I think this all rings true for where we are right now. The trend is that the agents are becoming superhuman in tasks for which there is a verifiable reward, and analysing a business problem, identifying inefficiencies and turning it into a software specification is not one of them.
However, things are changing so rapidly that I can see that starting to change as well. But it would take much better learning efficiency to understand unknown domains, 100% computer use reliability etc. I suspect we’ll see this by the end of the decade.
The way I see it is AI collapses hierarchy. Abstraction layers will become more flat both in software and in society. Because ultimately software and layers of heirarchy in society exists to serve a function. But now those functions are being replaced.
Imagine this you used to need a library for common things in your software project. Even if you just need one function but because it was easier to just import a library that would have been the standard practice. But now the AI will just go “I can just implement that thing you need in 10 lines”. You used to need things like react native or flutter if you wanted to build cross platform apps. Now not anymore you just need to tell the LLM and it does it in both, and you get better results too.
In society we also had all these layers of abstractions and hierarchies that we used to need but will become more and more irrelevant collapsing the hierarchy.
There is a saying “as above so below, as below so above” I think this applies here. It will propagate all through our social construct, software, society.
I am honestly _astonished_ that the current crop of agents are not all written natively for every platform.
I tried porting a moderately complex workout app that I’ve built over years (and which includes a whole agentic loop) from web to iOS native. It took me a couple of evenings.
There is now no excuse to not offer a native experience for every supported platform when you are a bigger company.
Make one of the platforms using good coding practices. Vibe code the others from the source using a proper test battery.
Because it’s too complex and not worth the effort. Most users will not care about or perceive the benefit. If they delegate that responsibility to Google via Electron, they can focus on features/bug fixes that move the needle.
What effort ? I’ve started seeing this attitude more and more. We had a thing that took a week to do before. Now it takes a morning. So we refuse to spend half a day polishing it because it’s too annoying.
Once the initial port is done, keeping platforms in sync is fast. Not to mention that a harness that would automate this would be valuable.
I run codex on an old rpi, it eats 190M of ram for what essentially is a telnet client.
I have a question about web to iOS native. How do you distribute that app? I’m assuming you’re paying $99 yearly to distribute through the App Store? Is this app meant just for you or for anyone to find?
Reason I’m asking is because I’d like to make native iOS apps just for myself rather than progressive web apps. But I haven’t understood the best way to distribute.
> You need audit, security, maintenance and accountability.
I think this is where the role of humans will move towards in the future. Things like accountability can't really be outsourced to AI.
This relates to solving the alignment problem: if you give the AI a specific goal and it has to plan sub-goals, how can you be sure the sub-goals align with your interests.
For tasks in an isloated environment, this doesn't really become an issue. You can give a sandbox the least privileges it needs to do the job you want it to do.
It becomes a problem if you give it open-ended access to systems that connect to the real world and which can have real consequences. There have been stories about openclaw deleting someones email inbox because it though that was what the owner wanted. AI taking control of public wikis to coordinate, which was recently posted is another one.
For software concerns with potential real catastrophic consequences are security, durability, availability. I think for future software systems these are the concerns where you need to limit the AI in a way that it cannot circumvent guarantees that you give around these concerns.
Minor nit:
> AI doesn’t change the question: it creates new choices and moves the thresholds.
This cause an adverse reaction when I read the post. This was probably not written by an LLM since the rest of the article doesn't look like it, but maybe in the future we need to all be more concious about leaving LLM-tells out of our writing.
Of those four, I think accountability is truly the only moat. The other three are varying amounts of both testable and iterated on via adversarial agents steel-manning the implementations.
Accountability will not come before AIs achieve legal personhood, and that probably will not happen in my lifetime. And if it does in some jurisdictions (which I am not betting money on, this is a full full AGI scenario after multiple more philosophical goalposts move first), I will completely bet it will not be a global recognition.
I think most developers and most dev work is now in the field of websites, apps, and other stuff that is by nature connected to the internet, and has some relatively monolithic backend that has not been built with "blast doors" or other internal safeguards. I guess it will take considerable time until most software is in a state where the impact of AI changes can be reliably estimated without looking at the changes themselves, and some patterns for how to achieve this probably still need to be discovered.
> maybe in the future we need to all be more concious about leaving LLM-tells out of our writing.
I expect the gruesome writing style of LLMs to be fixed relatively soon, seeing how it can already be prevented with just a few lines of stylistic instructions.
I just wonder: how long will it take us then, after the fix, to not be thrown off every time we see some unnecessary catchy or contrastive phrasing?
Being traumatized by poor writing was not on my bingo card for 21st century technological progress ...
A sane article. Looks like Ben Evans has real corporate experience and understands how things work in big corporations.
Giving people AI is just like giving people Google Wave (https://en.wikipedia.org/wiki/Google_Wave), which can do pretty much anything collaboratively, ended up doing nothing.
A lot of the long tail tools are used so rarely or for some very specific functions in certain organisations that there is a good case to be made that these can definitely be targets of simpler GPT guided automation.
There are a lot of redundancies within the tools and the pricing is such that one cannot do much about it, generally some companies pay for the brand, and some tools like SAP are integral to companies of certain size. A sufficiently integrated AI tool that learns the workflows might actually be able to find optimisations here as well, but definitely auditability, testing and other concerns will remain and this is what might become the USP of SAAS providers.
A lot of outsourced IT services jobs in India and other countries are cheap hourly wages for a lot of people maintaining these, certainly a lot of these will be under threat
While I do not believe AI to be a panacea and the non deterministic nature and costs once the scale keeps growing means the integration will be gradual. I am still excited for it to define what an organisation is and what do a lot of people actually do especially in fields like accounting etc. where repetitive work is billed at quite high rates.
Law etc. is a field where the gatekeepers might hold on much longer by adding more ridiculous rules and logic. Ultimately humans have decided what is constitutional and what is legal and subjective rules are what maintain human power.
I feel the right model is smart domain experts of humans making strong and useful harnesses that help AI be effective with the workflows of the organisation, however this might be the biggest fear of middle managers who will never let it happen easily
Accountants get to charge high rates largely because good advice saves their customers large amounts of tax. When we're talking tax accountants of course. There's all sorts of accountants though - audit, management, forensic etc.
I think "any apps" and probably "most apps" is too strong. It wouldn't be very productive to hand-roll your own filesystem, browser, cryptographic library, etc. At some point you have to delegate to another party. No person or corporation has the time and expertise to own and maintain every layer of the stack.
I think this all rings true for where we are right now. The trend is that the agents are becoming superhuman in tasks for which there is a verifiable reward, and analysing a business problem, identifying inefficiencies and turning it into a software specification is not one of them.
However, things are changing so rapidly that I can see that starting to change as well. But it would take much better learning efficiency to understand unknown domains, 100% computer use reliability etc. I suspect we’ll see this by the end of the decade.
The way I see it is AI collapses hierarchy. Abstraction layers will become more flat both in software and in society. Because ultimately software and layers of heirarchy in society exists to serve a function. But now those functions are being replaced.
Imagine this you used to need a library for common things in your software project. Even if you just need one function but because it was easier to just import a library that would have been the standard practice. But now the AI will just go “I can just implement that thing you need in 10 lines”. You used to need things like react native or flutter if you wanted to build cross platform apps. Now not anymore you just need to tell the LLM and it does it in both, and you get better results too.
In society we also had all these layers of abstractions and hierarchies that we used to need but will become more and more irrelevant collapsing the hierarchy.
There is a saying “as above so below, as below so above” I think this applies here. It will propagate all through our social construct, software, society.
I am honestly _astonished_ that the current crop of agents are not all written natively for every platform.
I tried porting a moderately complex workout app that I’ve built over years (and which includes a whole agentic loop) from web to iOS native. It took me a couple of evenings.
There is now no excuse to not offer a native experience for every supported platform when you are a bigger company.
Make one of the platforms using good coding practices. Vibe code the others from the source using a proper test battery.
Because it’s too complex and not worth the effort. Most users will not care about or perceive the benefit. If they delegate that responsibility to Google via Electron, they can focus on features/bug fixes that move the needle.
What effort ? I’ve started seeing this attitude more and more. We had a thing that took a week to do before. Now it takes a morning. So we refuse to spend half a day polishing it because it’s too annoying.
Once the initial port is done, keeping platforms in sync is fast. Not to mention that a harness that would automate this would be valuable.
I run codex on an old rpi, it eats 190M of ram for what essentially is a telnet client.
You're mistaken if you think it has ever taken a week to port (and support) an application to native primitives.
This isnt how a company which has "solved programming" would act.
Especially since users do prefer a native experience and having their RAM conserved.
I have a question about web to iOS native. How do you distribute that app? I’m assuming you’re paying $99 yearly to distribute through the App Store? Is this app meant just for you or for anyone to find?
Reason I’m asking is because I’d like to make native iOS apps just for myself rather than progressive web apps. But I haven’t understood the best way to distribute.
If just for yourself, get an Apple Dev account and use TestFlight
Do I need to upload a new build of the app every 90 days? I’d prefer the app to just work forever without thinking about it.
> You need audit, security, maintenance and accountability.
I think this is where the role of humans will move towards in the future. Things like accountability can't really be outsourced to AI.
This relates to solving the alignment problem: if you give the AI a specific goal and it has to plan sub-goals, how can you be sure the sub-goals align with your interests.
For tasks in an isloated environment, this doesn't really become an issue. You can give a sandbox the least privileges it needs to do the job you want it to do.
It becomes a problem if you give it open-ended access to systems that connect to the real world and which can have real consequences. There have been stories about openclaw deleting someones email inbox because it though that was what the owner wanted. AI taking control of public wikis to coordinate, which was recently posted is another one.
For software concerns with potential real catastrophic consequences are security, durability, availability. I think for future software systems these are the concerns where you need to limit the AI in a way that it cannot circumvent guarantees that you give around these concerns.
Minor nit:
> AI doesn’t change the question: it creates new choices and moves the thresholds.
This cause an adverse reaction when I read the post. This was probably not written by an LLM since the rest of the article doesn't look like it, but maybe in the future we need to all be more concious about leaving LLM-tells out of our writing.
Of those four, I think accountability is truly the only moat. The other three are varying amounts of both testable and iterated on via adversarial agents steel-manning the implementations.
Accountability will not come before AIs achieve legal personhood, and that probably will not happen in my lifetime. And if it does in some jurisdictions (which I am not betting money on, this is a full full AGI scenario after multiple more philosophical goalposts move first), I will completely bet it will not be a global recognition.
Why would the AI need to be accountable? Make the org that sells the tokens accountable.
This is like saying "writing code is solved, all we need now is somebody to blame because it doesnt work"
I think most developers and most dev work is now in the field of websites, apps, and other stuff that is by nature connected to the internet, and has some relatively monolithic backend that has not been built with "blast doors" or other internal safeguards. I guess it will take considerable time until most software is in a state where the impact of AI changes can be reliably estimated without looking at the changes themselves, and some patterns for how to achieve this probably still need to be discovered.
> maybe in the future we need to all be more concious about leaving LLM-tells out of our writing.
I expect the gruesome writing style of LLMs to be fixed relatively soon, seeing how it can already be prevented with just a few lines of stylistic instructions.
I just wonder: how long will it take us then, after the fix, to not be thrown off every time we see some unnecessary catchy or contrastive phrasing?
Being traumatized by poor writing was not on my bingo card for 21st century technological progress ...
A sane article. Looks like Ben Evans has real corporate experience and understands how things work in big corporations.
Giving people AI is just like giving people Google Wave (https://en.wikipedia.org/wiki/Google_Wave), which can do pretty much anything collaboratively, ended up doing nothing.
A lot of the long tail tools are used so rarely or for some very specific functions in certain organisations that there is a good case to be made that these can definitely be targets of simpler GPT guided automation.
There are a lot of redundancies within the tools and the pricing is such that one cannot do much about it, generally some companies pay for the brand, and some tools like SAP are integral to companies of certain size. A sufficiently integrated AI tool that learns the workflows might actually be able to find optimisations here as well, but definitely auditability, testing and other concerns will remain and this is what might become the USP of SAAS providers.
A lot of outsourced IT services jobs in India and other countries are cheap hourly wages for a lot of people maintaining these, certainly a lot of these will be under threat
While I do not believe AI to be a panacea and the non deterministic nature and costs once the scale keeps growing means the integration will be gradual. I am still excited for it to define what an organisation is and what do a lot of people actually do especially in fields like accounting etc. where repetitive work is billed at quite high rates.
Law etc. is a field where the gatekeepers might hold on much longer by adding more ridiculous rules and logic. Ultimately humans have decided what is constitutional and what is legal and subjective rules are what maintain human power.
I feel the right model is smart domain experts of humans making strong and useful harnesses that help AI be effective with the workflows of the organisation, however this might be the biggest fear of middle managers who will never let it happen easily
Accountants get to charge high rates largely because good advice saves their customers large amounts of tax. When we're talking tax accountants of course. There's all sorts of accountants though - audit, management, forensic etc.
It doesn't look like ai will enhance existing tools. AI allows everyone to build their own. Or at all eliminate necessity of any apps
I think "any apps" and probably "most apps" is too strong. It wouldn't be very productive to hand-roll your own filesystem, browser, cryptographic library, etc. At some point you have to delegate to another party. No person or corporation has the time and expertise to own and maintain every layer of the stack.
This article is a direct rebuttal of that idea.