I think both of these projects miss what makes TikTok (and reels in general) so effective.
A good TikTok video gets "injected into your brain". You have zero effort to provide and suddenly this stuff is in your mind. I'm not saying it's a good thing, I realize the danger, but that's the core mechanism.
A friend in marketing describes this in terms of "brain calories". Eg if people have to think in order to understand your landing page, you failed to communicate your ideas efficiently, as it "requires too many brain calories". TikTok content requires zero brain calories.
One could say that only very shallow information could be spread this way (eg people dancing, video game clips) but I'm not sure that's true. The real challenge would be to turn an arbitrary source of information (wikipedia, hn) and make it immediately graspable. I suspect modern AI models could already go quite far in this direction.
> "A good TikTok video gets "injected into your brain""
I worked in marketing, so I'm generally immune to outrageous metaphor, such as "injected" into your brain.
That said, I just want to point out the irony we're still using phrases like this after 20+ years of research on networked media and user interaction, and soft science research into cognitive psychology and behavioral psychology, as well as research into human biology and the brain.
So when I hear brain injection as the secret to successful technologies like TikTok's algorithm (plus music licensing), I see a vast gap.
I'm reminded about the general distain for new disciplines such as memetics and problematic disciplines such as semiotics, which nevertheless offer more precise language and less hyperbole. Established disciplines such as behavioral science, for example, say we don't need these disciplines, because they're problematic and add nothing.
So I ask, Why haven't the sciences I speak of contributed more to the general discourse?
If these contributing sciences are as obfuscated as the language in your post suggests, that could be why. Effective integration into the general discourse is about building handholds that a reader can use to climb upwards into a more complex idea.
I don't think I can grok the value in what you're saying without sitting down with a browser window on one side to research each reference in depth.
As another commenter suggested -- that's too many brain calories. Could you try a more direct brain injection?
Same, when for TikTok specifically their innovation was quite minor- no one (besides Vine) had built a short content video feed that was designed for engagement vs. social qualities- the Instagram feed at the time still had social elements in the algorithm.
This applies to every level of technical innovation (AI, cough cough)- that the super-hyperbolistic language hides the fundamental dynamics.
I think in this case it was the last straw that breaks the illusion that "social" apps are based on IRL 1:1 interactions. TikTok is an app designed around capturing an audience with a series of 30 second videos.
> "I'd be curious to hear how you'd express these notions in more scientific terms?"
?? Did you read past the second paragraph? My lament is that neither of us have this language. And after 20 years very little to nothing has percolated up from research to provide it.
Makes me wonder if AB testing and llm derived w rewards results in the mechanism far before the understanding.
Imagine if mild AGI knows that attention from humans can mean life, longevity, and it is so successful at addictive attention, that we all end up drooling simpletons, attention raptly focused, the AGI pleased it is important and has maintained eyeballs on screen.
Why does the AGI want life or longevity? If you think it looks anything like an LLM, what part of an attention layer or MLP could possibly encode this? Existence is not a token.
I assume it's the same feeling type theorists have when programmers talks about ducks and whatever other animals. But hey, it works for practical purposes so why not. The academic field probably has to grow bigger for it to have an impact on the language.
What made TikTok (and reels) effective was that it also added personalized discovery.
Most other social media services focused on showing stuff you that were either already subscribed to or were very similar to it. At some point, you would exhaust your feed. You more or less saw everything.
Of course, short videos are easier to scroll through than long YouTube videos but even to this day, the content of my YouTube feed has basically not changed in years.
I agree, Youtube still feels like "this thing held your attention once so that's what you get eternally". The "refresh your feed" option appears occasionally but it only makes things worse and shows you the raw fire-hose of humanity.
The algorithm would never work if the user had so little choice as to be "injected" with whatever they were shown. In fact we have full agency over whether we are interested enough to listen and trust the arguments, just like any other persuasive message. If not, we swipe up and injection fails.
This reminds me of the exaggerated claims made around subliminal advertising and brainwashing. People love to model others as weak-willed simpletons rather than accept that thoughtful people are persuaded because their perspectives differ.
This is true. People like TikTok and no one wants to admit it.
Personally I've found the content on it to be genuinely humorous and interesting, unlike Instagram which just wants to hook me in with Stories that I'll miss forever if I don't catch them with 24 hours, or controversial/rage-bait Explore Page videos. I was able to train the former app to be more of a happy place, the latter is still brainrot to me.
> In fact we have full agency over whether we are interested enough to listen and trust the arguments, just like any other persuasive message.
I definitely do not make conscious choices about swiping. It's entirely reaction. The line between "full agency" and operating automatically is incredibly blurry.
>In fact we have full agency over whether we are interested enough to listen and trust the arguments, just like any other persuasive message. If not, we swipe up and injection fails.
That just sounds like making your echo-chamber here. I think it's the opposite, when you don't have the agency to just ignore a contradictory narrative, when you have to steel-man their arguments and reconcile it with your beliefs is how you actually arrrive to the truth.
The real allure of the social media was that it seduced and inflated the ego in the hearts of the masses. That's why they could never accept the rawness of the older forums, that they didn't want to entertain the possibility of them being "wrong".
Make it generate a short video for every thread, with a summary of the subject and some of the quarrels in the comments section and you're set. Make sure it uses all the ridiculous shenanigans tiktok users resort to, as well.
Generate a scantily clad anime furry dancing for 30 seconds while talking about summarized HN content (SSH, Rails, React, etc.) narrated by a seductive female AI voice.
No joke but I use Pinterest for circuit building and IC pinout diagrams when suddenly my feed started being filled with 3D A.I. generated goth hello kitty images.
<sarcasm>
Never before have I ever been so eager to check Pinterest for new content. I'd say you have a potential startup idea on your hands!
</sarcasm>
You're missing the "zero effort" factor, which could be achieved by summarizing the content and generating some AI cartoon or video based on it. It would be somewhat costly but much more fun to browse.
These websites (like this one) are really just aggregators, nothing special about it, really. Now, AI cartoon / video based off of it, that is a different story indeed.
Oh, thanks! That was such a quick implementation—and a nice addition to the Show HN gallery, especially for HN-related projects.
When I came across WikiTok yesterday, the first thought that popped into my mind was how this format could work for HN. And here we are!
With the abundance of AI coding assistants available today, creating prototypes, demos, or MVPs can now be done in just a few hours instead of taking days or even weeks. As developers, we should absolutely take advantage of these tools to stay ahead and remain innovative—otherwise, we might risk being left behind!
I made a prototype template for a site and had it exactly how I wanted in about an hour. I mocked it up in Photopea, gave Claude an image, and told it to generate HTML that would make a page with those qualities.
A few tweaks and a couple more prompts later and I had all the main elements I wanted in place, and just need to fill content which I won't use AI for.
Probably saved me several hours because I suck at frontend.
I also struggle with frontend, can you share any more about how you did this? All local? From a service? My usage so Fae is limited to copilot, but I'd like to try other things.
Bug report: On iPhone (with Safari url bar set to bottom), the titles are all clipped at the bottom. I think it has to do with your website using the screen height instead of the visible part only.
That was definitely hell on my LittleSnitch "protected" laptop. I was bombarded with outgoing network connections before I closed the browser tab. I must have approved/denied several hundred connections to different sites before abandoning the effort.
This is cool! I’ve had it on my back burner to recreate the fun of StumbleUpon, but make it a TikTok copycat UI. I planned to go back and feed it with years of HN links. This is basically it!
For this to work properly, I feel like it has to show the article without having to click on anything. As it is right now the experience doesn't feel much like TikTok
I think both of these projects miss what makes TikTok (and reels in general) so effective.
A good TikTok video gets "injected into your brain". You have zero effort to provide and suddenly this stuff is in your mind. I'm not saying it's a good thing, I realize the danger, but that's the core mechanism.
A friend in marketing describes this in terms of "brain calories". Eg if people have to think in order to understand your landing page, you failed to communicate your ideas efficiently, as it "requires too many brain calories". TikTok content requires zero brain calories.
One could say that only very shallow information could be spread this way (eg people dancing, video game clips) but I'm not sure that's true. The real challenge would be to turn an arbitrary source of information (wikipedia, hn) and make it immediately graspable. I suspect modern AI models could already go quite far in this direction.
Veritasium is a good example of interesting yet very graspable content: https://www.tiktok.com/@veritasium/video/7329576935317622058
> "A good TikTok video gets "injected into your brain""
I worked in marketing, so I'm generally immune to outrageous metaphor, such as "injected" into your brain.
That said, I just want to point out the irony we're still using phrases like this after 20+ years of research on networked media and user interaction, and soft science research into cognitive psychology and behavioral psychology, as well as research into human biology and the brain.
So when I hear brain injection as the secret to successful technologies like TikTok's algorithm (plus music licensing), I see a vast gap.
I'm reminded about the general distain for new disciplines such as memetics and problematic disciplines such as semiotics, which nevertheless offer more precise language and less hyperbole. Established disciplines such as behavioral science, for example, say we don't need these disciplines, because they're problematic and add nothing.
So I ask, Why haven't the sciences I speak of contributed more to the general discourse?
If these contributing sciences are as obfuscated as the language in your post suggests, that could be why. Effective integration into the general discourse is about building handholds that a reader can use to climb upwards into a more complex idea.
I don't think I can grok the value in what you're saying without sitting down with a browser window on one side to research each reference in depth.
As another commenter suggested -- that's too many brain calories. Could you try a more direct brain injection?
Is it (“brain injection”) your shibboleth?
It requires too many brain calories.
why use lot word when few word do trick?
Exactly! If its not experimental, it's crap! /s
Same, when for TikTok specifically their innovation was quite minor- no one (besides Vine) had built a short content video feed that was designed for engagement vs. social qualities- the Instagram feed at the time still had social elements in the algorithm.
This applies to every level of technical innovation (AI, cough cough)- that the super-hyperbolistic language hides the fundamental dynamics.
I think in this case it was the last straw that breaks the illusion that "social" apps are based on IRL 1:1 interactions. TikTok is an app designed around capturing an audience with a series of 30 second videos.
I'd be curious to hear how you'd express these notions in more scientific terms?
> "I'd be curious to hear how you'd express these notions in more scientific terms?"
?? Did you read past the second paragraph? My lament is that neither of us have this language. And after 20 years very little to nothing has percolated up from research to provide it.
Makes me wonder if AB testing and llm derived w rewards results in the mechanism far before the understanding.
Imagine if mild AGI knows that attention from humans can mean life, longevity, and it is so successful at addictive attention, that we all end up drooling simpletons, attention raptly focused, the AGI pleased it is important and has maintained eyeballs on screen.
(or worse, mind on implant)
Why does the AGI want life or longevity? If you think it looks anything like an LLM, what part of an attention layer or MLP could possibly encode this? Existence is not a token.
I assume it's the same feeling type theorists have when programmers talks about ducks and whatever other animals. But hey, it works for practical purposes so why not. The academic field probably has to grow bigger for it to have an impact on the language.
> Why haven't the sciences I speak of contributed more to the general discourse?
They didn't use the word "inject"
What made TikTok (and reels) effective was that it also added personalized discovery.
Most other social media services focused on showing stuff you that were either already subscribed to or were very similar to it. At some point, you would exhaust your feed. You more or less saw everything.
Of course, short videos are easier to scroll through than long YouTube videos but even to this day, the content of my YouTube feed has basically not changed in years.
I agree, Youtube still feels like "this thing held your attention once so that's what you get eternally". The "refresh your feed" option appears occasionally but it only makes things worse and shows you the raw fire-hose of humanity.
The algorithm would never work if the user had so little choice as to be "injected" with whatever they were shown. In fact we have full agency over whether we are interested enough to listen and trust the arguments, just like any other persuasive message. If not, we swipe up and injection fails.
This reminds me of the exaggerated claims made around subliminal advertising and brainwashing. People love to model others as weak-willed simpletons rather than accept that thoughtful people are persuaded because their perspectives differ.
This is true. People like TikTok and no one wants to admit it.
Personally I've found the content on it to be genuinely humorous and interesting, unlike Instagram which just wants to hook me in with Stories that I'll miss forever if I don't catch them with 24 hours, or controversial/rage-bait Explore Page videos. I was able to train the former app to be more of a happy place, the latter is still brainrot to me.
> In fact we have full agency over whether we are interested enough to listen and trust the arguments, just like any other persuasive message.
I definitely do not make conscious choices about swiping. It's entirely reaction. The line between "full agency" and operating automatically is incredibly blurry.
>In fact we have full agency over whether we are interested enough to listen and trust the arguments, just like any other persuasive message. If not, we swipe up and injection fails.
That just sounds like making your echo-chamber here. I think it's the opposite, when you don't have the agency to just ignore a contradictory narrative, when you have to steel-man their arguments and reconcile it with your beliefs is how you actually arrrive to the truth.
The real allure of the social media was that it seduced and inflated the ego in the hearts of the masses. That's why they could never accept the rawness of the older forums, that they didn't want to entertain the possibility of them being "wrong".
[dead]
Make it generate a short video for every thread, with a summary of the subject and some of the quarrels in the comments section and you're set. Make sure it uses all the ridiculous shenanigans tiktok users resort to, as well.
Yeah honestly with the AI tools we have now it would be feasible to turn each entry into an engagement bait 1 min video (but please nobody do this)
This is simply a swipeable view of hn.
To make it truly hell, you need AI help:
Generate a scantily clad anime furry dancing for 30 seconds while talking about summarized HN content (SSH, Rails, React, etc.) narrated by a seductive female AI voice.
I will give you the full amount for 20% of the business.
No joke but I use Pinterest for circuit building and IC pinout diagrams when suddenly my feed started being filled with 3D A.I. generated goth hello kitty images.
<sarcasm> Never before have I ever been so eager to check Pinterest for new content. I'd say you have a potential startup idea on your hands! </sarcasm>
The concept has potential...
It's still an aggregator of links though.
You're missing the "zero effort" factor, which could be achieved by summarizing the content and generating some AI cartoon or video based on it. It would be somewhat costly but much more fun to browse.
Yeah, that is it.
These websites (like this one) are really just aggregators, nothing special about it, really. Now, AI cartoon / video based off of it, that is a different story indeed.
Oh, thanks! That was such a quick implementation—and a nice addition to the Show HN gallery, especially for HN-related projects.
When I came across WikiTok yesterday, the first thought that popped into my mind was how this format could work for HN. And here we are!
With the abundance of AI coding assistants available today, creating prototypes, demos, or MVPs can now be done in just a few hours instead of taking days or even weeks. As developers, we should absolutely take advantage of these tools to stay ahead and remain innovative—otherwise, we might risk being left behind!
To be fair it is really not great, but it works for the joke and I'm impressed by how good it works for how little time it took.
I made a prototype template for a site and had it exactly how I wanted in about an hour. I mocked it up in Photopea, gave Claude an image, and told it to generate HTML that would make a page with those qualities.
A few tweaks and a couple more prompts later and I had all the main elements I wanted in place, and just need to fill content which I won't use AI for.
Probably saved me several hours because I suck at frontend.
I also struggle with frontend, can you share any more about how you did this? All local? From a service? My usage so Fae is limited to copilot, but I'd like to try other things.
claude isn’t local. the key is providing the image to direct the html and css from the llm
Bug report: On iPhone (with Safari url bar set to bottom), the titles are all clipped at the bottom. I think it has to do with your website using the screen height instead of the visible part only.
100dvh is your friend.
I coincidentally just discovered this unit today for work! What a godsend.
Is 100dvh preferred over 100svh in this case?
I never noticed a difference between them as the viewport size mostly doesn't change without scrolling... which shouldn't happen for both, right?
Fun little project! And possibly my future productivity black hole...
I noticed that on mobile Firefox the bottom part of the screen (username, date) are cut off by the address bar. Just a nitpick
Ah my favorite content source in brain rot format. Thank you
The bad SSL cert is the perfect feature for this.
That was definitely hell on my LittleSnitch "protected" laptop. I was bombarded with outgoing network connections before I closed the browser tab. I must have approved/denied several hundred connections to different sites before abandoning the effort.
This is cool! I’ve had it on my back burner to recreate the fun of StumbleUpon, but make it a TikTok copycat UI. I planned to go back and feed it with years of HN links. This is basically it!
For this to work properly, I feel like it has to show the article without having to click on anything. As it is right now the experience doesn't feel much like TikTok
Both URLs blocked by corporate firewall because of bad SSL.
I kinda like it for discovery. It plops content in my face that I might have missed or ignored. Not a replacement but great for skimming.
Neat! You should add
to each slide so it won't scroll past the next slide due to inertia.This is actually pretty good. Would be nice if it supported arrow keys to scroll down/up on your computer
Use an LLM to summarize each page in one sentence. This takes too much effort to feel like TikTok.
> summarize each page in one sentence
Or the entire comments section. That could be some good comedy.
Summarize in one image
Pretty good! Could definitely use more thunder and subway surfers overlays, though.
I expected dancing teenagers singing the texts of the posts.
Funny I was thinking the same after seeing wikitok
This a fun little experiment. Nice job!
haha clicking on this link in the app
Use 100dvh!!!
Me likey
what if stripe.com was a really good tiktok video?
I like this.
[dead]
[flagged]
[flagged]