about 300 days to contract with a community of foxes and started "going out as a fox". At first I didn't want to explain it to people at all but when I botched my explanation and my wife gave me some tough love about "cultural appropriation" I realized I needed a cover story and it is "I go out as a character to do street photography and make people smile"
And it became real. My son didn't believe it until he saw me work: my son saw these two college girls walking down the street and thought "there is no way they would talk to me" and next thing he knows they flagged me down and are asking me questions and I took their picture. I went to fireworks in Groton, NY and the next day in Little York [1] I haven't even parked my car and people I saw the day before are excited to see me.
When I get approached every day it becomes effortless and automatic to make an approach and when I feel like I am getting a 50% take rate I think "I am doing something wrong and I need to regroup", my usual take rate is in the 80-90% range but it feels like 100%. I don't even talk with other street photographers about it on forums because I'm drawing from an entirely different probability distribution.
Thanks to this work, it looks like I will be teaching about magic soon [2] so I was working on a short monologue about blessings on the drive up to Little York which primed me to really respond when people I photographed said simple things like "Enjoy the Fourth of July" and eating soft serve ice cream from a vendor felt like the feast that Martin Prechtel says is the only ritual -- and I am still feeling high from it just like I felt high from going to Litha put on my my pagan friend.
See https://mastodon.social/@UP8/tagged/foxwork
[1] one of the five Yorks of New York
[2] Magic is real, full stop. Science is real too, and there is nothing more scientific than the double-blind test that accounts for the placebo effect which is as I see it: "the patient experiences the treatment as receiving a blessing and feels better" and "the doctor experiences the treatment as giving a blessing and perceives the patient differently"
Ah. I was expecting some technical jargon. You do you, man...
Maybe the glasses. After other companies have failed, they may have something here.
Meta retooled all their internal models to be LEM / LLM based recently.
Meta Glasses might be the next big thing.
AR/VR is the next big thing since 2015, the cumulated revenues since then don't even match the yearly revenues they were predicting for 2020 alone. This is going to end up like 3d TVs
Any day now.
And it is even worst than Facebook and Social Media. We are doomed.
glasses maybe but not meta glasses. Its not like meta has some patent on it.
Rayban glasses with the HUD in your eyeglass will go mainstream when they get a 10 hr battery instead of the 1.5 hr battery of today.
- Whether or not his opinion is correct, tech CEOs are notorious trend-chasers, and his opinion could set the tone for other companies
exactly why I won't listen to someone who is falling behind.
Being a CEO of a large corporation doesn't guarantee you the latest tech insight. You will be easily surrounded by yes men who speak what you wanted to hear.
Agreed, all I'm saying is that his perception of the tech -- whether correct or incorrect could potentially either be trend-setting, or could be indicative of trends already in motion.
Case in point: Metaverse.
My thoughts exactly. I think he is well positioned to see edge of ai tech as the money he spent (if allocated well) could give him bit more insights. Although questionoble how much he is still able with his background to understand, I don’t think he is as much engineer as he was used to be
I like that the most sober thing a CEO has said in a long time is getting the “who cares what CEOs think?” treatment.
I'd say that about several other CEOs too if their fanboys wouldn't jump down my throat about it. Zuckerberg is just universally unpopular enough that you can get away with it
What is a popular CEO? They're generally untrustworthy people.
Gabe Newell (president, not CEO) comes to mind
I really worry about what happens to Valve as a company after he steps down.
Steve Jobs was. Elon Musk is.
Outside of tech Warren Buffett seems popular.
I don't know what the baseline is for CEO popularity (I guess most people don't know that many CEOs by name?) but I'd be hard-pressed to call a 37/50 favorable/unfavorable ratio popular: https://www.natesilver.net/p/elon-musk-polls-popularity-nate...
Maybe he's no longer popular, but he had a cameo in Iron Man 2 and hosted SNL. That speaks to popularity.
Also, is favorable/unfavorable a good measure of popularity?
Yea he only runs one of the biggest technology companies in the world.
1 ex-PHP-dev CEO != 100,000+ SWEs who actually build technology of FB/Meta/IG/Whasapp
You are correct. An individual developer's narrowly-scoped view of their work is not the same as an executive's view of the entire business.
A what ?
what does "runs" actually mean? zoom calls from his private Hawai island to a bunch of buddies other C-level execs?
It means he has a controlling share over the money and all the power.
This thread is bizarre.
He can run the day to day string pulling however he wants, he is literally the one with all the power.
Is it bizarre? Or is it you surrealist line of thinking here? What's your angle here? Man with power and money == A cornerstone knowledge base if you want know anything about AI? Wtf.
The question was, ‘who is in charge’, not ‘who I think should be, if it was my money’.
Y’all are delusional.
If you think they’re idiots, go start your own companies and eat their lunch. They got rich by eating someone else’s lunch, after all.
Right now, it is their game to lose.
Trump, Musk, Andreessen, Thiel, Biden, and Bankman-Fried all have or had a lot of power, and I would look to none of them if I wanted to get a better understanding of reality. If anything, I expect this kind of skin in the game to be strongly anticorrelated with providing useful takes; these are the people with the greatest incentives to mislead the public.
Even if zero net new insight, he has access to information on most things, not just AI, that the layperson might never have, even in the future. Not intended as praise at all (i'd say this is true for a lot of C-level in a lot of big companies).
He leads a Trillion dollar company that employs a lot of AI scientists and researchers.
I really wish people would stop using market cap as a standin for value. Investors are incredibly stupid. Meta has actual revenue to point at.
It's a concrete measure of available resources. More so than revenue.
It is precisely the opposite of a concrete measure of anything but investor valuation. Investor valuation (or even just asset value, for that matter) doesn't have much weight when figuring out how seriously to take a CEO's words. Revenue at least has a modicum of signal of competency.
Investor valuation directly correlates to capital availability both in the sense of available capital to be raised in the market and as credit lines.
And this is Meta we are talking about not Tesla, it's P/E is like 20 something. The valuation is very reasonably correlated to the earnings.
This comment reads so angry only because the OP doesn’t like what they read
All this is happening because many Silicon Valley CEOs are totally disconnected from humanity. I actually dont think Mark even sees his employees as human beings but rather as training data from a future where AI agents will just replace most of his workforce.
Regardless of wether this is right or wrong, and not even getting into the correctness of such claims, the fact is that he fired 10000 employees, from what used to be, from the outside, an engineering first company. And he sent hints to the market that more layoffs would come as agents become better.
The first problem is: it looks we still need human beings, AI is great, but its not as awesome as we initially thought.
The second problem is: all people who can leave are leaving, and those who can't are looking for a job. Humans, we need stability, a steady flow of income, and joy in what we do. By firing people and putting them in permanent observation, for an AI that will replace them, he is destroying any reason anybody would ever want to work for such a company.
Mark would make a great Lumon CEO
Be honest, in the counterfactual where he had said that "AI" agents are the next industrial revolution and there is no future for human skills and so on -- would you have still valued his opinion as being effectively worthless?
> He is a PHP developer from the early '00s who got lucky with Facebook
Nice summary of a good hacker's viewpoint. All his money, power, and other non-technical signals don't really give him the technical authority to weigh on what really makes a technical difference.
That being said, depending on how good information-sharing is at the company, he might have a valuable perspective. But given that morale is low, it's doubtful that engineers feel free to tell the straight unvarnished truth without fear. And in typical large-company fashion, managers above them likely spin things positively as much as possible.
>Who cares
He spoke at an internal town hall for Meta. His employees care...
> He's not an AI scientist or an AI researcher
Perhaps he employs some, you know, to inform him on the topic?
The same folks who informed him about how awesome Behemoth was?
Snide remarks, both of you :)
My point was not that Meta has the best AI researchers/engineers.
My point was it should not matter if Zuck is a PHP developer or whatever, instead of an AI researcher, as he employs a lot of people with those exact skills.
I made no comment about how Meta is doing in the AI race, so it’s a bit besides the point IMO, but you did make me smile :)
...and the same people who informed him about this Metaverse-Trend-thingy :-D
Unfortunately all that money also made React and Next.js a reality.
He didn't "get lucky" with Facebook. He put in the time and effort to build, dropping out of Harvard, etc. It's all very well documented.
The cold hard reality is that, despite the attempt at creating a cult of personality around the man, he's far further along his journey of computer science than being a php dev from 2 decades ago. Boiling down his credentials to being simply a dev from a bygone era is hilarious
On that note, does anyone have the authority to speak about AI agents that would satisfy your requirements? It's not exactly quantum chromodynamics, it's glorified automation that costs a lot more and does it poorly.
Nice take, but aside from all this, I think he could whoop Elon's butt in a fight.. lol
unlike sam altman who basically invented the transfomer
I thought that he invented the computer chip?
Those who've made AI their entire personality really do seem to lack the emotional fortitude to deal with someone else having an opinion.
I'm anti Zuck, but it's mostly because he's a lizard man with no morals. You are anti Zuck because he's mentioned that the promise of AI isn't going to the moon as promised. We are not the same.
The most correct opinion
No disrespect to anyone that worked or works at meta:
We can build a better version of his whole platform in 30 mins.
at that scale? They serve an awful lot of requests, so I don't think 30m will cover it.
He's incredibly rich. Americans greatly respect rich people, the richer the more respect. It's apparently a bizarre reinterpretation/misunderstanding of Calvinism.
The respect only goes as far as to listen to what they have to say. We navigate our experience based on the whims of the rich so it helps to know what they are thinking, but it certainly doesn't confer much respect. Most are completely out of touch.
I trolled there, with a kernel of truth. The neutral version is that wealth is a contributor to respect that's unusually strong in the US and I do think that it goes back to Calvinism.
My diagnosis is.. You are working with people. Smart ones at that, who have been working for a very long time in their own, niche, specific way. To assume all engineers will become AI-pilled like you and hop on Claude Code whenever asked is the wrong approach. You can make the greatest tools, but if it doesn't fit in the behaviour and the way of working of the engineer, they will simply discard it.
I see this issue with clients and prospects all the time. A client's team of 5 produces 1 foobar widget in 2 weeks. That's 50 human days spent. Then, someone demonstrates the same thing can be produced (at an equal ot higher level of quality, mind you) with AI in 2 hours. Management might celebrate but teams will continue programming by hand as they always have, now asking ChatGPT about their build tool errors instead of using Stack Overflow.
Handing out tools and telling them it's cool is not enough. You'll need to understand, work with, and guide the engineering teams properly step by step. You'll have to change their behaviour. That does not happen overnight. Unfortunately the present approach is to drop the people who don't perform in the new era of AI-assisted software engineering. That is not the right approach in my eyes.
The last two years have been perfect for accumulating tech debt.
2023 you would have probably implemented your Agents with LangChain and RAG
2025 you'd use MCP and OpenAI/Anthropic Agent SDK.
2027 you will use a workspace frameworks (Amazon, Microsoft) sensor libraries and world models.
Agents are a fantastic generational technologies, but in mid-2026 the environment they are operating in is quickly changing.
The only way forward is to stay agile, understand model and vendor risk.
>2027 you will use a workspace frameworks (Amazon, Microsoft) sensor libraries and world models.
The only people who'll be using Microsoft for anything AI are those whose employer forces them, like with Teams. All their AI offerings are overwhelmingly inferior for anything code related.
Microsoft own Enterprise if you like it or not. And Microsoft will also own a large piece of the AI in Enterprise cake. https://www.microsoft.com/en-us/frontier-company
If AI was a productivity boon: wouldn't a company employ the same or more employees to capture more market share at a competitive edge. Shedding employees because they are more efficient seems like shooting yourself in the foot to try and stand in the same spot in a race.
AI should have caused a job market boon: because less skilled employees would have been more hirable/useful. That this is not the case leads me to suspect that AI is an excuse to reduce employee count, but not the root cause.
Myopia.
This technology isn't even a decade old. It hasn't even been really useful until the last 3 years or so. Why would you expect it to be transformative already?
"Internet is just a fancy fax machine" stuff.
yup - its the j curve of productivity phenomenon in action. and we'll see smaller companies reaping benefits faster becasue they can start with AI native development.
That's exactly what Zuckerberg said a year ago or so :/
No one needs Meta AI. They have missed the boat with chatbots / agents / harnesses like Claude Code.
Can you picture anyone you know stopping using the AI services they currently use for an AI service provided by Meta?
They have such huge opportunities in selling access to their data centres to Anthropic etc, and improving their own ad models for better targeted adds (using their proprietary data and their own infra!), it is maddening to watch them try to make SOTA LLM models and harnesses no one will ever use.
I think that over long developers will desperately be needed to handle AI.
In my experience, within weeks now concepts written in stone get shattered and the next paradigm has to be used in order to max out AI in an development environment.
What is the case for AI? To handle basic work? Augment the work? Add work?
Why I think dev will be in a good spot if they adapt is the simple fact, that while laymen are using ChatGPT etc. every day, this is like driving a Tesla vs a formula 1 car.
If you take ChatGPT away from the laymen, they are helpless with IT. Devs aren't.
AI isn't static, and every turn evolves into complexity, only devs may handle when they adapt to frequent paradigm shifts and go into high level mode.
It will be again the interface between men and machine, laymen and AI. The gap won't close anytime as expected (The programming manager - remember 6 month ago?), but widens more and more.
What I see is that in day to day work many services have arms race with AI updates. The managers are more and more overwhelmed by the workload but how to automate systems is still devs' area to shine.
The business case is still hidden and unclear, but only one aspect is clear to me: low level programming is mostly configuration work now and bug fixing for AI very seldomly now.
Can't think of a better poster child of complete corporate waste that benefits no one whose assets should be seized and redistributed to the masses.
For the amount that Meta wastes on LLM spending you can pay for things like universal childcare, public community college, and providing free lunch to all public students.
If you care about things like money, look up the dollar returns on feeding children during their development or when you tell families they don't have be an economic burden for simply existing.
A better world is possible.
$80 billion written off for the metaverse.
Think about the number of kids that were harmed being fed ads and nonsense content to enable this... this a scandal IMO.
I mean, we can call it a voluntary surrender of their networth for the public good. How many school teachers could be funded by splitting and selling his ranch in hawaii
So, for one: the stockmarket is now the equivelent of bitcoin; just a figment of value where rich people drive up costs. Just like a car is _invaluable_ to you not because of it's material value but because what it does out strips it's raw goods, facebook is mostly a bunch of tiny bubbles.
So you ask yourself, _if this thing disappeared tomorrow_, what would be the actual loss. It's definitely not it's valuation.
The ranch in hawaii is a real asset...
It could certainly not fund Hawaii’s department of education for a single year. And what would you do the next year after selling it and spending the money?
Move on to the next valuable thing he owns? Sell that, rinse and repeat until it's all gone? Doesn't seem difficult..
Maybe we should start with a smaller prototype and test the theory on your or my assets. Still seem like a good idea?
use it for housing more than 1 person.
How much of your money do you spend on paying for kids school lunches, paying medical bills of terminally ill kids and paying off the student loans of graduates?
It's very easy to say that someone/some oeganization's wealth should be confiscated, yet I have yet to see those proposing it actually putting any of their own money where their mouth is.
> How much of your money do you spend on paying for kids school lunches, paying medical bills of terminally ill kids and paying off the student loans of graduates?
At least in the society I live all of those are partially paid by me through taxes.
I'm very glad to do it since the existence of kids school lunches, free healthcare (including for the terminally ill), and free universities make my life much better since society as a whole is better off. Even as an immigrant which did not use any of those services, I'm glad to do my part to pay for them, it's just the cost of a good society.
So you're just making the money that you are forced by the government to pay or face going to prison into a virtue.
Do you actually put any of your own money to help support children/sick individuals other than just getting the money forcefully taken from you and being told that it's totally going to the kids/healtchare, while 50% of it gets burned up by government beurocrats?
I specifically said "I gladly pay it", it's not the threat of imprisonment which compels me to pay my taxes. Instead it's all the benefits I see from my taxes around me: education, transportation, public amenities, healthcare, so on and so forth. Each aspect being taken care of improves society, the holistic whole is larger than the sum of its parts, there is synergy when your population is educated, don't need to spend lots in transportation, lives in great well-taken care of urban environments, can access well-maintained public parks, pools, sports facilities, don't need to be afraid of getting sick, etc.
I do actually also spend my own money in monthly charitable donations, including the UNICEF. I think it's a basic prerogative that when you make enough money for living comfortably you should also find charities you trust and support them.
> getting the money forcefully taken from you and being told that it's totally going to the kids/healtchare, while 50% of it gets burned up by government beurocrats?
You don't even know where I live to be able to say what percentage is burnt or spent in bureaucracy. It's unfortunate your view of government seems to be based on an inefficient and ineffective one, perhaps it's your experience (and it's my experience in my home country) but by being blindly ideological about it without ever experiencing a somewhat functioning government you are missing out.
People like Zuck are incredible outliers when it comes to wealth and most people are basically underwater financially. A lot of us do send some money to the food bank or the local arts initiative but the amount left after the bills for living a regular life come due is a lot less for me than a guy with over $200b and no good ideas. That's a crazy order of magnitude difference, a thousand times a thousand times 200k which most people aren't even making. For all of the pretty gross stuff it seems Bill Gates has been up to, at least he did spend a lot on stuff like mosquito nets and AIDS prevention.
> How much of your money do you spend on paying for kids school lunches, paying medical bills of terminally ill kids and paying off the student loans of graduates?
A third of my salary each month.
As in the government takes it out of your paycheck automatically, and you'll go to prison if you refuse to pay it?
Yep, and that's literally the same exact mechanism we need the Epstein-class parasites to pay into as well.
Yeah, because the solution to all the corruption is just more taxes and giving politicians more power and money.
A marginal rate of 52%. Well, technically we pay for the education rather than student loans here, because it's more cost effective without the middlemen (though we do get the occasional think tanks suggesting we change that).
Zuckerberg is proof you only need to get really lucky once. He's basically flailed at leading facebook ever since, going all in on monumentally dumb bets and erratically making headcount and project decisions, and still he somehow keeps getting richer.
Even if only for how well Instagram and WhatsApp were managed post acquisition, your message doesn’t hold up.
Most large tech companies buy late and struggle to integrate the products and teams they acquire without creating massive damage. He didn’t. Twice.
I am quite far from a fan of what he brought to the world (privacy nightmares), but to say that he’s a proof you only need to get lucky once is quite far off.
also: face tagging in photos, the fb app platform and the like button pioneered concepts that we take for granted today. also, how product is done internally, growth metrics, all borne from facebook.
The like button is an ingenious and insidious invention. It compelled every other external web property to willingly load fb sdk. It’s mind blowing how effective it is.
I am morally against Facebook, but anyone that says zucky was lucky, as a tech person, likely just sour grapes.
We take for granted a lot of things, doesn't mean they meaningfully contribute to our lives, why is face tagging or thumbs up so special? I believe we already had upvotes and reactions on other platforms like early forums and chat platforms. Even if it was the first one to popularize it, it didn't really come about in a vacuum, the Web 2.0 at that time was all about user engagement and interactions.
Zuck is a world class operator. He’s led one of the top richest companies of all time.
I think the world is worse for it but that doesn’t mean it didn’t happen. The list is just to refute the idea that it was one weird trick that gave him the keys to the planet.
and it’s not about him being first or coding it himself, it’s that it happened with him at the helm. that’s just how it goes.
> Zuck is a world class operator.
I think his recent fumbles ironically prove your point. How many people could afford to fail so many times in a row (Remember their crypto coin? VR? Now LLMs?) and still be in the big leagues?
[dead]
I don't like Zuckerberg at all but Instagram and WhatsApp purchases were _very_ good calls
Sure, they worked out pretty well, but those were 12 and 14 years ago at this point, and aren't the only calls Facebook has made over the years:
https://en.wikipedia.org/wiki/List_of_mergers_and_acquisitio...
IMO the actual prescient call he made early on is to hang on to 60% of the total voting power of the company. I have to imagine there are several points where he'd have been pushed out quite a while ago if he hadn't have done that.
But if you take the same perspective as a VC: 95% of your bets can fail if the other 5% are spectacular.
He made spectacular bets with Instagram and WhatsApp. Most CEOs don’t make spectacular bets because their business doesn’t allow it, meaning the VC analogy doesn’t hold for them.
Hasn’t Meta failed to monetise WhatsApp?
So it may be popular but they have ultimately failed commercially so far. They just bought an Indian startup and immediately put the founder in charge of WhatsApp.
Only this week they announced WhatsApp would allow usernames rather than rely on phone numbers. All the other competitors apps (signal,telegram etc) did this years ago.
Whatsapp is by far meta biggest potential. If they could turn it into a superApp like WeChat they would make a lot of money but only if that don't destroy the app.
Think that means 95% of bets over a small period of time? (A single fund within the VC) as opposed to the whole VC since inception.
And so not comparable.
VC investing is also a tax write off for rich investors which is offset against their gains elsewhere, wasted company money is a loss for the share buyer.
They spied on their mobile users with the Facebook app to detect Instagram and WhatsApp growth early
corporate espionage at global scale to get ahead of existential threats is some pretty insightful leadership, it’s all a matter of what team you’re rooting for.
By the time FB bought WhatsApp it had nearly a billion users. No need for technology conspiracy nonsense.
I guess it's hard to say how much them monitoring their (potential) competitors influenced their decisions but it's definitely _not_ conspiracy nonsense and pretty well documented:
https://en.wikipedia.org/wiki/Onavo
https://www.bbc.com/news/technology-47281906
Indeed they were.
Also, it’s interesting to note that in both cases they where two very lean companies (Instagram had 13 employees and WhatsApp had 55) with pretty a pretty large and engaged user base (30 million for Insta and 450 million for WhatsApp)
Yeah, and he threw a huge bunch of cash to acquire them. pretty easy to acquire. Harder to build. He got lucky with facebook because myspace sucked. Good timing.
I always hear reverence around Myspace. Is it really rose-tinted nostalgia? It seemed like a cool core product.
MySpace gave users a lot of freedom. That was great in certain scenarios but at it got bigger it meant there was just more noise and the general atmosphere was very messy for lack of a better word. Facebook was very clean by comparison and it felt natural to connect to people who couldn't handle MySpace, people in your family or at work etc. Then eventually it got big enough through that that it just became the place to connect with your cool friends too because "everyone" was there.
So yeah, it's kinda mixed. The core product was cool but it wasn't as suitable for as a replacement for the phone book in the way Facebook was.
Yes and then Facebook did everything they could to subvert granular privacy controls (Myspace was decent, it had a wall but that was about it) which is of paramount importance for people with real safety issues
FB would have been invented. It's just another layer of authentication anyway. Myspace was an interesting experiment for its time
...Comparisons of Tom to Zuck are rife I'm aware but somehow Myspace isn't known for "they trust me, dumb fucks"
...but most of all, Samy is my hero!
> Instagram and WhatsApp purchases were _very_ good calls
Elementary call when you have the data no-one else has thanks to illicit means and explicitly decided on a buy or bury strategy
A very HN comment... At the time, almost nobody felt spending billions acquiring these companies was "elementary". Instagram was bought so early that spending a billion on a startup was unheard of. And WhatsApp was so expensive that they questioned Zuck's sanity.
You mention two strategic enablers that made these unusual acquisitions seem obvious to FB leadership. So were those two not good calls?
The parent comment is referring to a “VPN” installed by Facebook on users phone that surreptitiously sent telemetry data to Facebook. No one else had that information.
And also even if they never had Onavo you have Zuck furiously backpedaling in an email about I'm not saaaaaaaying we should acquire IG to wipe them out from the landscape of competition, but........
And then you have the WA founder (Acton) famously saying delete facebook
Some acquisitions make the product better or are altruistic. Zuck's war chest isn't one of them
As another commenter said that he acquired both with the money made on Facebook, they didn't originate within the company.
Just because he fails publicly sometimes does not mean he only makes bad decisions. Facebook is still incredibly successful. The market may apply a Zuckerberg discount compared to other tech companies, but that is not the same thing as saying he has been a bad operator.
As a founder who has gone through scaling processes, I am quite impressed at Zuckerberg's ability to remain relatively competent and relevant at the top role throughout the whole journey. The skillset required changes quite a lot in each phase.
There are very few cases where the archetypical young garage-founder has stayed on top in the long-run.
Don't mistake tenure for competence. Meta's stock structure is such that Zuck cannot be forced out by the board or the shareholders regardless of how many mistakes he makes.
HN is so extreme sometimes. I remember when Llama was gaining popularity, everyone here was praising him for releasing open models. Same for other works in this area. Also remember praise for some VR devices.
And it's all back to Facebook, nothing in between. It doesn't look any different from general population influenced by TV and other social media.
> I remember when Llama was gaining popularity, everyone here was praising him for releasing open models. Same for other works in this area. Also remember praise for some VR devices
Neither of which have been great commercial successes. Open source LLM's and VR devices have have been cool products for HN readers, but haven't made much impact on the public or Meta's revenue.
Whilst Instagram and Facebook Ad revenue is coming in it means Zuckerberg could spend billions pivoting the company to country and western themed steak houses and the share holders wouldn't (couldn't) do anything about it.
I am not HN. I'm just one guy. And, I have not, in recent memory, said anything nice about Zuckerberg.
> and still he somehow keeps getting richer.
Just a stab in the dark, but it's likely because FB and meta are wildly successful businesses, and it's almost impossible to attribute that to one "lucky" event at the inception of the business.
If I'm being honest I'm actually kind of shocked at the engagement this comment is getting and the fact it's still visible. Like him or not, if you're capable of being objective, he is doing the opposite of failing...
You know that a lot of acquisitions is equity hire. How do you get engineers to be excited about fixing something low level in Facebook, a hard task; you don't. Better to wrap those problems in green field projects and have engineers excited, and adopt that tech back into Facebook. Google done this forever (DeepMind). Seems like the work of a genius?
Zuckerbergs comment from 2010: «We have not once bought a company for the company. We buy companies to get excellent people.»
Getting a company to IPO is not about getting lucky once - it's making hundreds of good decisions daily. From hiring to product direction to fundraising.
Like or hate the guy you gotta give credit where it's due. (Same for Musk, etc.)
There was no luck involved, just stealing and legal swindling.
A real proper true American! No wonder he gets along well with Trump
The power of network effects. Same with Elon and Twitter
The man has its flaws but to deny that he has been a very good operator is just putting your head in the sand.
- acquired WhatsApp and Instagram and grew them into the most popular apps on the planet - created an gigantic ad business that enables entire new types of SMBs to exist that rely on Meta to find their customers - which is pretty much a money volcano - bought a ton of GPUs right at the inflection of early LLM take-off
None of this was obvious when he started Facebook but obviously well executed in hindsight.
People who never built a company have the craziest takes on company building, I swear.
You talking about me?
And the guy will never ever be laid off by the board while he lays off thousands and tens of thousands each time he fails and loses billions
All thanks to Silicon Valley's ridiculous stock structures
He’s got that Epstein class luck
I feel as thought Meta, compared to other tech giants, have a vested interest in saying that AI failed, as they are the only major tech company that has almost unequivocally lost the AI race.
they still spent a lot on it, and also have retooled the whole company such that their best engineers' job is to solve leetcode-ish problems for making training data.
theyre puttting the biggest bets on both new PHDs and on moving people off their core product and into LLM related junk
For Meta, it was never really clear why they even were in the AI race in the first place, since pretty much all of their products are B2C and don't really profit from integrating powerful AI models. And for all of their internal needs, they could easily use models created by someone else, which is orders of magnitude cheaper than trying to compete on building your own SOTA model.
<laughs in Apple stock>
This article is at least the sixth restatement of a single Reuters article that has been posted here.
Because it tells people something they desperately want to be true: AI will not take their jobs and CEOs will regret trying to do so.
Zuckerberg was always excellent at knowing how to capture the attention of the internet....
> said a review of a recent data security incident with the company's controversial mouse-tracking software indicated that no employee data was included in AI training.
That's... not quite right. The employee data is used in AI training and is intended to be used this way. But despite not correctly ACLing the data for a couple weeks, it is believed it was not accessed inappropriately.
I wonder what will be the next big thing for Zuck after metaverse failure and now AI coming to nothing? Perpetual motion machines?
No need to guess: https://www.reuters.com/business/mark-zuckerberg-directed-me...
CEO Mark Zuckerberg recently dispatched a small team at his company to create a smartphone app similar to Polymarket and Kalshi, the New York Times reported on Tuesday, citing two employees with knowledge of the matter.
The app will probably rely on a video game-like points system instead of users wagering money, though the company has not ruled out betting real money eventually, according to the report.
Gambling! They are already working on a prediction market product. I could see slot machines next.
Breaking news: Meta reallocates 100,000 AI workers to trade tulip bulb derivatives
I'd almost be surprised if these large companies didn't have secret trading divisions.
A lot of companies have Bloomberg terminals for e.g. futures hedging.
Meta will continue trying to build a platform that they can control. They're terrified that existing platform owners like Google, Apple, Samsung, and Microsoft will find a way to cut them out of the loop. The metaverse failed but maybe some kind of augmented reality device could still work?
They're pushing hard for a unified platform. For WhatsApps new username feature one can only choose from usernames not already used on instagram or facebook.
Meta can unify their own products but they're still sharecroppers on someone else's farm.
They're going to try to push those stupid glasses next. Not convinced they'll succeed since it's not just the technology problem they need to solve. How do I find PMF for a product which encourages others to physically assault my users, while also not having it banned in various countries? We're talking about a societal shift that would need to happen and nobody's going to trust Meta again where that stuff's concerned.
They’re DOA
Only apple has the trust of its users to pull it off. And apple will make sure they do what they can to keep meta out
Possibly Snapchat? Snapchat already has spectacles whereas apple doesn't have their own glasses
I don't want to be a Zuckerberg hater, but is there any thing Zuckerberg has done or said in the last few years that wasn't mundane, reactionary or self-aggrandizing?
In other words, was there a single decision or take he made that turned out in his favor?
Gubernatorial run in California.
I think the reanimated corpse of Hitler endorsed by the reanimated corpse of Epstein would get more votes than the Zucc.
As long as Meta‘s ad platform keeps printing money they can afford to lose a lot on metaverse and AI.
Have they tried feeding macadamia nuts to the llama?
The company that helps connect kids with the adults who want to harm them is having a hard time replacing humans? Shocking.
They also provide a platform to terrorists who want to livestream their killing spree.
Maybe Zuck is doing the soft walk back so he can justify a few more H1Bs now that he laid off lots of expensive Americans.
Aka I thought the stuff that these other guys are doing was not so difficult. No one can replace me, of course.
Many such cases.
So what happened to Meta after those successful llama 3 model releases? They really made competing models back then. If felt like they have right people, strategy and good results. Now it feels they have neither of those…
it is amazing that meta continues to burn money and keep afloat. ads DO really make money, especially when they are going to overtake google in this [1].
to all the critics, i would suggest letting him/them cook. the snowden-era privacy concerns was exactly around ai getting trained or using personal data, and they have a treasure trove of it.
their money dump do sometimes also spend money on r&d unrelated to the current mindhive topics. we never know what the scatterbrain approach may give birth to tomorrow.
do be critical and vocal on how they squeeze money out of other areas or how they increase their revenue by ruining our lives instead.
[1] https://news.ycombinator.com/item?id=47758569
You'd have thought that Zuck's previous failures to make the things he dreams about (e.g. Metaverse, decent in-house AI) materialize might have made him a bit more cautious about betting the farm on things that don't exist, especially when he's expecting someone else (the AI agent folks) to make it happen!
I suppose you have to admire the conviction: I'll fire my developers today because REAL SOON NOW I'll be able to replace them with AGI!
I can't imagine they'd ever get enough people to opt-in to mouse tracking to generate enough data for their model training. It's either a DOA AI project, employees will be pressured into accepting it (making it part of their review), or their managers will accept the monitoring on their behalf.
I wonder when he'll admit his hopes were baseless
right after he stops trying to steal everyone's privacy. Not only on the internet but IRL too with those Meta Raybans
If a Meta employee screws up a major project, what happens? What will happen to the executives behind these mass firings and realignment - executives of one of the very top SV companies whose job is dealing with the landscape of disruptive technology development and overreacted to the latest thing? What is the standard for them?
Reminder: the "expected" result is replaced all programmers and white collar employees. Even they don't necessarily state it explicitly, that's what is in their minds.
Anything less than that is slower than expected.
with coding, you have sort of a framework for doing it right, if you have good specs, good testing practices, strict grounding in expected results deterministically, good linting, etc... this is much easier to automate with AI for the coding part within that assuming you did your homework around it... i don't have experience with all the business layers but it seems a bit more nuanced and fuzzy as you get away from that "harness" of sorts as it doesn't have to work in the same way as code for execution and evaluation... and even if code works, it still needs tastemakers in the final ok. maybe the taste maker ability still needs a lot of work/scale to be feasible, idk, like its still earlier than later on that. maybe Elon already cracked this to an extent given his automation in various companies.
By "AI agent development" does he mean using agents and such for in-house coding? Or does he mean their initiative to train their own agentic model based on the data he's having a bunch of employees generate?
Why don’t they just got Claude Fable to do it for them? Are they stupid?
Because they already have a Gemini contract.
So they are stupid it seems.
Libra, Metaverse now THE AI company? Meta seems to have a need on some strategic consultancy
> At the time, he said, executives were "super optimistic" about tools like Claude Code from AI startup Anthropic.
Some guy in sales at Anthropic has a new yacht though.
built an agent to handle my email triage. it now flags every message as "needs human review" and sends me a daily summary of how overwhelmed it is.
AI hype comes quickly, and goes with the wind quickly as well.
Facebook is already renting out their excess GPU capacity. Nvidia is giving out GPU capacity for investment stakes in startups. OpenAI is trying to get the government to take a 5% stake and needs to make the same amount of revenue within 3 years as Google does now or else they're going to default on their payments to Oracle. It may not be the same as the 'dark fiber' from the Dot-com bubble but it sure seems similar. Especially when this hardware is purchased ahead of delivery and ages so quickly it's not even clear it will be reusable after a few years. It's the scale of the investment in this stuff that is the issue. In 3 years, AI spending is off the charts compared to the railway bubble, Dot-com bubble, Roaring 20s, or really any other bubble. I don't think it's going to end well.
Isn't he just trying to tank competitors prices? Try to burst the bubble in a race he is way behind?
Echoes the same as when AWS would mean no need for IT anymore. Is there less? Yes. Gone? No.
> that executives had miscalculated on the timing of the changes
Hmmm... so who is going to be thrown under the bus?
"I was hoping AI had progressed enough so I could fire you. But you failed to make it so. Therefore, you're fired!"
tokenmaxxing will be a funny footnote like nfts on the tonight show 2 years post-hype
Or: you wasted too much money on failing to replace yourselves so now I have to lay you off. Which is one of the two possible grand outcomes of the AI bubble, which both result in laying people off, because that is all these companies know how to do as a response to stress.
I am not sure that it has to be so zero sum. The AI truth is probably somewhere in the middle; it probably doesn't replace software engineers and it probably won't be deleted as completely useless. My current feeling is that it's a powerful tool I'm happy to pay to use; it doesn't replace me, but it makes it easier to do higher quality work. It feels a lot like IntelliSense, or faster compilers, or getting a 32" monitor. That probably doesn't sustain the bubble, but it's something that people are going to be poking at and making money off of for a long time.
I agree that people are investing as though the world is going to run itself while the ultra-wealthy run off in yachts to compare sizes. If it wasn't AI, it would just be tulips or something. That's just how people are. But maybe they'll be right, who knows.
> The AI truth is probably somewhere in the middle; it probably doesn't replace software engineers and it probably won't be deleted as completely useless.
This is not really somewhere in the middle, I think. It is very close to one of the ends. Because the fear-promise to the idiot-investor class was that it would have those impacts across all industries, not just us nerds. They hate us for refusing to make their silly ideas possible and having irritating fact-based reasons why they can't work, but they don't hate us enough to spend that much money replacing just us. They have lots of other people they hate paying too, and we haven't even made a dent.
LLMs are genuinely useful tools if used as such and not some magic people replacer.
I do wonder if their provided utility will still justify the cost once the bubble pops and frontier models aren't subsidized money furnaces, though.
Yup, a nice tool.
I think we can learn from history about where this is headed. I remember when each developer needed a $600 compiler to write C++. Eventually that industry was replaced by free compilers. Kind of a similar thing going on with frontier models; I really like Fable 5, but if I didn't have any money I'd use GLM 5.2 or something. Since my employer does have money to spend on it, I enjoy the productivity boost (just like Intel's C++ compiler produced faster code than gcc for a while), but you can see the direction things are going. Eventually the moat around Anthropic and OpenAI collapses, and AI coding assistance is something built into every text editor.
There is still probably a lot of upside on betting that some lab produces a model that is light years ahead of what we have today. But I feel less "I'd better start planning for a new career" than I did a year ago.
Bottom-line win-win! All hail the shareholder value!
Who is the genius who told him development will get faster?
The man can't catch a break!
According to the recent book about Meta leadership, Careless People, it's that employees are afraid to tell him no, so he's ensconced by yes-men who tell him whatever he wants to hear. He probably has no grasp of market and product development realities.
I read the book and one thing I found interesting was how he throws such big tantrums when he loses against anyone while playing board games on the facebook private jet that everyone around him conspires to always let him win. Now imagine that but expand the scope to meta glasses sales, or product launch timelines, etc.
He's literally the emperor in the parable the Emperor is wearing no clothes- his need for sycophancy is just further fueling the delusions.
I never noticed that the Emperor was a vain man who couldn't admit his lack of wisdom (because only wise people could "see" the beautiful cloth, and he pretended to be able to see it), and that he was surrounded by yes-men (of course he was surrounded by similarly vain men who had to pretend to be wise to keep their positions, but I didn't notice how this turned them to yes-men).
Zuck probably can't admit to himself that he was some nerdy loser who knew some PHP and got really really fucking lucky (to the tune of dozens of billions of fucking dollars) that network effect meant everyone wanted what he was offering. I'm guessing he thinks those billions must be proof that he's smart... So smart that he's unbeatable at any board game.
> how he throws such big tantrums when he loses against anyone while playing board games on the facebook private jet that everyone around him conspires to always let him win
It's hard to believe that that is a real person and not a fictional person being written against some trope.
there are plenty of orgs where writing more code is not a good thing, in fact it's the last thing they want. but yet these orgs would still employ 80%+ of all developers, not necessarily to write code though.
Investor money being pissed up against the wall, for little purpose? What else is new.
I'm amazed the backers are on board. I wouldn't be. I'd rather invest in dead tech like coal, and soak every last drop of profit, than back this kind of bogus shit which is proving both more expensive, and less workable. Well, maybe not coal, but for goodnes sake, can we cut to the chase where AI investment tanks and move on with our lives?
Maybe if you spent 2 Billions in tokens instead of 1 Zucky.
AI agents are no good.
I'm sorry if it's a non sequitur but I feel even beyond superintelligence/AI/LLM whatever of the last few years... they've always done this, it's always been somewhat hamfisted
Examples abound of "I reported Nazi hate page. Didn't violate community guidelines. I called my friend a jerk, jokingly, got a month ban
For years. Not restricted to when ChatGPT et al arrived on the scene
(Because, AI in theory makes sense. If you want to monitor things at scale you might use AI - however that's defined - to make your workload easier. When is an account being hijacked? When are bad actors infiltrating the system? Or whatever)
Since when Zuckerberg is an AI expert?
I'm guessing this is specifically about Avocado which everyone at Meta would acknowledge is terrible.
I think there are seriously misplaced expectations here. The primary role of AI is transference of effort, while "increased productivity" is just a side-effect (since computers are so much faster than humans at highly repetitive tasks). It's about not having to directly do X anymore (or as often), even though it may take a few rounds to get X to a satisfactory point. But even if following up is needed, most of the effort budget can then be used for Y.
Also those with very heavy investment in AI are looking for bonkers results, which is the cause of their disappointment. They need to reduce their expectations. I for one am loving the results so far.
We're getting closer to the bubble-bursting peak. Yet another tech corp that fired a ton of experienced people hoping to squeeze more $ for a quarter, finding out AI hype is not all it's cracked up to be. Loving that for them.
Second thought, I wonder if zuckburger has turned into the Cramer of tech. First, the metaverse fiasco, now another massive overinvestment into a dead-ish end?
My instinct (for better or worse) is usually contrarian. Most people seem very skeptical of what Meta is doing with AI. But, what if, in a way at least, it makes sense?
Maybe Wang has correctly identified that the programming and agentic ability that Anthropic and OpenAI models have has largely come from armies of software engineers creating massive datasets by writing out coding and agentic problems and solutions?
So he told Zuckerberg that. The reason it may be turning into so much friction is that at companies like Anthropic or OpenAI, training engineers were either hired specifically for that purpose or probably mostly handled through contracts with third parties (which again, hired them to train AI). And honestly many of them may be overseas or just happy to have a job in a difficult period. But anyway they wouldn't have very high salary expectations etc.
But Zuckerberg already had 25000 engineers. Why not take say 1/5 of them and get them working on the the dataset? The problem is that those engineers were hired for different prestigious highly paid positions at Meta/Facebook. They were not hired to do tedious grading of AI answers or quiz construction.
But Zuckerberg either has to do this, or spend additional billions on doing it all with external contractors. A third option would be to try to create a massive distillation operation. Or just hope that his engineers could invent some magical new training trick that manifested the agentic and programming skills without the large scale human input.
Or he could release a model trained largely by existing open weights models. Which without some huge breakthrough probably has no chance of surpassing them, so is pointless.
I think most of the substantive criticism of Zuckerberg has been about burning funds. If he gives up the "your job is to grade AI homework now" plan because his engineers refuse, he would need to go through third parties. The additional billions and billions this would cost would create more pressure on the bottom line and shareholder pressure.
It would also give up any potential advantage that Wang may have optimistically sold the operation as, on that using "real" engineers as opposed to lower paid data labelling engineers might result in a higher quality dataset.
At some point, model architectures that don't need such massive datasets or can be created automatically in a way that advances the frontier will probably come about. But right now it doesn't exist.
Further, the way AI works currently, business advantage from AI comes from encoding existing internal intelligence and knowledge. Meta's massive engineering corp effectively has that in their heads. Having them create these datasets is possibly the only way to leverage this knowledge asset in this paradigm.
I guess the problem is it means forcing thousands of people to do a different job from the one they were hired for.
None of that makes sense.
What's the end goal? Meta-specific engineering, with baked-in knowledge of how FB, Threads, and WhatsApp work? General and/or coding products to compete with Anthropic and OpenAI? Some special Magic Thing which only Meta can invent which will bedazzle Meta's users?
You don't need giant datasets unless you know what you're going to do with them. OpAI and Anthropic are having enough issues making their products profitable. And those are, if not beloved, then at least respected, with a real, if patchy, reputation for usefulness.
What was Meta's pitch in this market? There were hints of interest when LeCun was still doing original R&D, and there was some distant possibility of a next-gen revolutionary product.
But now the goal seems to be to flail around doing something incoherently AI-branded with no obvious strategy.
The troops are being marched around, but no one knows where the battle is supposed to be.
Ai remains a solution looking for a problem.
Code autocomplete is a success, password reset via ai is a failure - everything else ... still busy tokenmaxxxing in search of a problem it fits into.
They are making more money than ever before. Maybe Meta leadership doesn’t really care about having a coherent strategy at this point. They can afford to flail around to see if something sticks. Reminds me of Rich kids who have ability to travel the world and find themselves before settling into a career
One problem is that the AI agent market is fiercely competitive. Why build when you can buy? For the foreseeable future there will be a number of competitive models on the "efficient frontier" and I don't think one vendor will pull ahead.
In that market you can build a model and spend a lot of money on it and at best get something that's on the same frontier as everybody else but just as likely end up with uncompetitive models like the ones they have now.
You might save a bit running your own models, doing your own inference, etc. Why not take advantage of "last mover advantage" and buy whatever is best when you need it and figure the odds are good that everybody else is going to buy more GPUs than they need and as a large customer you'll be able to buy in bulk at fire sale prices?
That makes sense in a way, but remember that Meta had previously seen some brief developer glory in the initial Llama release. Going the off-the-shelf route would essentially be giving up on being on the technology frontier in this area, and not monetizing their knowledge assets.
The most effective use of that knowledge might be feeding it into RAG instead of feeding into the base of the pyramid.
> I think most of the substantive criticism of Zuckerberg has been about burning funds.
The 2017 Rohingya massacre in Myanmar? They handed him the death toll. He filed it under growth.
>I think most of the substantive criticism of Zuckerberg has been about burning funds.
I'm not in the org myself I know some Meta SWEs tangentially. My understanding is that the biggest criticism is just the chaos of it all. Jumping constantly from one thing to another like headless chickens and accomplishing nothing.
It created an environment where it's kind of impossible to plan and progress your career.
While I mostly agree with your post, I do want to point out one thing:
> Or he could release a model trained largely by existing open weights models. Which without some huge breakthrough probably has no chance of surpassing them, so is pointless.
This seems to be categorically untrue. Composer 2.5 is a substantial improvement on its underlying Kimi base model.
If that is backed up by benchmarks then maybe they should imitate whatever Cursor did. What did they do?
They may eventually have to do that. Or they might be starting with an existing Llama model. Maybe I should have said "huge breakthrough or additional dataset".
Ahh the classic bootlicker's bias, this is how oppressed developers move and click their sad cursor hoping to keep their job while they train their AI replacement!
>"For people who are comfortable, that's great, they can contribute to this kind of great human survey. To people who are not, it is not an issue,"
Related:
Meta’s chaotic AI strategy
https://news.ycombinator.com/item?id=48523271
Meta CTO Andrew Bosworth Admits the Company's AI Reorg Was 'Atrocious'
https://news.ycombinator.com/item?id=48548461
Maybe they'd make faster progress if they worked in the Metaverse.
You always get more work done when you have no legs to worry about
Did they really think they could record all their employees screens for a couple months and one-shot the agent thing? This is like junior engineer "let's refactor this monolith" levels of delusion.
why havent big tech employees formed a union?
That's a great question. Some form of union has already started in some companies, as far as I know. But not many employees have joined those unions. Probably because most employees have high income and don't really feel like they need collective bargaining power, compared to other low-income laborers.
Poor skills at working together and love of high salaries over work place autonomy.
Fun fact: https://news.ycombinator.com/item?id=23496533
Did he announce it in the metaverse? Bro is never right.
Want with one hand, shit in the other. Which fills first?
i bet he wants some calculative shit
Or some fuzzy yet inevitably reliable shit.
The modern trend is to think intelligence is generative “like compression” or “predicting next in sequence” rather than iteratively reducing uncertainty, like those fault tolerant humans.
Compression can be defined as reducing uncertainty. If you can predict the next sequence you can compress it to 0 bytes using arithmetic coding. Reliable prediction is what enables compression and it's the link between compression and AI that everyone is talking about.
No one ever in comp sci says artificial intelligence is "like compression", they correctly state that "artificial intelligence IS compression". It's absolutely known and accepted that artificial intelligence (defined as predicting outcomes with a measure of certainty and taking chosen actions towards goals using those predictions) has equivalence to compression in a very hard science way. The hardest part of artificial intelligence is compression and the remaining part, the choice of actions based on predictions is just a tree search to a goal.
Compression in image, video, sound, and text. These items to compressed are all created by humans and we will say represented by files. The difference between an instant of reality and the files is vast. Reality also doesn’t stand still and each instant needs to be captured and interpreted before AI happens.
AI can be just like compression but currently the compute power is no match for details.
Finally these reality details need consideration in any successful implementation. Which means the implementator needs to be aware of the details and successfully relate them to everything else in the model.
I think anyone surprised by these things is not fully engaged with what they are doing.
The factor that is missing in that analysis to me is a time based dynamic stability perspective. Humans have a pretty good ability to go off the rails in reasoning one day and wake up reasonable; a pretty good ability to pursue tasks, despite a multitude of distractions, for ten years or longer. The best models get appreciably worse over a half million tokens. Even using a bunch of limited context agents over time, they lack mental stability. They keep coming up with ideas contrary to the long term idea, and every so often generate ideas that make no sense but they have a hard time letting go of. So the pure functional LLM is compression, but AGI needs some centering process, some high level of dynamic stability to stay sane over time and in the face of 10,000 shiny pretty things to chase.
The harnesses get better, but I haven’t seen much experimentation on long term stability, at least since the “let the LLM run the candy machine” papers from a while ago.
Because the thing missing, even with the largest agentic swarms, is independent intelligence, where it’s given something to own, like say “end to end data quality as we add more clients” (for a SaaS) and it just figures out what that means at each time, mutating its role and solutions to fix the external world, without getting silly.
the things as developer i see no practical gain for us when we should and found model routing better than keep pouring investment into agentic model like, i rather use a smaller faster model with harnessing to do drama script generation instead of leaving it to Fable or Gpt5.5
i blame Wang!
Wang should be CEO. Zuck is the one at fault here.
[flagged]
[dead]
[flagged]
[flagged]
[flagged]
[dead]
[dead]
[dead]
[flagged]
How does he get to decide what's "enough"? Reality will tell us, he can only place bets, whether it pans out isn't something that he has any say in.
Mark is really a bad leader with a mwah mwah vision. He is maybe correct in some things. But the execution is really really poor. Plus he does not have followers and believers. He only got money that can simulate followers to a certain extent
If it was still possible to get verbatim results from Google then I believe "mwah mwah vision" would have been an authentic Googlewhack pointing at this comment thread.
Nice thought, but I think strict googlewhacking frowned upon quote usage.
Actually, agents have been developing incredibly fast. It is just that Zuckerberg has some unrealistic expectations; the man is a lunatic.
Over the past six months or so, OpenAI's internal team has completely shifted from being heavy ChatGPT users to using Codex. Once you start using an agent like Codex, it is very hard to go back.This shift is truly transformative.
I am also aware that some of the consumer agent products on the market are growing very rapidly, such as Manus and GenSpark. Not to mention Claude Code and Codex.