- Updated: March 30, 2026
- 55 min read
Coding Model Limits: Key Takeaways from Hacker News Discussion
Hacker Newsnew | past | comments | ask | show | jobs | submitloginAsk HN: Where have you found the coding limits of current models?15 points by logicallee 51 minutes ago | hide | past | favorite | 19 commentsI’d love to hear where you’ve run into the limits of current coding models, whether it’s based on lines of code, systems integration, databases, etc. help WhatsName 0 minutes ago | next [–] I hate the tendency to make things up and I dont mean hallucinations.I once had claude code write a python script that emulated the output of my training script, including prentending that loss is decreasing. Why? Because it was unable to install a python dependency.Everytime I use a coding agent, I need to double check that it’s not cutting corners, hard coding things that shouldn’t be or straight up rewriting failing test cases.What I need more is honesty.replyspicyusername 31 minutes ago | prev | next [–] Writing too much at once with under specified prompts.If you stick to targeted problems with well-described prompts, acceptance criteria, and lots of linting, unit testing, and integration testing, you’ll typically get what you want with code that looks okay. And when things start to stray, it’s easy to get things back on track.It’s when you start trying to have LLMs write too much without a human review that you start getting unnecessary function chains, abstractions that aren’t needed, code that doesn’t really match the existing style, duplicate code, missing functionality, hallucinated functionality, tautological tests, etc.It works best when there’s regular feedback in the loop about what’s good and what’s not good.Testing and linting can fill in some of that, but we still need a human in the loop with “taste”, so to speak.replysaagarjha 5 minutes ago | prev | next [–] I find that AI models are very bad at doing performance work because they keep guessing how their changes affect things or not really understanding how the profiler results work, leading to them going in circles and taking forever. I have noticed this effect in surprisingly few lines of code (hundreds).replymichaelmior 1 minute ago | parent | next [–] One thing I’ve found that I’ve found super helpful for this is converting profiling results to Markdown and feeding it back into the agent in a loop. I’ve done it with a bit of manual orchestration, but it could probably be automated pretty well. Specifically, pprof-rs[0] and pprof-to-md[1] have worked pretty well for me, YMMV.[0] https://github.com/tikv/pprof-rs[1] https://github.com/platformatic/pprof-to-mdreplybluGill 9 minutes ago | prev | next [–] It doesn’t push back enough when I ask it to implement something that is a bad idea. I’ve caught it breaking threading code, and then fixing it by putting atomic in the wrong place: which might have fixed in that a few racey tests can now run a thousands times, but it wasn’t the right place and so the race still exist.It has changed constants when moving the constant to a different file (the move was required, but the constant were not expected to change).I have found it very helpful to ask AI to review the latest changes – it often finds serious problems in review of code it just wrote.replyNormal_gaussian 22 minutes ago | prev | next [–] Respecting instructions around tool use, even on small codebases, where the tool isn’t its favourite way of doing things.For example, models repeatedly try and not use my Makefile’s (build / run / test / test-file / check) and instead spend several cycles attempting to invoke using system tools and getting it wrong.I’ve got to the point where I run most models with an extra path folder which overrides typically bad commands tells them off and redirects them back (more effective and portable than hooks). But then sometimes the model reads the makefile, discovers the workaround, and does the stupid anyway.replydatsci_est_2015 11 minutes ago | prev | next [–] Hard limit: – Plenty of API hallucination happening on cutting edge Spark (4.0.0+) functionality, especially PySpark. Spark bares some blame here for broken and incomplete documentation. Takes a human in the loop to realize that the documentation is misleading or wrong or missing. Soft limit: – API design.I’ve found that, unless specifically steered towards “good” API design (highly subjective), agents tend to just add another endpoint / function to satisfy the exact task at hand, with total disregard to how the rest of the API looks. (Pretty much exactly what a junior engineer would do…)replymichaelmior 0 minutes ago | parent | next [–] On this note, one thing I’ve found Codex to do is worry more than necessary about breaking changes for internal APIs.Maybe a bit more prompting would fix this, but I found even when iteratively implementing larger new features, it worries about breaking APIs that aren’t used by anything but the new code yet.replyreedf1 32 minutes ago | prev | next [–] Layers of abstraction. Most noticeably with inheritance and general OOP concepts.I’ve tried to force it, assuming it prefers a more functional or simple class style; but it genuinely struggles to generate (but not understand) what I might call a typical system in an OOP paradigm with well defined abstractions.replyBalinares 16 minutes ago | prev | next [–] Most recently: Opus 4.6 screwed up using a common and well documented API (Qt), then when asked to debug the observed issues, blamed “a Qt bug” and wrote a whole layer on top of the API to work around the issue caused by its incorrect use of the API.It did the above twice in a row around different parts of the API.The thought that there are almost certainly devs out there merging Claude PRs without the skills or volition to push back on its screwups is not comfortable.replyhypendev 27 minutes ago | prev | next [–] Depends – in the pure technical “implementation” level, only limits I’ve found are the ones with outdated knowledge – libraries, platforms, availability of some things.But one big limit is the DX. Their choice of DX is usually abysmal – ironically, just like an average devs.They seem to lack the aesthetic instinct for code, so you have to really point them hard into the direction or provide a sample of the expected DX, for them to still fight against it at every turn.While understandable in a way, as they are trained on average code and most code will now be written by the machines anyways making the DX “less relevant”, it’s also a giant code smell, as bad DX tends to point towards bad internals and wrong decisions along the way.So not really a technical limit – they swallow anything you throw at them, even the most complex cases – but more of an aesthetic limit in terms of taste.replynetdevphoenix 33 minutes ago | prev | next [–] I think consistency levels when operating autonomously is still a challenge which you could refer to as their limit. You need to do so much around them to keep the consistency AND a decent level of performance. It’s like a savant with a short-attention span.replywood_spirit 33 minutes ago | prev | next [–] I found sonnet 4.5 struggled with a two pointer interval merge (two sorted lists of things with start stop timestamps), but opus 4.5 managed. Then it took opus 4.6 to make it a three way or k way merge. So it reminded me of the classic Simple made Easy talk by Rich Hickey where he talks about braids. Sonnet couldn’t track two twisted braids and opus 4.6 managed a weave.But the last few weeks Opus 4.6 seems to have got dumb again.Now it is making way more mistakes and forgetting useful things and recent context it used to manage.I am guessing this is just Anthropic quietly dialling down the real effort as they either downsize or free up compute for someone or something else.replyrockbruno 20 minutes ago | prev | next [–] Problems that require deep knowledge of multiple repositories, e.g. when trying to debug issues involving dependencies.The models get confused very fast even with all code available locally, due to the size of the problem. But in my experience any kind of deep integration already messes up the models, even within a single repo.replysshine 13 minutes ago | prev | next [–] Dendritic Nix (either too new or too underrepresented in training data)Proper escaping of layered syntaxes in Ansible on the first attemptWriting bare-metal embedded Rust, although this was long ago, so not current modelsreplydgb23 11 minutes ago | prev | next [–] I would push back on that question a little, because it has a baked in assumption about how these things work that conflict with my mental model and experience with them.The reason is that sometimes it spits out something or does a workflow that’s pretty sophisticated, and sometimes it fails spectacularly in the most basic ways.I don’t think there is a complexity or domain knowledge limit as there would be with a human. Or at least not in the same sense. As long as it can repeat and remix patterns that it is trained on, then it will do its thing well. The same seems to be true for “reasoning” loops and workflows.It can spit out code that has been done N times before in a similar manner for a large N.They can still break down because of very trivial issues and assumptions that happen to be baked in, go off the rails and get stuck long loops that are completely insane if you think of them as imitating human programmers.When I use an agent, I always interview it first about the task. Ask how they would go about it, probe them, give them info that they lack.Never go from prompt to action.Have them define their approach first, then split the approach into pieces, from gathering data to cleaning it up and so on. If applicable, front-load work that can be achieved with scripts, so you have testable and repeatable steps rather than let it go wild.So the TLDR is: I think the limitation is simply that it’s a non-deterministic token machine that produces useful results enough of the time so it appears to be reasonable.replykaelandt 32 minutes ago | prev | next [–] It’s just really hard for them to write non-verbose code. I don’t know if this is incentives from the providers to generate more tokens, but even with guidance on compact code, simple, etc, they just can’t really do it right now.replykeyle 35 minutes ago | prev | next [–] I haven’t found their limit, but I have found my limits, of waiting for them to do their thing.I tend to re-enjoy handrolling code more. I delegate the stuff that annoys me.We have a large multiplatform codebase, the issue seems to be more the time it takes to navigate the code and reason about it, rather than the size. Arguably the size is causing ‘them’ to be slower in that regards, but I haven’t found the limit yet. And with compaction, it’s even less of a problem.my 5c.replyadventured 27 minutes ago | prev [–] They don’t understand esoteric areas of computer science very well at all.I had a mistake in which a large back-up file deletion event happened during a robocopy. 600gb of files got ‘deleted’ (file headers toast etc). Trying to get the LLMs to understand the hunt parameters, what to focus on, what not to focus on – none of them could reasonably come close to doing file content recovery properly.I needed to build a custom solution because the available industry options couldn’t do what was required and the LLMs were useless for that (including the latest versions of Claude, Gemini and GPT). They just went around in circles, capped by their apparently weak knowledge of file recovery as a field. That is, creativity was their limitation.reply Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact Search: [{},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”Ask HN: Where have you found the coding limits of current models?15 points by logicallee 51 minutes ago | hide | past | favorite | 19 commentsI’d love to hear where you’ve run into the limits of current coding models, whether it’s based on lines of code, systems integration, databases, etc.help\n\nWhatsName 0 minutes ago | next [–]\nI hate the tendency to make things up and I dont mean hallucinations.I once had claude code write a python script that emulated the output of my training script, including prentending that loss is decreasing. Why? Because it was unable to install a python dependency.Everytime I use a coding agent, I need to double check that it’s not cutting corners, hard coding things that shouldn’t be or straight up rewriting failing test cases.What I need more is honesty.replyspicyusername 31 minutes ago | prev | next [–]\nWriting too much at once with under specified prompts.If you stick to targeted problems with well-described prompts, acceptance criteria, and lots of linting, unit testing, and integration testing, you’ll typically get what you want with code that looks okay. And when things start to stray, it’s easy to get things back on track.It’s when you start trying to have LLMs write too much without a human review that you start getting unnecessary function chains, abstractions that aren’t needed, code that doesn’t really match the existing style, duplicate code, missing functionality, hallucinated functionality, tautological tests, etc.It works best when there’s regular feedback in the loop about what’s good and what’s not good.Testing and linting can fill in some of that, but we still need a human in the loop with \”taste\”, so to speak.replysaagarjha 5 minutes ago | prev | next [–]\nI find that AI models are very bad at doing performance work because they keep guessing how their changes affect things or not really understanding how the profiler results work, leading to them going in circles and taking forever. I have noticed this effect in surprisingly few lines of code (hundreds).replymichaelmior 1 minute ago | parent | next [–]\nOne thing I’ve found that I’ve found super helpful for this is converting profiling results to Markdown and feeding it back into the agent in a loop. I’ve done it with a bit of manual orchestration, but it could probably be automated pretty well. Specifically, pprof-rs[0] and pprof-to-md[1] have worked pretty well for me, YMMV.[0] https://github.com/tikv/pprof-rs[1] https://github.com/platformatic/pprof-to-mdreplybluGill 9 minutes ago | prev | next [–]\nIt doesn’t push back enough when I ask it to implement something that is a bad idea. I’ve caught it breaking threading code, and then fixing it by putting atomic in the wrong place: which might have fixed in that a few racey tests can now run a thousands times, but it wasn’t the right place and so the race still exist.It has changed constants when moving the constant to a different file (the move was required, but the constant were not expected to change).I have found it very helpful to ask AI to review the latest changes – it often finds serious problems in review of code it just wrote.replyNormal_gaussian 22 minutes ago | prev | next [–]\nRespecting instructions around tool use, even on small codebases, where the tool isn’t its favourite way of doing things.For example, models repeatedly try and not use my Makefile’s (build / run / test / test-file / check) and instead spend several cycles attempting to invoke using system tools and getting it wrong.I’ve got to the point where I run most models with an extra path folder which overrides typically bad commands tells them off and redirects them back (more effective and portable than hooks). But then sometimes the model reads the makefile, discovers the workaround, and does the stupid anyway.replydatsci_est_2015 11 minutes ago | prev | next [–]\nHard limit: – Plenty of API hallucination happening on cutting edge Spark (4.0.0+) functionality, especially PySpark. Spark bares some blame here for broken and incomplete documentation. Takes a human in the loop to realize that the documentation is misleading or wrong or missing.\n\nSoft limit: – API design.I’ve found that, unless specifically steered towards “good” API design (highly subjective), agents tend to just add another endpoint / function to satisfy the exact task at hand, with total disregard to how the rest of the API looks. (Pretty much exactly what a junior engineer would do…)replymichaelmior 0 minutes ago | parent | next [–]\nOn this note, one thing I’ve found Codex to do is worry more than necessary about breaking changes for internal APIs.Maybe a bit more prompting would fix this, but I found even when iteratively implementing larger new features, it worries about breaking APIs that aren’t used by anything but the new code yet.replyreedf1 32 minutes ago | prev | next [–]\nLayers of abstraction. Most noticeably with inheritance and general OOP concepts.I’ve tried to force it, assuming it prefers a more functional or simple class style; but it genuinely struggles to generate (but not understand) what I might call a typical system in an OOP paradigm with well defined abstractions.replyBalinares 16 minutes ago | prev | next [–]\nMost recently: Opus 4.6 screwed up using a common and well documented API (Qt), then when asked to debug the observed issues, blamed \”a Qt bug\” and wrote a whole layer on top of the API to work around the issue caused by its incorrect use of the API.It did the above twice in a row around different parts of the API.The thought that there are almost certainly devs out there merging Claude PRs without the skills or volition to push back on its screwups is not comfortable.replyhypendev 27 minutes ago | prev | next [–]\nDepends – in the pure technical \”implementation\” level, only limits I’ve found are the ones with outdated knowledge – libraries, platforms, availability of some things.But one big limit is the DX. Their choice of DX is usually abysmal – ironically, just like an average devs.They seem to lack the aesthetic instinct for code, so you have to really point them hard into the direction or provide a sample of the expected DX, for them to still fight against it at every turn.While understandable in a way, as they are trained on average code and most code will now be written by the machines anyways making the DX \”less relevant\”, it’s also a giant code smell, as bad DX tends to point towards bad internals and wrong decisions along the way.So not really a technical limit – they swallow anything you throw at them, even the most complex cases – but more of an aesthetic limit in terms of taste.replynetdevphoenix 33 minutes ago | prev | next [–]\nI think consistency levels when operating autonomously is still a challenge which you could refer to as their limit. You need to do so much around them to keep the consistency AND a decent level of performance. It’s like a savant with a short-attention span.replywood_spirit 33 minutes ago | prev | next [–]\nI found sonnet 4.5 struggled with a two pointer interval merge (two sorted lists of things with start stop timestamps), but opus 4.5 managed. Then it took opus 4.6 to make it a three way or k way merge. So it reminded me of the classic Simple made Easy talk by Rich Hickey where he talks about braids. Sonnet couldn’t track two twisted braids and opus 4.6 managed a weave.But the last few weeks Opus 4.6 seems to have got dumb again.Now it is making way more mistakes and forgetting useful things and recent context it used to manage.I am guessing this is just Anthropic quietly dialling down the real effort as they either downsize or free up compute for someone or something else.replyrockbruno 20 minutes ago | prev | next [–]\nProblems that require deep knowledge of multiple repositories, e.g. when trying to debug issues involving dependencies.The models get confused very fast even with all code available locally, due to the size of the problem. But in my experience any kind of deep integration already messes up the models, even within a single repo.replysshine 13 minutes ago | prev | next [–]\nDendritic Nix (either too new or too underrepresented in training data)Proper escaping of layered syntaxes in Ansible on the first attemptWriting bare-metal embedded Rust, although this was long ago, so not current modelsreplydgb23 11 minutes ago | prev | next [–]\nI would push back on that question a little, because it has a baked in assumption about how these things work that conflict with my mental model and experience with them.The reason is that sometimes it spits out something or does a workflow that’s pretty sophisticated, and sometimes it fails spectacularly in the most basic ways.I don’t think there is a complexity or domain knowledge limit as there would be with a human. Or at least not in the same sense. As long as it can repeat and remix patterns that it is trained on, then it will do its thing well. The same seems to be true for \”reasoning\” loops and workflows.It can spit out code that has been done N times before in a similar manner for a large N.They can still break down because of very trivial issues and assumptions that happen to be baked in, go off the rails and get stuck long loops that are completely insane if you think of them as imitating human programmers.When I use an agent, I always interview it first about the task. Ask how they would go about it, probe them, give them info that they lack.Never go from prompt to action.Have them define their approach first, then split the approach into pieces, from gathering data to cleaning it up and so on. If applicable, front-load work that can be achieved with scripts, so you have testable and repeatable steps rather than let it go wild.So the TLDR is: I think the limitation is simply that it’s a non-deterministic token machine that produces useful results enough of the time so it appears to be reasonable.replykaelandt 32 minutes ago | prev | next [–]\nIt’s just really hard for them to write non-verbose code. I don’t know if this is incentives from the providers to generate more tokens, but even with guidance on compact code, simple, etc, they just can’t really do it right now.replykeyle 35 minutes ago | prev | next [–]\nI haven’t found their limit, but I have found my limits, of waiting for them to do their thing.I tend to re-enjoy handrolling code more. I delegate the stuff that annoys me.We have a large multiplatform codebase, the issue seems to be more the time it takes to navigate the code and reason about it, rather than the size. Arguably the size is causing ‘them’ to be slower in that regards, but I haven’t found the limit yet. And with compaction, it’s even less of a problem.my 5c.replyadventured 27 minutes ago | prev [–]\nThey don’t understand esoteric areas of computer science very well at all.I had a mistake in which a large back-up file deletion event happened during a robocopy. 600gb of files got ‘deleted’ (file headers toast etc). Trying to get the LLMs to understand the hunt parameters, what to focus on, what not to focus on – none of them could reasonably come close to doing file content recovery properly.I needed to build a custom solution because the available industry options couldn’t do what was required and the LLMs were useless for that (including the latest versions of Claude, Gemini and GPT). They just went around in circles, capped by their apparently weak knowledge of file recovery as a field. That is, creativity was their limitation.reply”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””,”login”:””,”Ask HN: Where have you found the coding limits of current models?”:”Ask HN: Where have you found the coding limits of current models?”,”WhatsName 0 minutes ago | next [–]\nI hate the tendency to make things up and I dont mean hallucinations.I once had claude code write a python script that emulated the output of my training script, including prentending that loss is decreasing. Why? Because it was unable to install a python dependency.Everytime I use a coding agent, I need to double check that it’s not cutting corners, hard coding things that shouldn’t be or straight up rewriting failing test cases.What I need more is honesty.reply”:”15 points by logicallee 51 minutes ago | hide | past | favorite | 19 comments”,”spicyusername 31 minutes ago | prev | next [–]\nWriting too much at once with under specified prompts.If you stick to targeted problems with well-described prompts, acceptance criteria, and lots of linting, unit testing, and integration testing, you’ll typically get what you want with code that looks okay. And when things start to stray, it’s easy to get things back on track.It’s when you start trying to have LLMs write too much without a human review that you start getting unnecessary function chains, abstractions that aren’t needed, code that doesn’t really match the existing style, duplicate code, missing functionality, hallucinated functionality, tautological tests, etc.It works best when there’s regular feedback in the loop about what’s good and what’s not good.Testing and linting can fill in some of that, but we still need a human in the loop with \”taste\”, so to speak.reply”:””,”saagarjha 5 minutes ago | prev | next [–]\nI find that AI models are very bad at doing performance work because they keep guessing how their changes affect things or not really understanding how the profiler results work, leading to them going in circles and taking forever. I have noticed this effect in surprisingly few lines of code (hundreds).reply”:”I’d love to hear where you’ve run into the limits of current coding models, whether it’s based on lines of code, systems integration, databases, etc.”,”michaelmior 1 minute ago | parent | next [–]\nOne thing I’ve found that I’ve found super helpful for this is converting profiling results to Markdown and feeding it back into the agent in a loop. I’ve done it with a bit of manual orchestration, but it could probably be automated pretty well.Specifically, pprof-rs[0] and pprof-to-md[1] have worked pretty well for me, YMMV.[0] https://github.com/tikv/pprof-rs[1] https://github.com/platformatic/pprof-to-mdreply”:””,”bluGill 9 minutes ago | prev | next [–]\nIt doesn’t push back enough when I ask it to implement something that is a bad idea.I’ve caught it breaking threading code, and then fixing it by putting atomic in the wrong place: which might have fixed in that a few racey tests can now run a thousands times, but it wasn’t the right place and so the race still exist. It has changed constants when moving the constant to a different file (the move was required, but the constant were not expected to change).I have found it very helpful to ask AI to review the latest changes – it often finds serious problems in review of code it just wrote.reply”:”help”,”Normal_gaussian 22 minutes ago | prev | next [–]\nRespecting instructions around tool use, even on small codebases, where the tool isn’t its favourite way of doing things.For example, models repeatedly try and not use my Makefile’s (build / run / test / test-file / check) and instead spend several cycles attempting to invoke using system tools and getting it wrong.I’ve got to the point where I run most models with an extra path folder which overrides typically bad commands tells them off and redirects them back (more effective and portable than hooks). But then sometimes the model reads the makefile, discovers the workaround, and does the stupid anyway.reply”:”WhatsName 0 minutes ago | next [–]\nI hate the tendency to make things up and I dont mean hallucinations.I once had claude code write a python script that emulated the output of my training script, including prentending that loss is decreasing. Why? Because it was unable to install a python dependency.Everytime I use a coding agent, I need to double check that it’s not cutting corners, hard coding things that shouldn’t be or straight up rewriting failing test cases.What I need more is honesty.reply”,”datsci_est_2015 11 minutes ago | prev | next [–]\nHard limit: – Plenty of API hallucination happening on cutting edge Spark (4.0.0+) functionality, especially PySpark. Spark bares some blame here for broken and incomplete documentation. Takes a human in the loop to realize that the documentation is misleading or wrong or missing.\n\nSoft limit: – API design.I’ve found that, unless specifically steered towards “good” API design (highly subjective), agents tend to just add another endpoint / function to satisfy the exact task at hand, with total disregard to how the rest of the API looks. (Pretty much exactly what a junior engineer would do…)reply”:””,”michaelmior 0 minutes ago | parent | next [–]\nOn this note, one thing I’ve found Codex to do is worry more than necessary about breaking changes for internal APIs.Maybe a bit more prompting would fix this, but I found even when iteratively implementing larger new features, it worries about breaking APIs that aren’t used by anything but the new code yet.reply”:””,”reedf1 32 minutes ago | prev | next [–]\nLayers of abstraction. Most noticeably with inheritance and general OOP concepts.I’ve tried to force it, assuming it prefers a more functional or simple class style; but it genuinely struggles to generate (but not understand) what I might call a typical system in an OOP paradigm with well defined abstractions.reply”:”WhatsName 0 minutes ago | next [–]\nI hate the tendency to make things up and I dont mean hallucinations.I once had claude code write a python script that emulated the output of my training script, including prentending that loss is decreasing. Why?Because it was unable to install a python dependency.Everytime I use a coding agent, I need to double check that it’s not cutting corners, hard coding things that shouldn’t be or straight up rewriting failing test cases.What I need more is honesty.reply”,”Balinares 16 minutes ago | prev | next [–]\nMost recently: Opus 4.6 screwed up using a common and well documented API (Qt), then when asked to debug the observed issues, blamed \”a Qt bug\” and wrote a whole layer on top of the API to work around the issue caused by its incorrect use of the API.It did the above twice in a row around different parts of the API.The thought that there are almost certainly devs out there merging Claude PRs without the skills or volition to push back on its screwups is not comfortable.reply”:”spicyusername 31 minutes ago | prev | next [–]\nWriting too much at once with under specified prompts.If you stick to targeted problems with well-described prompts, acceptance criteria, and lots of linting, unit testing, and integration testing, you’ll typically get what you want with code that looks okay. And when things start to stray, it’s easy to get things back on track.It’s when you start trying to have LLMs write too much without a human review that you start getting unnecessary function chains, abstractions that aren’t needed, code that doesn’t really match the existing style, duplicate code, missing functionality, hallucinated functionality, tautological tests, etc.It works best when there’s regular feedback in the loop about what’s good and what’s not good.Testing and linting can fill in some of that, but we still need a human in the loop with \”taste\”, so to speak.reply”,”hypendev 27 minutes ago | prev | next [–]\nDepends – in the pure technical \”implementation\” level, only limits I’ve found are the ones with outdated knowledge – libraries, platforms, availability of some things.But one big limit is the DX. Their choice of DX is usually abysmal – ironically, just like an average devs.They seem to lack the aesthetic instinct for code, so you have to really point them hard into the direction or provide a sample of the expected DX, for them to still fight against it at every turn.While understandable in a way, as they are trained on average code and most code will now be written by the machines anyways making the DX \”less relevant\”, it’s also a giant code smell, as bad DX tends to point towards bad internals and wrong decisions along the way.So not really a technical limit – they swallow anything you throw at them, even the most complex cases – but more of an aesthetic limit in terms of taste.reply”:””,”netdevphoenix 33 minutes ago | prev | next [–]\nI think consistency levels when operating autonomously is still a challenge which you could refer to as their limit. You need to do so much around them to keep the consistency AND a decent level of performance. It’s like a savant with a short-attention span.reply”:””,”wood_spirit 33 minutes ago | prev | next [–]\nI found sonnet 4.5 struggled with a two pointer interval merge (two sorted lists of things with start stop timestamps), but opus 4.5 managed. Then it took opus 4.6 to make it a three way or k way merge. So it reminded me of the classic Simple made Easy talk by Rich Hickey where he talks about braids. Sonnet couldn’t track two twisted braids and opus 4.6 managed a weave.But the last few weeks Opus 4.6 seems to have got dumb again.Now it is making way more mistakes and forgetting useful things and recent context it used to manage.I am guessing this is just Anthropic quietly dialling down the real effort as they either downsize or free up compute for someone or something else.reply”:”spicyusername 31 minutes ago | prev | next [–]\nWriting too much at once with under specified prompts.If you stick to targeted problems with well-described prompts, acceptance criteria, and lots of linting, unit testing, and integration testing, you’ll typically get what you want with code that looks okay. And when things start to stray, it’s easy to get things back on track.It’s when you start trying to have LLMs write too much without a human review that you start getting unnecessary function chains, abstractions that aren’t needed, code that doesn’t really match the existing style, duplicate code, missing functionality, hallucinated functionality, tautological tests, etc.It works best when there’s regular feedback in the loop about what’s good and what’s not good.Testing and linting can fill in some of that, but we still need a human in the loop with \”taste\”, so to speak.reply”,”rockbruno 20 minutes ago | prev | next [–]\nProblems that require deep knowledge of multiple repositories, e.g. when trying to debug issues involving dependencies. The models get confused very fast even with all code available locally, due to the size of the problem. But in my experience any kind of deep integration already messes up the models, even within a single repo.reply”:”saagarjha 5 minutes ago | prev | next [–]\nI find that AI models are very bad at doing performance work because they keep guessing how their changes affect things or not really understanding how the profiler results work, leading to them going in circles and taking forever. I have noticed this effect in surprisingly few lines of code (hundreds).reply”,”sshine 13 minutes ago | prev | next [–]\nDendritic Nix (either too new or too underrepresented in training data)Proper escaping of layered syntaxes in Ansible on the first attemptWriting bare-metal embedded Rust, although this was long ago, so not current modelsreply”:””,”dgb23 11 minutes ago | prev | next [–]\nI would push back on that question a little, because it has a baked in assumption about how these things work that conflict with my mental model and experience with them.The reason is that sometimes it spits out something or does a workflow that’s pretty sophisticated, and sometimes it fails spectacularly in the most basic ways.I don’t think there is a complexity or domain knowledge limit as there would be with a human. Or at least not in the same sense. As long as it can repeat and remix patterns that it is trained on, then it will do its thing well. The same seems to be true for \”reasoning\” loops and workflows.It can spit out code that has been done N times before in a similar manner for a large N.They can still break down because of very trivial issues and assumptions that happen to be baked in, go off the rails and get stuck long loops that are completely insane if you think of them as imitating human programmers.When I use an agent, I always interview it first about the task. Ask how they would go about it, probe them, give them info that they lack.Never go from prompt to action.Have them define their approach first, then split the approach into pieces, from gathering data to cleaning it up and so on. If applicable, front-load work that can be achieved with scripts, so you have testable and repeatable steps rather than let it go wild.So the TLDR is: I think the limitation is simply that it’s a non-deterministic token machine that produces useful results enough of the time so it appears to be reasonable.reply”:””,”kaelandt 32 minutes ago | prev | next [–]\nIt’s just really hard for them to write non-verbose code. I don’t know if this is incentives from the providers to generate more tokens, but even with guidance on compact code, simple, etc, they just can’t really do it right now.reply”:”saagarjha 5 minutes ago | prev | next [–]\nI find that AI models are very bad at doing performance work because they keep guessing how their changes affect things or not really understanding how the profiler results work, leading to them going in circles and taking forever. I have noticed this effect in surprisingly few lines of code (hundreds).reply”,”keyle 35 minutes ago | prev | next [–]\nI haven’t found their limit, but I have found my limits, of waiting for them to do their thing.I tend to re-enjoy handrolling code more. I delegate the stuff that annoys me.We have a large multiplatform codebase, the issue seems to be more the time it takes to navigate the code and reason about it, rather than the size. Arguably the size is causing ‘them’ to be slower in that regards, but I haven’t found the limit yet. And with compaction, it’s even less of a problem.my 5c.reply”:”michaelmior 1 minute ago | parent | next [–]\nOne thing I’ve found that I’ve found super helpful for this is converting profiling results to Markdown and feeding it back into the agent in a loop. I’ve done it with a bit of manual orchestration, but it could probably be automated pretty well. Specifically, pprof-rs[0] and pprof-to-md[1] have worked pretty well for me, YMMV.[0] https://github.com/tikv/pprof-rs[1] https://github.com/platformatic/pprof-to-mdreply”,”adventured 27 minutes ago | prev [–]\nThey don’t understand esoteric areas of computer science very well at all.I had a mistake in which a large back-up file deletion event happened during a robocopy. 600gb of files got ‘deleted’ (file headers toast etc). Trying to get the LLMs to understand the hunt parameters, what to focus on, what not to focus on – none of them could reasonably come close to doing file content recovery properly.I needed to build a custom solution because the available industry options couldn’t do what was required and the LLMs were useless for that (including the latest versions of Claude, Gemini and GPT). They just went around in circles, capped by their apparently weak knowledge of file recovery as a field. That is, creativity was their limitation.reply”:””},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:””,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:”15 points by logicallee 51 minutes ago | hide | past | favorite | 19 comments”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:””,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:”I’d love to hear where you’ve run into the limits of current coding models, whether it’s based on lines of code, systems integration, databases, etc.”},{},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:””,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:”help”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”spicyusername 31 minutes ago | prev | next [–]\nWriting too much at once with under specified prompts.If you stick to targeted problems with well-described prompts, acceptance criteria, and lots of linting, unit testing, and integration testing, you’ll typically get what you want with code that looks okay. And when things start to stray, it’s easy to get things back on track.It’s when you start trying to have LLMs write too much without a human review that you start getting unnecessary function chains, abstractions that aren’t needed, code that doesn’t really match the existing style, duplicate code, missing functionality, hallucinated functionality, tautological tests, etc.It works best when there’s regular feedback in the loop about what’s good and what’s not good.Testing and linting can fill in some of that, but we still need a human in the loop with \”taste\”, so to speak.reply”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””,”login”:””,”Ask HN: Where have you found the coding limits of current models?”:”spicyusername 31 minutes ago | prev | next [–]\nWriting too much at once with under specified prompts.If you stick to targeted problems with well-described prompts, acceptance criteria, and lots of linting, unit testing, and integration testing, you’ll typically get what you want with code that looks okay. And when things start to stray, it’s easy to get things back on track.It’s when you start trying to have LLMs write too much without a human review that you start getting unnecessary function chains, abstractions that aren’t needed, code that doesn’t really match the existing style, duplicate code, missing functionality, hallucinated functionality, tautological tests, etc.It works best when there’s regular feedback in the loop about what’s good and what’s not good.Testing and linting can fill in some of that, but we still need a human in the loop with \”taste\”, so to speak.reply”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”saagarjha 5 minutes ago | prev | next [–]\nI find that AI models are very bad at doing performance work because they keep guessing how their changes affect things or not really understanding how the profiler results work, leading to them going in circles and taking forever.I have noticed this effect in surprisingly few lines of code (hundreds).reply”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””,”login”:””,”Ask HN: Where have you found the coding limits of current models?”:”saagarjha 5 minutes ago | prev | next [–]\nI find that AI models are very bad at doing performance work because they keep guessing how their changes affect things or not really understanding how the profiler results work, leading to them going in circles and taking forever. I have noticed this effect in surprisingly few lines of code (hundreds).reply”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”michaelmior 1 minute ago | parent | next [–]\nOne thing I’ve found that I’ve found super helpful for this is converting profiling results to Markdown and feeding it back into the agent in a loop. I’ve done it with a bit of manual orchestration, but it could probably be automated pretty well.Specifically, pprof-rs[0] and pprof-to-md[1] have worked pretty well for me, YMMV.[0] https://github.com/tikv/pprof-rs[1] https://github.com/platformatic/pprof-to-mdreply”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””,”login”:””,”Ask HN: Where have you found the coding limits of current models?”:”michaelmior 1 minute ago | parent | next [–]\nOne thing I’ve found that I’ve found super helpful for this is converting profiling results to Markdown and feeding it back into the agent in a loop. I’ve done it with a bit of manual orchestration, but it could probably be automated pretty well. Specifically, pprof-rs[0] and pprof-to-md[1] have worked pretty well for me, YMMV.[0] https://github.com/tikv/pprof-rs[1] https://github.com/platformatic/pprof-to-mdreply”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”bluGill 9 minutes ago | prev | next [–]\nIt doesn’t push back enough when I ask it to implement something that is a bad idea. I’ve caught it breaking threading code, and then fixing it by putting atomic in the wrong place: which might have fixed in that a few racey tests can now run a thousands times, but it wasn’t the right place and so the race still exist.It has changed constants when moving the constant to a different file (the move was required, but the constant were not expected to change).I have found it very helpful to ask AI to review the latest changes – it often finds serious problems in review of code it just wrote.reply”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””,”login”:””,”Ask HN: Where have you found the coding limits of current models?”:”bluGill 9 minutes ago | prev | next [–]\nIt doesn’t push back enough when I ask it to implement something that is a bad idea. I’ve caught it breaking threading code, and then fixing it by putting atomic in the wrong place: which might have fixed in that a few racey tests can now run a thousands times, but it wasn’t the right place and so the race still exist.It has changed constants when moving the constant to a different file (the move was required, but the constant were not expected to change).I have found it very helpful to ask AI to review the latest changes – it often finds serious problems in review of code it just wrote.reply”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”Normal_gaussian 22 minutes ago | prev | next [–]\nRespecting instructions around tool use, even on small codebases, where the tool isn’t its favourite way of doing things.For example, models repeatedly try and not use my Makefile’s (build / run / test / test-file / check) and instead spend several cycles attempting to invoke using system tools and getting it wrong.I’ve got to the point where I run most models with an extra path folder which overrides typically bad commands tells them off and redirects them back (more effective and portable than hooks). But then sometimes the model reads the makefile, discovers the workaround, and does the stupid anyway.reply”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””,”login”:””,”Ask HN: Where have you found the coding limits of current models?”:”Normal_gaussian 22 minutes ago | prev | next [–]\nRespecting instructions around tool use, even on small codebases, where the tool isn’t its favourite way of doing things.For example, models repeatedly try and not use my Makefile’s (build / run / test / test-file / check) and instead spend several cycles attempting to invoke using system tools and getting it wrong.I’ve got to the point where I run most models with an extra path folder which overrides typically bad commands tells them off and redirects them back (more effective and portable than hooks). But then sometimes the model reads the makefile, discovers the workaround, and does the stupid anyway.reply”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”datsci_est_2015 11 minutes ago | prev | next [–]\nHard limit: – Plenty of API hallucination happening on cutting edge Spark (4.0.0+) functionality, especially PySpark. Spark bares some blame here for broken and incomplete documentation. Takes a human in the loop to realize that the documentation is misleading or wrong or missing.\n\nSoft limit: – API design. I’ve found that, unless specifically steered towards “good” API design (highly subjective), agents tend to just add another endpoint / function to satisfy the exact task at hand, with total disregard to how the rest of the API looks.(Pretty much exactly what a junior engineer would do…)reply”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””,”login”:””,”Ask HN: Where have you found the coding limits of current models?”:”datsci_est_2015 11 minutes ago | prev | next [–]\nHard limit: – Plenty of API hallucination happening on cutting edge Spark (4.0.0+) functionality, especially PySpark. Spark bares some blame here for broken and incomplete documentation.Takes a human in the loop to realize that the documentation is misleading or wrong or missing.\n\nSoft limit: – API design. I’ve found that, unless specifically steered towards “good” API design (highly subjective), agents tend to just add another endpoint / function to satisfy the exact task at hand, with total disregard to how the rest of the API looks.(Pretty much exactly what a junior engineer would do…)reply”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”michaelmior 0 minutes ago | parent | next [–]\nOn this note, one thing I’ve found Codex to do is worry more than necessary about breaking changes for internal APIs. Maybe a bit more prompting would fix this, but I found even when iteratively implementing larger new features, it worries about breaking APIs that aren’t used by anything but the new code yet.reply”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””,”login”:””,”Ask HN: Where have you found the coding limits of current models?”:”michaelmior 0 minutes ago | parent | next [–]\nOn this note, one thing I’ve found Codex to do is worry more than necessary about breaking changes for internal APIs.Maybe a bit more prompting would fix this, but I found even when iteratively implementing larger new features, it worries about breaking APIs that aren’t used by anything but the new code yet.reply”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”reedf1 32 minutes ago | prev | next [–]\nLayers of abstraction. Most noticeably with inheritance and general OOP concepts.I’ve tried to force it, assuming it prefers a more functional or simple class style; but it genuinely struggles to generate (but not understand) what I might call a typical system in an OOP paradigm with well defined abstractions.reply”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””,”login”:””,”Ask HN: Where have you found the coding limits of current models?”:”reedf1 32 minutes ago | prev | next [–]\nLayers of abstraction.Most noticeably with inheritance and general OOP concepts. I’ve tried to force it, assuming it prefers a more functional or simple class style; but it genuinely struggles to generate (but not understand) what I might call a typical system in an OOP paradigm with well defined abstractions.reply”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”Balinares 16 minutes ago | prev | next [–]\nMost recently: Opus 4.6 screwed up using a common and well documented API (Qt), then when asked to debug the observed issues, blamed \”a Qt bug\” and wrote a whole layer on top of the API to work around the issue caused by its incorrect use of the API.It did the above twice in a row around different parts of the API.The thought that there are almost certainly devs out there merging Claude PRs without the skills or volition to push back on its screwups is not comfortable.reply”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””,”login”:””,”Ask HN: Where have you found the coding limits of current models?”:”Balinares 16 minutes ago | prev | next [–]\nMost recently: Opus 4.6 screwed up using a common and well documented API (Qt), then when asked to debug the observed issues, blamed \”a Qt bug\” and wrote a whole layer on top of the API to work around the issue caused by its incorrect use of the API.It did the above twice in a row around different parts of the API.The thought that there are almost certainly devs out there merging Claude PRs without the skills or volition to push back on its screwups is not comfortable.reply”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”hypendev 27 minutes ago | prev | next [–]\nDepends – in the pure technical \”implementation\” level, only limits I’ve found are the ones with outdated knowledge – libraries, platforms, availability of some things.But one big limit is the DX. Their choice of DX is usually abysmal – ironically, just like an average devs.They seem to lack the aesthetic instinct for code, so you have to really point them hard into the direction or provide a sample of the expected DX, for them to still fight against it at every turn.While understandable in a way, as they are trained on average code and most code will now be written by the machines anyways making the DX \”less relevant\”, it’s also a giant code smell, as bad DX tends to point towards bad internals and wrong decisions along the way.So not really a technical limit – they swallow anything you throw at them, even the most complex cases – but more of an aesthetic limit in terms of taste.reply”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””,”login”:””,”Ask HN: Where have you found the coding limits of current models?”:”hypendev 27 minutes ago | prev | next [–]\nDepends – in the pure technical \”implementation\” level, only limits I’ve found are the ones with outdated knowledge – libraries, platforms, availability of some things.But one big limit is the DX. Their choice of DX is usually abysmal – ironically, just like an average devs.They seem to lack the aesthetic instinct for code, so you have to really point them hard into the direction or provide a sample of the expected DX, for them to still fight against it at every turn.While understandable in a way, as they are trained on average code and most code will now be written by the machines anyways making the DX \”less relevant\”, it’s also a giant code smell, as bad DX tends to point towards bad internals and wrong decisions along the way.So not really a technical limit – they swallow anything you throw at them, even the most complex cases – but more of an aesthetic limit in terms of taste.reply”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”netdevphoenix 33 minutes ago | prev | next [–]\nI think consistency levels when operating autonomously is still a challenge which you could refer to as their limit. You need to do so much around them to keep the consistency AND a decent level of performance.It’s like a savant with a short-attention span.reply”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””,”login”:””,”Ask HN: Where have you found the coding limits of current models?”:”netdevphoenix 33 minutes ago | prev | next [–]\nI think consistency levels when operating autonomously is still a challenge which you could refer to as their limit. You need to do so much around them to keep the consistency AND a decent level of performance.It’s like a savant with a short-attention span.reply”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”wood_spirit 33 minutes ago | prev | next [–]\nI found sonnet 4.5 struggled with a two pointer interval merge (two sorted lists of things with start stop timestamps), but opus 4.5 managed. Then it took opus 4.6 to make it a three way or k way merge. So it reminded me of the classic Simple made Easy talk by Rich Hickey where he talks about braids.Sonnet couldn’t track two twisted braids and opus 4.6 managed a weave.But the last few weeks Opus 4.6 seems to have got dumb again. Now it is making way more mistakes and forgetting useful things and recent context it used to manage.I am guessing this is just Anthropic quietly dialling down the real effort as they either downsize or free up compute for someone or something else.reply”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””,”login”:””,”Ask HN: Where have you found the coding limits of current models?”:”wood_spirit 33 minutes ago | prev | next [–]\nI found sonnet 4.5 struggled with a two pointer interval merge (two sorted lists of things with start stop timestamps), but opus 4.5 managed. Then it took opus 4.6 to make it a three way or k way merge. So it reminded me of the classic Simple made Easy talk by Rich Hickey where he talks about braids. Sonnet couldn’t track two twisted braids and opus 4.6 managed a weave.But the last few weeks Opus 4.6 seems to have got dumb again. Now it is making way more mistakes and forgetting useful things and recent context it used to manage.I am guessing this is just Anthropic quietly dialling down the real effort as they either downsize or free up compute for someone or something else.reply”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”rockbruno 20 minutes ago | prev | next [–]\nProblems that require deep knowledge of multiple repositories, e.g. when trying to debug issues involving dependencies. The models get confused very fast even with all code available locally, due to the size of the problem. But in my experience any kind of deep integration already messes up the models, even within a single repo.reply”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””,”login”:””,”Ask HN: Where have you found the coding limits of current models?”:”rockbruno 20 minutes ago | prev | next [–]\nProblems that require deep knowledge of multiple repositories, e.g. when trying to debug issues involving dependencies. The models get confused very fast even with all code available locally, due to the size of the problem.But in my experience any kind of deep integration already messes up the models, even within a single repo.reply”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”sshine 13 minutes ago | prev | next [–]\nDendritic Nix (either too new or too underrepresented in training data)Proper escaping of layered syntaxes in Ansible on the first attemptWriting bare-metal embedded Rust, although this was long ago, so not current modelsreply”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””,”login”:””,”Ask HN: Where have you found the coding limits of current models?”:”sshine 13 minutes ago | prev | next [–]\nDendritic Nix (either too new or too underrepresented in training data)Proper escaping of layered syntaxes in Ansible on the first attemptWriting bare-metal embedded Rust, although this was long ago, so not current modelsreply”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”dgb23 11 minutes ago | prev | next [–]\nI would push back on that question a little, because it has a baked in assumption about how these things work that conflict with my mental model and experience with them.The reason is that sometimes it spits out something or does a workflow that’s pretty sophisticated, and sometimes it fails spectacularly in the most basic ways.I don’t think there is a complexity or domain knowledge limit as there would be with a human. Or at least not in the same sense. As long as it can repeat and remix patterns that it is trained on, then it will do its thing well. The same seems to be true for \”reasoning\” loops and workflows.It can spit out code that has been done N times before in a similar manner for a large N.They can still break down because of very trivial issues and assumptions that happen to be baked in, go off the rails and get stuck long loops that are completely insane if you think of them as imitating human programmers.When I use an agent, I always interview it first about the task. Ask how they would go about it, probe them, give them info that they lack.Never go from prompt to action.Have them define their approach first, then split the approach into pieces, from gathering data to cleaning it up and so on. If applicable, front-load work that can be achieved with scripts, so you have testable and repeatable steps rather than let it go wild.So the TLDR is: I think the limitation is simply that it’s a non-deterministic token machine that produces useful results enough of the time so it appears to be reasonable.reply”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””,”login”:””,”Ask HN: Where have you found the coding limits of current models?”:”dgb23 11 minutes ago | prev | next [–]\nI would push back on that question a little, because it has a baked in assumption about how these things work that conflict with my mental model and experience with them.The reason is that sometimes it spits out something or does a workflow that’s pretty sophisticated, and sometimes it fails spectacularly in the most basic ways.I don’t think there is a complexity or domain knowledge limit as there would be with a human. Or at least not in the same sense. As long as it can repeat and remix patterns that it is trained on, then it will do its thing well. The same seems to be true for \”reasoning\” loops and workflows.It can spit out code that has been done N times before in a similar manner for a large N.They can still break down because of very trivial issues and assumptions that happen to be baked in, go off the rails and get stuck long loops that are completely insane if you think of them as imitating human programmers.When I use an agent, I always interview it first about the task. Ask how they would go about it, probe them, give them info that they lack.Never go from prompt to action.Have them define their approach first, then split the approach into pieces, from gathering data to cleaning it up and so on. If applicable, front-load work that can be achieved with scripts, so you have testable and repeatable steps rather than let it go wild.So the TLDR is: I think the limitation is simply that it’s a non-deterministic token machine that produces useful results enough of the time so it appears to be reasonable.reply”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”kaelandt 32 minutes ago | prev | next [–]\nIt’s just really hard for them to write non-verbose code. I don’t know if this is incentives from the providers to generate more tokens, but even with guidance on compact code, simple, etc, they just can’t really do it right now.reply”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””,”login”:””,”Ask HN: Where have you found the coding limits of current models?”:”kaelandt 32 minutes ago | prev | next [–]\nIt’s just really hard for them to write non-verbose code. I don’t know if this is incentives from the providers to generate more tokens, but even with guidance on compact code, simple, etc, they just can’t really do it right now.reply”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”keyle 35 minutes ago | prev | next [–]\nI haven’t found their limit, but I have found my limits, of waiting for them to do their thing.I tend to re-enjoy handrolling code more. I delegate the stuff that annoys me.We have a large multiplatform codebase, the issue seems to be more the time it takes to navigate the code and reason about it, rather than the size.Arguably the size is causing ‘them’ to be slower in that regards, but I haven’t found the limit yet. And with compaction, it’s even less of a problem.my 5c.reply”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””,”login”:””,”Ask HN: Where have you found the coding limits of current models?”:”keyle 35 minutes ago | prev | next [–]\nI haven’t found their limit, but I have found my limits, of waiting for them to do their thing.I tend to re-enjoy handrolling code more.I delegate the stuff that annoys me.We have a large multiplatform codebase, the issue seems to be more the time it takes to navigate the code and reason about it, rather than the size. Arguably the size is causing ‘them’ to be slower in that regards, but I haven’t found the limit yet. And with compaction, it’s even less of a problem.my 5c.reply”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”adventured 27 minutes ago | prev [–]\nThey don’t understand esoteric areas of computer science very well at all.I had a mistake in which a large back-up file deletion event happened during a robocopy. 600gb of files got ‘deleted’ (file headers toast etc).Trying to get the LLMs to understand the hunt parameters, what to focus on, what not to focus on – none of them could reasonably come close to doing file content recovery properly. I needed to build a custom solution because the available industry options couldn’t do what was required and the LLMs were useless for that (including the latest versions of Claude, Gemini and GPT). They just went around in circles, capped by their apparently weak knowledge of file recovery as a field.That is, creativity was their limitation.reply”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””,”login”:””,”Ask HN: Where have you found the coding limits of current models?”:”adventured 27 minutes ago | prev [–]\nThey don’t understand esoteric areas of computer science very well at all.I had a mistake in which a large back-up file deletion event happened during a robocopy. 600gb of files got ‘deleted’ (file headers toast etc).Trying to get the LLMs to understand the hunt parameters, what to focus on, what not to focus on – none of them could reasonably come close to doing file content recovery properly. I needed to build a custom solution because the available industry options couldn’t do what was required and the LLMs were useless for that (including the latest versions of Claude, Gemini and GPT). They just went around in circles, capped by their apparently weak knowledge of file recovery as a field.That is, creativity was their limitation.reply”},{“Hacker Newsnew | past | comments | ask | show | jobs | submitlogin”:”Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact\nSearch:”,”Hacker Newsnew | past | comments | ask | show | jobs | submit”:””}]
Andrii Bidochko
CTO UBOS
Andrii Bidochko is an AI entrepreneur and researcher focused on AI agents, reinforcement learning, and autonomous systems. He writes about the technologies shaping the future of machine intelligence, from frontier models and agent architectures to real-world AI applications.