I am not a big fan of AI-generated educational content, mostly because it's a great way to confidently learn falsehoods and misconceptions. I would prefer to learn from a reliable and reputable source.
I am also not a big fan of trying to beat doomscrolling. One of the defining properties of doomscrolling is that it is mindless and addicting. The moment you try to create a mindful, healthy alternative, you've already lost. No product will ever beat doomscrolling, only individuals dedicated to their own mental health are capable of clearing this hurdle.
This was a couple of years ago, but I remember using ChatGPT to try and study for a certification by generating quiz questions.
It would always start to make every correct answer option "C" over time, no matter what I tried. Eventually I was so focused on whether or not it was stuck in a "C" loop that I started overthinking all of the questions and wasting time.
Flash forward to testing Sonnet 4.6 recently to try and see if it could effectively teach me something new, I got about 5 prompts in before I had to point out an oversight, and it gave me the classic "you're absolutely right, ignore that suggestion".
This is anecdotal of course, but at least LLMs are helping to build my skills of fact verification and citation checking!
This is neat. I did see one bug early in the conversation: "The submission shows an escaped/unfinished string: find . -name \." which looks like an error converting what the user chose back to an escaped string for the AI.
For feedback:
* Please change the colours. I can't read white text on light purple. By all means keep purple, just increase the contrast.
* Maybe keep it in a narrower column, again for readability especially since we're constantly switching sides left to right as we read the AI text and the 'You submitted' text. I also feel it's quite duplicated, I can visually see which option was chosen yet it's restated in 'You submitted'. Only show what's necessary.
* In the replay you shared I can still change active radio buttons. Make it static.
To more focused feedback:
* I really like the idea. Instead of an AI telling you the answer, you learn -- this is fantastic.
* Multiple choice questions are great, but eventually please move to freeform. An AI can evaluate that. (For example, after learning a bit via examples, it might prompt, 'How do you find all PDFs in your home folder?' and give a text entry field. In the past, you'd have to type exactly the string the app expected; with an AI, it can evaluate flexibly. So take advantage of this. It's an immense opportunity vs earlier learning platforms.) Also, people learn better at some point when you're not given answers (aka multiple choice) but need to write.
I think you're on to something. AI as thinking replacement is a worry. AI as a guided teacher is great. I look forward to what you do with it.
Re the more focused feedback, I totally agree re the questioning styles. In the prompt I ask for it to not do so many multiple choice questions, but I think it is addicted/the conversation history skews the context.
I'm going to introduce a settings panel (easily accessible during the conversation), which will let you move to "chat mode" (to discuss instead of be asked questions), and also to configure the types of questions you're asked and the ratio (if I can get the llm to oblige). I'm also going to see if I can come up with some different question formats beyond multiple choice, free-form and multi select (which the llm doesn't use too much).
My feedback reiterates what you're hearing about the cogintive load being unruly. If you want to "beat" doom scrolling, you need your mechanics to be as simple -- quick swipes and taps.
One potential direction, simple cards that are True/False.
If you want to beat doom scrolling, put your phone down?
Directed learning towards some goal is always going to be useful. These replacements for doom scrolling are addressing a surface level issue. Just change your habit to be spaced repetition on flash cards.
An interesting approach, Good luck with it!
A nit to pick: find is not a bash command. You can run it for example from a Windows DOS command line as:
wsl find ...
You can run all Linux commands this way. Also, pretty sure that find's "-o" is the Boolean "or", not "otherwise". (Yet another example of why learning from LLMs is dangerous, I suppose).
otherwise is actually an accurate description of what -o actually means. It means do the thing on the left if it works, otherwise do the thing on the right. That is, if the clause on the left succeeds then the clause on the right will be ignored.
A naive interpretation of or in the light of Boolean algebra would be: do both and return true if either succeeds.
The example you gave is just so buggy; why do you think that this is worth sharing with the world yet?
The very first question is full of obvious bugs.
You have 'find . -name "notes.txt"' selected, it then says 'You submitted: ls -R | grep notes.txt: find . -name "notes.txt"', then it responds:
'Thanks — your answer looks like it was partially entered.
'You picked find . -name "notes.txt" (good choice). The submission shows an escaped/unfinished string: find . -name \. The correct full command is find . -name "notes.txt", which searches recursively from the current directory for files or directories named exactly notes.txt.'
There seems to be some weird kind of quoting issue going on there. I would fix obvious issues like that before sharing this with the world.
> why do you think that this is worth sharing with the world yet?
It's worth sharing because it's cool, even if it's not perfect yet!
You'll never make every single person on HN happy. But if you share your stuff early and make one person happy at the very least that means you should keep working on it!
Don't let perfectionism get in the way of good enough :]
Show HNs never pleased everyone, and it'd be silly to try, but until recently there was a bit of a "it's not perfect... but the person has spent more time working on this than I have even spent thinking on the problem" kind of expectation whereas now many of them feel like the comments section ends up doing more thought about the submission than was put into refining it.
It'll be really interesting to see where this settles. In the meantime, erring on the side of kindness tends to work best!
Personally, I found this idea cool, and a fresh change from the onslaught of clawdbot posts on this platform. It was enough to make my time on this site a little bit more enjoyable. There's no need to nitpick to the point of bringing down another human for their work.
Sorry, I shouldn't have been that harsh about it; I apologize. Doesn't look like I can delet the comment, though.
Anyhow, I'm just a bit bitter about the onslaught of obviously vibe coded projects that very little effort has been put into. And OP did ask for feedback. But I shouldn't have been so mean about it. I know it can be fun to be able to whip up something quicker than you could before, but I really wish people would spend a tiny bit more effort on it before asking for feedback.
Hey, I appreciate your views (both positive and negative). There actually is a bit of a back story behind it all. I spent a month working on a different version of rebrain, (still available here: https://dev.rebrain.gg). After putting that work in I started to show it to some friends. It was clear from how they responded that it was not as promising as I had hoped. I listened to their feedback and came up with a different idea for LLM-based education (which is the version that's live now). I did vibe code it in about 3/4 days. But I vowed to try to get feedback sooner rather than later, which is why I posted it on ShowHN pretty early in its development. I do want to improve it... so please let me know if there are things which really frustrate you.
I checked the website and you weren't harsh enough. It reads more like a post modern art about vibecoding, some kind of performative art than anything.
Does this form based interaction offer any specific advantages over chatting directly with ChatGPT or Claude? Essentially the site is still powered by their APIs and it is just the interface that is different.
I like something here. The interactive - incremental steps of learning or teaching. But having to go through sort of quizes to just get info I need - may not be best for all use cases.
There is a reason chat bots work - it mimics natural human interaction.
But this can be interest pattern - for say AI driven personal tutor for math topic
Lastly - for love of god - pls do something with the UI - all colors and bubbles and I am totally lost just trying to make sense of what is going on. Look at reditt if you just want back and forth thread convo style. Life is simpler that way.
It would be great to be able to kick of a chatbot convo to go deeper on something each time you have answered. Chat for a bit to deepen understanding and read around, then pop back out and continue to the next question when your satisfied you've understood the concept.
Agree on colours. The Electronics topic was white on yellow, completely unreadable.
I thought this would teach you random facts in one-screen-sized tiles. This doesn't look like doom scrolling. But I like the idea of the user picking a subject, maybe combine both: pick a subject, it gives you facts about it, maybe a tile to ask for which path to take once in a while (swipe left to keep learning about find, swipe right to learn about xargs).
Maybe if you can deliver information in small chunks in an interesting way that doesn't require a large attention span, it will work.
But if I have to read a wall of text, it's not an alternative to doomscrolling, it's an alternative to reading a book or documentation. And in that case I'd rather read a book or documentation.
Great idea. It sadly doesn’t work for me, I went to pick a topic and some of the answers were pre-submitted, some of the questions did nothing after clicking “check”, then it just ended. Maybe it’s the HN hug of death?
That's only visible when it's not your personal conversation (you can't interact with someone else's). In a way it's designed to be distracting so you know how to start your own conversation.
I think it's a cool idea and I personally find using LLMs as a teaching tool to still be the most rewarding way of interacting with them, if done right.
Your obvious first port of call IMO is correctness of material, where there's room for improvement [1]. I deliberately picked Gleam because it's still a less known language.
For what it's worth, prompting Opus 4.6 in chat got me this result [2]. Sonnet 4.6 via the Workbench also got it right.
Agree with other comments about UX and design, and maybe also some of those around improving teaching style or gamification aspects, but the above is more important.
Yes, at the moment it's an issue of cost. I can't use the best models because it is not affordable. Hopefully as performance improves over the years this will become less of an issue. Maybe I can build in a websearch to verify info though...
I hear you. Yes, I think "seeding" an LLM with docs or other learning material is one of the fundamentals of effectively and efficiently using it for learning, maybe you can build more in that direction?
Yeah I actually started there. Https://dev.rebrain.gg has the old version up.
You upload a source and it generates questions from it. However when showing it to friends I found that the barrier to usage was too high as most people don’t have a source ready. But I think adding it as an option would be pretty cool and doable
Great idea! I too forget the many switches that the 'find' command can use. Anything that teaches and helps one understand the core concepts of any subject is always a good idea. I'm just not a fan of creating an account and password. I have way, too, many of those. Good luck :)
Thanks very much for the feedback. If you have any ideas of how to communicate the retention mechanisms please let me know. If you want to discuss further, please leave feedback on the site (you can do this if you're signed in) and mention you are nivcmo from hn and I'll be in touch via email.
the gap we kept hitting when building 45d was exactly this - there's a real difference between "AI that retrieves content" and "AI that understands how this specific person learns." the former is a better search engine. the latter is something new.
our learner model tracks 500+ signals per person - not just difficulty preferences but how they engage with confusion, where they disengage, what formats actually create retention for them specifically. the course structure changes as a result, not just the difficulty level.
curious what you found on the memory/context side. longitudinal context across sessions is where most AI tutors fall apart.
I am not a big fan of AI-generated educational content, mostly because it's a great way to confidently learn falsehoods and misconceptions. I would prefer to learn from a reliable and reputable source.
I am also not a big fan of trying to beat doomscrolling. One of the defining properties of doomscrolling is that it is mindless and addicting. The moment you try to create a mindful, healthy alternative, you've already lost. No product will ever beat doomscrolling, only individuals dedicated to their own mental health are capable of clearing this hurdle.
This was a couple of years ago, but I remember using ChatGPT to try and study for a certification by generating quiz questions.
It would always start to make every correct answer option "C" over time, no matter what I tried. Eventually I was so focused on whether or not it was stuck in a "C" loop that I started overthinking all of the questions and wasting time.
Flash forward to testing Sonnet 4.6 recently to try and see if it could effectively teach me something new, I got about 5 prompts in before I had to point out an oversight, and it gave me the classic "you're absolutely right, ignore that suggestion".
This is anecdotal of course, but at least LLMs are helping to build my skills of fact verification and citation checking!
This is neat. I did see one bug early in the conversation: "The submission shows an escaped/unfinished string: find . -name \." which looks like an error converting what the user chose back to an escaped string for the AI.
For feedback:
* Please change the colours. I can't read white text on light purple. By all means keep purple, just increase the contrast.
* Maybe keep it in a narrower column, again for readability especially since we're constantly switching sides left to right as we read the AI text and the 'You submitted' text. I also feel it's quite duplicated, I can visually see which option was chosen yet it's restated in 'You submitted'. Only show what's necessary.
* In the replay you shared I can still change active radio buttons. Make it static.
To more focused feedback:
* I really like the idea. Instead of an AI telling you the answer, you learn -- this is fantastic.
* Multiple choice questions are great, but eventually please move to freeform. An AI can evaluate that. (For example, after learning a bit via examples, it might prompt, 'How do you find all PDFs in your home folder?' and give a text entry field. In the past, you'd have to type exactly the string the app expected; with an AI, it can evaluate flexibly. So take advantage of this. It's an immense opportunity vs earlier learning platforms.) Also, people learn better at some point when you're not given answers (aka multiple choice) but need to write.
I think you're on to something. AI as thinking replacement is a worry. AI as a guided teacher is great. I look forward to what you do with it.
Hey, thank you so much for the detailed feedback.
Re the more focused feedback, I totally agree re the questioning styles. In the prompt I ask for it to not do so many multiple choice questions, but I think it is addicted/the conversation history skews the context.
I'm going to introduce a settings panel (easily accessible during the conversation), which will let you move to "chat mode" (to discuss instead of be asked questions), and also to configure the types of questions you're asked and the ratio (if I can get the llm to oblige). I'm also going to see if I can come up with some different question formats beyond multiple choice, free-form and multi select (which the llm doesn't use too much).
My feedback reiterates what you're hearing about the cogintive load being unruly. If you want to "beat" doom scrolling, you need your mechanics to be as simple -- quick swipes and taps.
One potential direction, simple cards that are True/False.
If you want to beat doom scrolling, put your phone down?
Directed learning towards some goal is always going to be useful. These replacements for doom scrolling are addressing a surface level issue. Just change your habit to be spaced repetition on flash cards.
An interesting approach, Good luck with it! A nit to pick: find is not a bash command. You can run it for example from a Windows DOS command line as:
wsl find ...
You can run all Linux commands this way. Also, pretty sure that find's "-o" is the Boolean "or", not "otherwise". (Yet another example of why learning from LLMs is dangerous, I suppose).
otherwise is actually an accurate description of what -o actually means. It means do the thing on the left if it works, otherwise do the thing on the right. That is, if the clause on the left succeeds then the clause on the right will be ignored.
A naive interpretation of or in the light of Boolean algebra would be: do both and return true if either succeeds.
Haha, thanks for the further education!
The example you gave is just so buggy; why do you think that this is worth sharing with the world yet?
The very first question is full of obvious bugs.
You have 'find . -name "notes.txt"' selected, it then says 'You submitted: ls -R | grep notes.txt: find . -name "notes.txt"', then it responds:
'Thanks — your answer looks like it was partially entered.
'You picked find . -name "notes.txt" (good choice). The submission shows an escaped/unfinished string: find . -name \. The correct full command is find . -name "notes.txt", which searches recursively from the current directory for files or directories named exactly notes.txt.'
There seems to be some weird kind of quoting issue going on there. I would fix obvious issues like that before sharing this with the world.
You'll never make every single person on HN happy. But if you share your stuff early and make one person happy at the very least that means you should keep working on it!
Don't let perfectionism get in the way of good enough :]
I think there has just been a bit of whiplash with "Show HN" here lately, it was recently discussed here https://www.arthurcnops.blog/death-of-show-hn/ with much good discussion here https://news.ycombinator.com/item?id=47045804. That's no excuse for being overly harsh, but it's not really quite "anything is great to share!" either.
Show HNs never pleased everyone, and it'd be silly to try, but until recently there was a bit of a "it's not perfect... but the person has spent more time working on this than I have even spent thinking on the problem" kind of expectation whereas now many of them feel like the comments section ends up doing more thought about the submission than was put into refining it.
It'll be really interesting to see where this settles. In the meantime, erring on the side of kindness tends to work best!
Personally, I found this idea cool, and a fresh change from the onslaught of clawdbot posts on this platform. It was enough to make my time on this site a little bit more enjoyable. There's no need to nitpick to the point of bringing down another human for their work.
Sorry, I shouldn't have been that harsh about it; I apologize. Doesn't look like I can delet the comment, though.
Anyhow, I'm just a bit bitter about the onslaught of obviously vibe coded projects that very little effort has been put into. And OP did ask for feedback. But I shouldn't have been so mean about it. I know it can be fun to be able to whip up something quicker than you could before, but I really wish people would spend a tiny bit more effort on it before asking for feedback.
Hey, I appreciate your views (both positive and negative). There actually is a bit of a back story behind it all. I spent a month working on a different version of rebrain, (still available here: https://dev.rebrain.gg). After putting that work in I started to show it to some friends. It was clear from how they responded that it was not as promising as I had hoped. I listened to their feedback and came up with a different idea for LLM-based education (which is the version that's live now). I did vibe code it in about 3/4 days. But I vowed to try to get feedback sooner rather than later, which is why I posted it on ShowHN pretty early in its development. I do want to improve it... so please let me know if there are things which really frustrate you.
I checked the website and you weren't harsh enough. It reads more like a post modern art about vibecoding, some kind of performative art than anything.
I'm going to agree with you. Neat idea but the site is broken and clearly vibe coded, no Fs given to it actually working or making any sense.
This is far below most other "Show HN" posts and your first message was spot on.
Do you have any info on areas it was broken?
People are having fun with this. https://rebrain.gg/conversations/369
Does this form based interaction offer any specific advantages over chatting directly with ChatGPT or Claude? Essentially the site is still powered by their APIs and it is just the interface that is different.
I mean that's the question the author asks, the answer is to test it out yourself.
I like something here. The interactive - incremental steps of learning or teaching. But having to go through sort of quizes to just get info I need - may not be best for all use cases.
There is a reason chat bots work - it mimics natural human interaction.
But this can be interest pattern - for say AI driven personal tutor for math topic
Lastly - for love of god - pls do something with the UI - all colors and bubbles and I am totally lost just trying to make sense of what is going on. Look at reditt if you just want back and forth thread convo style. Life is simpler that way.
It would be great to be able to kick of a chatbot convo to go deeper on something each time you have answered. Chat for a bit to deepen understanding and read around, then pop back out and continue to the next question when your satisfied you've understood the concept.
Agree on colours. The Electronics topic was white on yellow, completely unreadable.
Yeah, I like your idea. Will implement
Improving the UI would only take you a few prompts
[flagged]
Doom scrolling is when you just have to flip up. Maybe swipe if it's an advanced app.
This is something else entirely and requires far too much thinking for the label.
I thought this would teach you random facts in one-screen-sized tiles. This doesn't look like doom scrolling. But I like the idea of the user picking a subject, maybe combine both: pick a subject, it gives you facts about it, maybe a tile to ask for which path to take once in a while (swipe left to keep learning about find, swipe right to learn about xargs).
Maybe if you can deliver information in small chunks in an interesting way that doesn't require a large attention span, it will work.
But if I have to read a wall of text, it's not an alternative to doomscrolling, it's an alternative to reading a book or documentation. And in that case I'd rather read a book or documentation.
The color scheme is painfully low-contrast. I can barely read it. Dear god increase the contrast.
You are right, I'll take out some of the worst colours
They seem all over the place. Instead of specific colours, you could go around the hue circle without touching saturation: https://developer.mozilla.org/en-US/docs/Web/CSS/Reference/V...
Great tool How is this internally implemented? Would you be writing an engineering blog post?
Great idea. It sadly doesn’t work for me, I went to pick a topic and some of the answers were pre-submitted, some of the questions did nothing after clicking “check”, then it just ended. Maybe it’s the HN hug of death?
You have to create a conversation for yourself. Currently you can't interact with other people's conversations. Sorry this wasn't clear enough!
There is also xikipedia.org if you want to doomscroll while learning
I clicked on that link, saw clashing colours and a general lack of UX design principles and immediately closed it.
Upon clicking ‘continue’ or ‘submit’, make it transition to a different color which will give a good user experience
Thanks for the interesting idea
wow. i like.
gripe: the "+New conversation" button is distracting, floating on the botton. what about floating a "+" in the upper right corner?
That's only visible when it's not your personal conversation (you can't interact with someone else's). In a way it's designed to be distracting so you know how to start your own conversation.
I think it's a cool idea and I personally find using LLMs as a teaching tool to still be the most rewarding way of interacting with them, if done right.
Your obvious first port of call IMO is correctness of material, where there's room for improvement [1]. I deliberately picked Gleam because it's still a less known language.
For what it's worth, prompting Opus 4.6 in chat got me this result [2]. Sonnet 4.6 via the Workbench also got it right.
Agree with other comments about UX and design, and maybe also some of those around improving teaching style or gamification aspects, but the above is more important.
Good luck with this, hope you crack it! :)
Yes, at the moment it's an issue of cost. I can't use the best models because it is not affordable. Hopefully as performance improves over the years this will become less of an issue. Maybe I can build in a websearch to verify info though...
I hear you. Yes, I think "seeding" an LLM with docs or other learning material is one of the fundamentals of effectively and efficiently using it for learning, maybe you can build more in that direction?
Yeah I actually started there. Https://dev.rebrain.gg has the old version up.
You upload a source and it generates questions from it. However when showing it to friends I found that the barrier to usage was too high as most people don’t have a source ready. But I think adding it as an option would be pretty cool and doable
I think you should use nano-banana to auto matically create a "youtube short" that might make this more addictive like doomscrolling
a lot of chat websites have a no registration mode so people can try it out
Great idea! I too forget the many switches that the 'find' command can use. Anything that teaches and helps one understand the core concepts of any subject is always a good idea. I'm just not a fan of creating an account and password. I have way, too, many of those. Good luck :)
No passwords FYI - I hate them too. Google sign in or "magic link". I know not everyone likes them, but trying to keep friction as low as possible
:)
Too many forms
[dead]
Thanks very much for the feedback. If you have any ideas of how to communicate the retention mechanisms please let me know. If you want to discuss further, please leave feedback on the site (you can do this if you're signed in) and mention you are nivcmo from hn and I'll be in touch via email.
the gap we kept hitting when building 45d was exactly this - there's a real difference between "AI that retrieves content" and "AI that understands how this specific person learns." the former is a better search engine. the latter is something new.
our learner model tracks 500+ signals per person - not just difficulty preferences but how they engage with confusion, where they disengage, what formats actually create retention for them specifically. the course structure changes as a result, not just the difficulty level.
curious what you found on the memory/context side. longitudinal context across sessions is where most AI tutors fall apart.