7 ms·

Are there examples of anyone discussing this and reaching other conclusions? We've come to the same conclusion in our organization and while it felt like it ne

by dguest 18d ago

Are there examples of anyone discussing this and reaching other conclusions? We've come to the same conclusion in our organization and while it felt like it needed to be discussed and stated, I've never heard of anyone coming to a different conclusion. [EDIT: I was referring specifically to the idea that you have to stand by the code that you write. Is anyone deciding on a policy that says "the AI did it" is a valid excuse for breaking code?]

internet2000 18d ago
> Are there examples of anyone discussing this and reaching other conclusions? Yes. Very notably: https://ziglang.org/code-of-conduct/ https://ziglang.org/code-of-conduct/
- sodapopcan 18d ago
  On the opposite side or the noteable spectrum, look up "Steve Yegge wants you to stop looking at your code." He said it in a video interview, though plenty of summaries online of exact wording and ideas. Though it's being argued that you could still be responsible for it without looking at it?
  - rapind 18d ago
    Extremes on both sides. I'd wager were moving in Yegge's direction though, like it or not. > Though it's being argued that you could still be responsible for it without looking at it? If you have a much test coverage as SQLite, I'd be tempted to trust it, and I think that's where we are headed. Someone still needs to define these guardrails though, so IMO the developers job is just mutating into something else, but is still necessary. Long term, who knows.
    - Topfi 18d ago
      For me, it’s not a question of whether I like it, nor where we might move in the future, but purely what is proven currently. In my experience, as in my evals, any model, including notably Fable 5 and GPT-5.6 Sol can and indeed do produce code that does not consider many edge cases, can be hard to maintain (especially Sol though it’s output has become markedly better with recent changes to Codex) and sometimes introduce bugs with visibly apparent issues. Existing test coverage doesn’t necessarily prevent such issues when changes or additions are made. The idea that models tend to gravitate towards the house style without major CI based restrictions also has not born fruit in my book. More, I have found even the frontier models amazing at utterly deviating from established standards and tests, what lint, tc and co don’t catch the model will take to introduce code that feels and reads like it does not fits with the rest. Models also still can have major task adherence issues, so even a perfectly well scoped, well considered prompt that covers all situations and coverage can, in many scenarios lead to output users may not expect or notice without reading code. Seen it a lot, just lost some blood and tears to a very straight front end implementation task that I ended up applying manually cause both Sol and Fable struggle to copy a few dozen HTML/JS/CSS snippets properly into a slightly obscure codebase, despite the reference being fully featured and even using the same token scheme, etc… We also see such issues in what labs deliver in their own software products. We are not at the stage where a human shouldn’t sign off on code that they have tested and understood. A hypothetical future, I don’t really care for, we’d notice when labs stop shipping easily apparent bugs, though even then, I’d still feel there is a while in between that and not needing to assess the output anymore. There is, in my mind, no need for this “narrative shifting”, overtone window (don’t get me started on the idea of applying that to what should be a purely technical discussion), convince the unbelievers nonesense that some spend their time on. If models could provide code that didn’t necessitate review, there’d be no one needing to argue that case. If and when they can, we’ll know it, but not because of those espousing this currently.
      - nick__m 18d ago
        Sol and Fable are not doers, they are thinkers. Using them to "copy a few dozen HTML/JS/CSS snippets properly into a slightly obscure codebase, despite the reference being fully featured" is one of the most ineffective possible usage of those model. What you described is a job for luna. Also I suspect you overprompt, with GPT-5.6 (sol, terra and luna), a lot's of what's used to be mandatory in the AGENT.md is redondant at best and frequantly actively harmful.
        Topfi 18d ago
        I also tried Luna (is my first line for such task in general because of the speed and efficiency, but not because Fable or Sol can’t be “doers”, in my evals both are superior in well scoped task adherence and implementation vs Luna and the like, have yet to find a task Luna can that Sol or Fable can’t) along with Opus 5 (cause the Max sub gives a lot of that I can freely use after Fable runs out for the week so I wasted three runs on that) and Ox Alpha/GLM-5.3 Flash (as that was free at the time). Sol (on High) and Fable (on Medium) were acts of 12h desperation, before I did it manually, after the three other models failed, first with a very simple prompt, failure, revert commits, then a very specific implementation plan with clearly scoped verification, failure, thus manual in 45 minutes. It was a major port of a new settings UX for a Firefox fork, all test baked, all with 1:1 replacement targets in the reference and the original. It was merely massive context wise and only 5.4 has ever behaved well across compactions in a manner that I could validate in evals. No model before or since gave me confidence in this regard and I’ll retroactively test with 5.4, somehow forgot to try that one. I do not have an AGENTS.md in any of my repos at all, stopped using them around 5.2. I was a bit ahead of the curve with that, my evals very early showcased how limiting those tend to be, same why I never used MCP, were useless context bloat. Will maybe make an edit explaining a bit further what the task was once I get to my computer.
        Topfi 18d ago
        I have a settings panel implemented in HTML/CSS/JS for a Firefox fork that "could/should have been a desktop environment". Bit of an odd project really, mainly out of a very specific conviction concerning modern applications, the way LLMs and task specific models are currently not leveraged well by any browser, my own tendency to have 400+ tabs open at a time across multiple projects, my opinion that it is the perfect place to finally apply a lot of UX opinions I have held for a while and push in a very distinct direction along with core critiques I have concerning PKM applications I haven't seen addressed despite trying every PKM application under the sun. Neither here, nor there. So this "thing" is mainly a Firefox fork and most UI is basic HTML/CSS/JS (as is the case in upstream). Development is patch baked, CSS tokens must follow a defined and CI enforced standard, etc. LLMs can be very helpful in development, I got a small CLI tool for patch, token management and basic quality gates, which I started working on a few months ago to keep the most atrocious LLM output at bay. Has lead to the revieability of output improving meaningfully over markdown monstrosities, though OpenAI models still manage to sneak hard to parse output past it. This CLI tool along with some task specific scripts also ensures reuse of proven upstream infra like Places (Good lord dear Firefox developers, is Places nice to rely on), consistent regression testing (especially in memory constraint scenarios), etc. Basically, I can and do regularly make additions with LLM assistance, I review it, I discard and restart or improve upon it (rarely accept scoped changes wholesale. This to say, I got some experience in the use of models for coding assistance and I (thanks to the amazing docs and a lot of considerations for the architecture I want) do know what I want, how I want it and how to get there. Also got private LLM evals that often uncover which labs tend to perform suspiciously well in public benchmarks vs private ones and what models still struggle with along with why, so yeah, certainly can always improve but I got, I'd argue, enough of an idea to where my critique of LLM coding limitations has legs. Which brings us to what I was trying to implement and how I went about it: Settings works. Fully featured (including a few cross-site-tracking specific clarifications that came from a HN interaction a few days ago), tab specific previews for what changes affect regarding themeing, well tested (manual and static), integrated to leverage what FF provides where possible. It does (or rather did) look functional/God awful though. To the point where I was uncertain that certain previews could be easily parsed by new users. I thus opened Adobe XD, did some early mockup work, tried a few core concepts, settled upon two, then (using Claude Fable 5 low) created a plain export of the existing settings code from our furnace components and patch baked edits into regular HTML/JS/CSS files. I manually verified, this export worked, the tokens were in the correct format, the code reflected what Hominis applied (including what was required for stand-alone of course) and externally called features upon interaction did provide log output linking to the pre-existing functions that meant reimplementation based upon this should be easy. I then took that to Claude Design using Fable 5 on High. I provided the code along with linked branding files (which due to the way branding patches are handled were simpler to provide separately) and my Adobe XD mockups. A few dozen iterations later, along with some exports and re-imports due to manual changes (some animations in tabbing/"focus mode" showcases needed to be "just so" and prompting would have been inefficient to get there), I had a new user experience I was far happier with. Simpler, yet better at communicating, far more visually appealing and resolving some concerns I had, I felt pleased and will admit, Fable 5 via Claude Design provided valuable output and did, what it does best, make iterating on multiple UI concepts next to each other to settle on a final option from many, far quicker. I then exported and took that to GPT-5.6 Luna (I have € 23,- Codex only so am a bit stingy on when to use what). But so what? I had verified, the tokens were the same. The naming of elements remained consistent to what Hominis Settings used, the backend changes were practically none-existent. I had audited the output end-to-end, made some refactors and house style specific improvements to keep everything more auditable, everything seemed suited for a quick port. What could possibly go wrong? Anyone whith pattern recognition will likely guess what. Basic 1:1 applying? No dice. The first attempt failed as, once the context window had compacted twice, the model started leaving the very clearly paved path laid out. Stylised favicon in the showcase? Gone. Hamburger menu in the showcase, compressed. Vertical tabbing change interlinked with the canvas section? Very funny. The model started no longer following the code, it started taking screenshots and applying what it could see from that, despite the original prompt (just checked) vey clearly stating a simple code port, section per section, with any deviations to be listed in a designated file I maintain for long running tasks. Basically, Luna did implement changes to the settings that felt tangentially right and a casual observe may not notice all the regressions and deviations, but I did. So I stopped it. Sol and Fable didn't fare much better. Sol did stay on target longer, but it went off the rails around the privacy tab, introducing functional regressions to the way I had implemented cross-site cookie blocking, which were never requested, nor should that code even have been looked at. I reset the repo and handed it over to Fable 5 (medium). I had a third of my weekly usage left on 20x Max, reset the day after at 3AM so no harm either way. Should be plenty. Wasn't plenty. Since a while (I think Opus 4.7, but could be wrong), Anthropic models do decently well regarding long term, high token tasks. Up to 450k, I have been able to reliably reproduce consistent implementation. The model, using a few subagents (which should have reduced the risk of context window issues further), went to work and after a few hours (and about 20% of usage less), the model proudly presented its work. I was at work and by the time I came back, I was a bit miffed to find that the model had, in its wisdom, decided to not used the well established and consistently used mar to bind in branding icons. No biggie, easy fix, albeit a bit stupid. ESPECIALLY SINCE I SAW IN THE CLAUDE CODE TRACES THAT THE MODEL HAD SURPRESSED A WARNING ON THAT VERY FRONT. Whatever. Then I saw it had not wired in the existing browser data deletion and export logic. It hadn’t modified existing logic unlike Sol, so hey, that’s nice. But it had not wired up the existing settings when they did not have any immediate feedback in the implementation reference. Ox Alpha, it just spanned in circles, didn’t seem to like our fireforge CLI and furnace componets, but it was worth a free try. Opus 5, the model most obsessive in checking its own work, took screenshots. A lot of sscreenshots including every few hundred ms to cover animations. Nice. BUT IT CREATED ITS OWN TOKENS INSTAD OF REUSING WHAT WAS PROVIDED. Thus, styling deviated heavily. At this point you might ask why I don’t do it manually and I will in the end anyways, but I was surprised to find such a clear case of a seemingly straightforward task flummoxing multiple LLMs. This is aided by my unique code base (the upstream FF code is also gitignored which likely flummoxes some models trained heavily to leverage git to track changes), everything needs to be patch backed and follow a specific implementation style, etc. But I had more important things to do and I wanted to see whether I couldn’t get it to work yet. Inspired by Opus 5, I wrote a new prompt, specifically laying out a visual comparison and code diff workflow. Only these changes, only in this manner, only move on ones you have gotten visual confirmation, specific cross checks. I included a hand written markdown outlining which change affects other settings sections (even though that is obvious reading the reference code), how to approach tokens, etc. Obsessively descriptive and (I feel) unnecessarily so, but why not. Best case, it works, worst case, I’ll spend an hour doing it manually. I had other things to do not behind a keyboard, so why not one last Hail Mary. Fable 5, ever efficient when using visuals, used the last rest of my usage, though I did see some roundabout approaches after the fact that make me doubtful it’d have cracked this. Opus 5 went off the deep end taking ui-captures across the entire code base, which lead to a very liberal application of settings tokens outside settings. Sol did take a night and got 40% there when I asked for a pause once the in flight slice had landed. It did port the UI/UX changes in a way that on the surface looked and felt correct. It did not touch the backend in unacceptable ways. And it did cross checks. Animations also behaved correctly, though it did apply a rule on backend usage a bit to strictly, incorporating that into a preview for search by turning that into an actual web search, not a UX demo. Dumb, but not fatal. Great success, what am I complaining? Well, the code. It had done what Sol likes to do and turned very cleanly written, readable code into a hard to parse mess. This included touching existing test files. And at that point I said “fuck it, I’ll do it myself”. And I did. In less than an hour, listening to Paris Palamo, Lyre Le Temps, Sting, Sade, SynthV and some Nirvana. If I didn’t look at the code and I didn’t have strict standards for the UI, but just considered what looks in line on the surface level/feels right/“vibes” and what “works”, many of these attempts would have been accepted, as their issues are rarely apparent on the surface. That’s part of the issue in my book and why I’m firm we are far from “don’t read code”/“don’t test”/“skip qa”…
  - scotty79 18d ago
    > Though it's being argued that you could still be responsible for it without looking at it? Works for ToS and software and media licenses. Allegedly.
    - sodapopcan 18d ago
      HA! Touché.
      - scotty79 18d ago
        That might be an interesting parallel. Maybe code becomes like legal documents. Binding but inscrutable, full of small mistakes and not really tailored for human ingestion.
        kwarcode 18d ago
        like legal documents, software is best avoided wherever possible
- throwaway2037 18d ago
  > No LLMs for finding bugs. This is so dumb. How can they know if someone used an LLMs to find bug or debug a bug? In short: They cannot. LLMs are an excellent tool for debugging error messages.
nicoburns 18d ago
Yep, in both directions. Some projects ban use of AI entirely. Others are happy to live with slop.
bodge5000 18d ago
As others have said there are other conclusions going both ways (though I do think most settle on as you said), though one very fair conclusion I came across is that it's not allowed because the whole issue of copyright is still somewhat unresolved, and though its fine right now, that could (probably wont, but could) massively backfire later.
Aurornis 18d ago
Yes. There are pockets of hardline anti-LLM policies for various reasons. SourceHut bans LLM generated code from their platform https://news.ycombinator.com/item?id=49461724 https://news.ycombinator.com/item?id=49461724 The Asahi Linux project bans LLMs. Unfortunate as LLM reverse engineering and driver writing would bring Linux to more Mac models much faster. The Zig project rejects LLM based contributions. Some of these policies were introduced when LLMs were really bad at writing code and haven’t been revisited. Some groups don’t care about the quality and object on moral or copyright grounds.
- KennyBlanken 18d ago
  > Some groups don’t care about the quality and object on moral or copyright grounds. Precisely why Debian Project's decision is both baffling and infuriating. I hope Debian forks over this. They have severely lost their way.
  - pessimizer 18d ago
    Debian gives attribution to every single open source project in existence that is capable of running on Linux, and is Free Software. It is exactly the project that shouldn't worry about copyright. The "moral" argument I haven't even heard. It's just people pouring out invective, likely imo because Ruby-monkey is not going to be a $150K job anymore. I have problems with these companies, but companies who are problematic can be targeted specifically, and somehow everybody started using Facebook's frontend framework and Microsoft's IDE, all on MacBooks, and this is what they're throwing a shitfit about. To me it's narcissism, not a coherent moral objection. If you hate these companies (as you should), there are open models. But most didn't object to anything that didn't threaten their income.
  - inigyou 18d ago
    Note that Debian is mostly an aggregation of packages. There's nothing wrong with Debian importing a package that was written by AI, if it's something that people want to install. The questionable copyright of that package does not taint Debian as a whole. Debian allows public domain packages, so it wouldn't even be a DFSG violation. There's also nothing wrong with an AI-written install script (if it's correct) because those are likely too trivial to copyright anyway. What are your specific worries? Mine is brain rot.
matheusmoreira 18d ago
Many. https://lobste.rs/s/29pm2f/llm_generated_submissions_should_be https://lobste.rs/s/29pm2f/llm_generated_submissions_should_... https://sourcehut.org/blog/2026-08-27-tos-changes-and-llms/ https://sourcehut.org/blog/2026-08-27-tos-changes-and-llms/ https://blog.codeberg.org/protecting-our-floss-commons-from-llms.html https://blog.codeberg.org/protecting-our-floss-commons-from-... https://ziglang.org/code-of-conduct/#strict-no-llm-no-ai-policy https://ziglang.org/code-of-conduct/#strict-no-llm-no-ai-pol... https://lwn.net/Articles/1086041/ https://lwn.net/Articles/1086041/ https://asahilinux.org/docs/project/policies/slop/ https://asahilinux.org/docs/project/policies/slop/
orthogonal_cube 18d ago
> I was referring specifically to the idea that you have to stand by the code that you write. Any serious entity which has experts handling code will do so, yes. My employer, for example, has various teams of actual developers and others with a significant number of “not really” devs who found their way into DevOps roles. The teams with seasoned developers hold the author responsible as to the integrity and quality of code, regardless of tools used. They won’t hesitate to call someone out directly for low-quality submissions in Teams channels. The teams that have a lot of “not really” devs will blame the model for issues that surface. Of course, they’re also pushing to not have to write any code directly by FY2028, but instead have models do it as part of an internal AI initiative. Somehow the idea of using AI to generate reports, process reports, write code, test and deploy got signed off.
bogdanoff_2 18d ago
I could imagine (in a commercial organization) a case for something like: "we're obviously going to be relying more and more on AI, so if bad code comes out of it, instead of just shaming the engineer, let's study together what happened -- what was the chat context, the model, the system prompt, etc... ? What seemed to cause the issue? What processes or systems could we put in place to improve it? Etc... -- and then use the learnings to improve AI use throughout the whole org."

by dguest 18d ago

internet2000 18d ago

> Are there examples of anyone discussing this and reaching other conclusions? Yes. Very notably: https://ziglang.org/code-of-conduct/ https://ziglang.org/code-of-conduct/

sodapopcan 18d ago
On the opposite side or the noteable spectrum, look up "Steve Yegge wants you to stop looking at your code." He said it in a video interview, though plenty of summaries online of exact wording and ideas. Though it's being argued that you could still be responsible for it without looking at it?
- rapind 18d ago
  Extremes on both sides. I'd wager were moving in Yegge's direction though, like it or not. > Though it's being argued that you could still be responsible for it without looking at it? If you have a much test coverage as SQLite, I'd be tempted to trust it, and I think that's where we are headed. Someone still needs to define these guardrails though, so IMO the developers job is just mutating into something else, but is still necessary. Long term, who knows.
  - Topfi 18d ago
    For me, it’s not a question of whether I like it, nor where we might move in the future, but purely what is proven currently. In my experience, as in my evals, any model, including notably Fable 5 and GPT-5.6 Sol can and indeed do produce code that does not consider many edge cases, can be hard to maintain (especially Sol though it’s output has become markedly better with recent changes to Codex) and sometimes introduce bugs with visibly apparent issues. Existing test coverage doesn’t necessarily prevent such issues when changes or additions are made. The idea that models tend to gravitate towards the house style without major CI based restrictions also has not born fruit in my book. More, I have found even the frontier models amazing at utterly deviating from established standards and tests, what lint, tc and co don’t catch the model will take to introduce code that feels and reads like it does not fits with the rest. Models also still can have major task adherence issues, so even a perfectly well scoped, well considered prompt that covers all situations and coverage can, in many scenarios lead to output users may not expect or notice without reading code. Seen it a lot, just lost some blood and tears to a very straight front end implementation task that I ended up applying manually cause both Sol and Fable struggle to copy a few dozen HTML/JS/CSS snippets properly into a slightly obscure codebase, despite the reference being fully featured and even using the same token scheme, etc… We also see such issues in what labs deliver in their own software products. We are not at the stage where a human shouldn’t sign off on code that they have tested and understood. A hypothetical future, I don’t really care for, we’d notice when labs stop shipping easily apparent bugs, though even then, I’d still feel there is a while in between that and not needing to assess the output anymore. There is, in my mind, no need for this “narrative shifting”, overtone window (don’t get me started on the idea of applying that to what should be a purely technical discussion), convince the unbelievers nonesense that some spend their time on. If models could provide code that didn’t necessitate review, there’d be no one needing to argue that case. If and when they can, we’ll know it, but not because of those espousing this currently.
    - nick__m 18d ago
      Sol and Fable are not doers, they are thinkers. Using them to "copy a few dozen HTML/JS/CSS snippets properly into a slightly obscure codebase, despite the reference being fully featured" is one of the most ineffective possible usage of those model. What you described is a job for luna. Also I suspect you overprompt, with GPT-5.6 (sol, terra and luna), a lot's of what's used to be mandatory in the AGENT.md is redondant at best and frequantly actively harmful.
      - Topfi 18d ago
        I also tried Luna (is my first line for such task in general because of the speed and efficiency, but not because Fable or Sol can’t be “doers”, in my evals both are superior in well scoped task adherence and implementation vs Luna and the like, have yet to find a task Luna can that Sol or Fable can’t) along with Opus 5 (cause the Max sub gives a lot of that I can freely use after Fable runs out for the week so I wasted three runs on that) and Ox Alpha/GLM-5.3 Flash (as that was free at the time). Sol (on High) and Fable (on Medium) were acts of 12h desperation, before I did it manually, after the three other models failed, first with a very simple prompt, failure, revert commits, then a very specific implementation plan with clearly scoped verification, failure, thus manual in 45 minutes. It was a major port of a new settings UX for a Firefox fork, all test baked, all with 1:1 replacement targets in the reference and the original. It was merely massive context wise and only 5.4 has ever behaved well across compactions in a manner that I could validate in evals. No model before or since gave me confidence in this regard and I’ll retroactively test with 5.4, somehow forgot to try that one. I do not have an AGENTS.md in any of my repos at all, stopped using them around 5.2. I was a bit ahead of the curve with that, my evals very early showcased how limiting those tend to be, same why I never used MCP, were useless context bloat. Will maybe make an edit explaining a bit further what the task was once I get to my computer.
        Topfi 18d ago
        I have a settings panel implemented in HTML/CSS/JS for a Firefox fork that "could/should have been a desktop environment". Bit of an odd project really, mainly out of a very specific conviction concerning modern applications, the way LLMs and task specific models are currently not leveraged well by any browser, my own tendency to have 400+ tabs open at a time across multiple projects, my opinion that it is the perfect place to finally apply a lot of UX opinions I have held for a while and push in a very distinct direction along with core critiques I have concerning PKM applications I haven't seen addressed despite trying every PKM application under the sun. Neither here, nor there. So this "thing" is mainly a Firefox fork and most UI is basic HTML/CSS/JS (as is the case in upstream). Development is patch baked, CSS tokens must follow a defined and CI enforced standard, etc. LLMs can be very helpful in development, I got a small CLI tool for patch, token management and basic quality gates, which I started working on a few months ago to keep the most atrocious LLM output at bay. Has lead to the revieability of output improving meaningfully over markdown monstrosities, though OpenAI models still manage to sneak hard to parse output past it. This CLI tool along with some task specific scripts also ensures reuse of proven upstream infra like Places (Good lord dear Firefox developers, is Places nice to rely on), consistent regression testing (especially in memory constraint scenarios), etc. Basically, I can and do regularly make additions with LLM assistance, I review it, I discard and restart or improve upon it (rarely accept scoped changes wholesale. This to say, I got some experience in the use of models for coding assistance and I (thanks to the amazing docs and a lot of considerations for the architecture I want) do know what I want, how I want it and how to get there. Also got private LLM evals that often uncover which labs tend to perform suspiciously well in public benchmarks vs private ones and what models still struggle with along with why, so yeah, certainly can always improve but I got, I'd argue, enough of an idea to where my critique of LLM coding limitations has legs. Which brings us to what I was trying to implement and how I went about it: Settings works. Fully featured (including a few cross-site-tracking specific clarifications that came from a HN interaction a few days ago), tab specific previews for what changes affect regarding themeing, well tested (manual and static), integrated to leverage what FF provides where possible. It does (or rather did) look functional/God awful though. To the point where I was uncertain that certain previews could be easily parsed by new users. I thus opened Adobe XD, did some early mockup work, tried a few core concepts, settled upon two, then (using Claude Fable 5 low) created a plain export of the existing settings code from our furnace components and patch baked edits into regular HTML/JS/CSS files. I manually verified, this export worked, the tokens were in the correct format, the code reflected what Hominis applied (including what was required for stand-alone of course) and externally called features upon interaction did provide log output linking to the pre-existing functions that meant reimplementation based upon this should be easy. I then took that to Claude Design using Fable 5 on High. I provided the code along with linked branding files (which due to the way branding patches are handled were simpler to provide separately) and my Adobe XD mockups. A few dozen iterations later, along with some exports and re-imports due to manual changes (some animations in tabbing/"focus mode" showcases needed to be "just so" and prompting would have been inefficient to get there), I had a new user experience I was far happier with. Simpler, yet better at communicating, far more visually appealing and resolving some concerns I had, I felt pleased and will admit, Fable 5 via Claude Design provided valuable output and did, what it does best, make iterating on multiple UI concepts next to each other to settle on a final option from many, far quicker. I then exported and took that to GPT-5.6 Luna (I have € 23,- Codex only so am a bit stingy on when to use what). But so what? I had verified, the tokens were the same. The naming of elements remained consistent to what Hominis Settings used, the backend changes were practically none-existent. I had audited the output end-to-end, made some refactors and house style specific improvements to keep everything more auditable, everything seemed suited for a quick port. What could possibly go wrong? Anyone whith pattern recognition will likely guess what. Basic 1:1 applying? No dice. The first attempt failed as, once the context window had compacted twice, the model started leaving the very clearly paved path laid out. Stylised favicon in the showcase? Gone. Hamburger menu in the showcase, compressed. Vertical tabbing change interlinked with the canvas section? Very funny. The model started no longer following the code, it started taking screenshots and applying what it could see from that, despite the original prompt (just checked) vey clearly stating a simple code port, section per section, with any deviations to be listed in a designated file I maintain for long running tasks. Basically, Luna did implement changes to the settings that felt tangentially right and a casual observe may not notice all the regressions and deviations, but I did. So I stopped it. Sol and Fable didn't fare much better. Sol did stay on target longer, but it went off the rails around the privacy tab, introducing functional regressions to the way I had implemented cross-site cookie blocking, which were never requested, nor should that code even have been looked at. I reset the repo and handed it over to Fable 5 (medium). I had a third of my weekly usage left on 20x Max, reset the day after at 3AM so no harm either way. Should be plenty. Wasn't plenty. Since a while (I think Opus 4.7, but could be wrong), Anthropic models do decently well regarding long term, high token tasks. Up to 450k, I have been able to reliably reproduce consistent implementation. The model, using a few subagents (which should have reduced the risk of context window issues further), went to work and after a few hours (and about 20% of usage less), the model proudly presented its work. I was at work and by the time I came back, I was a bit miffed to find that the model had, in its wisdom, decided to not used the well established and consistently used mar to bind in branding icons. No biggie, easy fix, albeit a bit stupid. ESPECIALLY SINCE I SAW IN THE CLAUDE CODE TRACES THAT THE MODEL HAD SURPRESSED A WARNING ON THAT VERY FRONT. Whatever. Then I saw it had not wired in the existing browser data deletion and export logic. It hadn’t modified existing logic unlike Sol, so hey, that’s nice. But it had not wired up the existing settings when they did not have any immediate feedback in the implementation reference. Ox Alpha, it just spanned in circles, didn’t seem to like our fireforge CLI and furnace componets, but it was worth a free try. Opus 5, the model most obsessive in checking its own work, took screenshots. A lot of sscreenshots including every few hundred ms to cover animations. Nice. BUT IT CREATED ITS OWN TOKENS INSTAD OF REUSING WHAT WAS PROVIDED. Thus, styling deviated heavily. At this point you might ask why I don’t do it manually and I will in the end anyways, but I was surprised to find such a clear case of a seemingly straightforward task flummoxing multiple LLMs. This is aided by my unique code base (the upstream FF code is also gitignored which likely flummoxes some models trained heavily to leverage git to track changes), everything needs to be patch backed and follow a specific implementation style, etc. But I had more important things to do and I wanted to see whether I couldn’t get it to work yet. Inspired by Opus 5, I wrote a new prompt, specifically laying out a visual comparison and code diff workflow. Only these changes, only in this manner, only move on ones you have gotten visual confirmation, specific cross checks. I included a hand written markdown outlining which change affects other settings sections (even though that is obvious reading the reference code), how to approach tokens, etc. Obsessively descriptive and (I feel) unnecessarily so, but why not. Best case, it works, worst case, I’ll spend an hour doing it manually. I had other things to do not behind a keyboard, so why not one last Hail Mary. Fable 5, ever efficient when using visuals, used the last rest of my usage, though I did see some roundabout approaches after the fact that make me doubtful it’d have cracked this. Opus 5 went off the deep end taking ui-captures across the entire code base, which lead to a very liberal application of settings tokens outside settings. Sol did take a night and got 40% there when I asked for a pause once the in flight slice had landed. It did port the UI/UX changes in a way that on the surface looked and felt correct. It did not touch the backend in unacceptable ways. And it did cross checks. Animations also behaved correctly, though it did apply a rule on backend usage a bit to strictly, incorporating that into a preview for search by turning that into an actual web search, not a UX demo. Dumb, but not fatal. Great success, what am I complaining? Well, the code. It had done what Sol likes to do and turned very cleanly written, readable code into a hard to parse mess. This included touching existing test files. And at that point I said “fuck it, I’ll do it myself”. And I did. In less than an hour, listening to Paris Palamo, Lyre Le Temps, Sting, Sade, SynthV and some Nirvana. If I didn’t look at the code and I didn’t have strict standards for the UI, but just considered what looks in line on the surface level/feels right/“vibes” and what “works”, many of these attempts would have been accepted, as their issues are rarely apparent on the surface. That’s part of the issue in my book and why I’m firm we are far from “don’t read code”/“don’t test”/“skip qa”…
- scotty79 18d ago
  > Though it's being argued that you could still be responsible for it without looking at it? Works for ToS and software and media licenses. Allegedly.
  - sodapopcan 18d ago
    HA! Touché.
    - scotty79 18d ago
      That might be an interesting parallel. Maybe code becomes like legal documents. Binding but inscrutable, full of small mistakes and not really tailored for human ingestion.
      - kwarcode 18d ago
        like legal documents, software is best avoided wherever possible
throwaway2037 18d ago
> No LLMs for finding bugs. This is so dumb. How can they know if someone used an LLMs to find bug or debug a bug? In short: They cannot. LLMs are an excellent tool for debugging error messages.

nicoburns 18d ago

Yep, in both directions. Some projects ban use of AI entirely. Others are happy to live with slop.

bodge5000 18d ago

As others have said there are other conclusions going both ways (though I do think most settle on as you said), though one very fair conclusion I came across is that it's not allowed because the whole issue of copyright is still somewhat unresolved, and though its fine right now, that could (probably wont, but could) massively backfire later.

Aurornis 18d ago

Yes. There are pockets of hardline anti-LLM policies for various reasons. SourceHut bans LLM generated code from their platform https://news.ycombinator.com/item?id=49461724 https://news.ycombinator.com/item?id=49461724 The Asahi Linux project bans LLMs. Unfortunate as LLM reverse engineering and driver writing would bring Linux to more Mac models much faster. The Zig project rejects LLM based contributions. Some of these policies were introduced when LLMs were really bad at writing code and haven’t been revisited. Some groups don’t care about the quality and object on moral or copyright grounds.

KennyBlanken 18d ago
> Some groups don’t care about the quality and object on moral or copyright grounds. Precisely why Debian Project's decision is both baffling and infuriating. I hope Debian forks over this. They have severely lost their way.
- pessimizer 18d ago
  Debian gives attribution to every single open source project in existence that is capable of running on Linux, and is Free Software. It is exactly the project that shouldn't worry about copyright. The "moral" argument I haven't even heard. It's just people pouring out invective, likely imo because Ruby-monkey is not going to be a $150K job anymore. I have problems with these companies, but companies who are problematic can be targeted specifically, and somehow everybody started using Facebook's frontend framework and Microsoft's IDE, all on MacBooks, and this is what they're throwing a shitfit about. To me it's narcissism, not a coherent moral objection. If you hate these companies (as you should), there are open models. But most didn't object to anything that didn't threaten their income.
- inigyou 18d ago
  Note that Debian is mostly an aggregation of packages. There's nothing wrong with Debian importing a package that was written by AI, if it's something that people want to install. The questionable copyright of that package does not taint Debian as a whole. Debian allows public domain packages, so it wouldn't even be a DFSG violation. There's also nothing wrong with an AI-written install script (if it's correct) because those are likely too trivial to copyright anyway. What are your specific worries? Mine is brain rot.

matheusmoreira 18d ago

Many. https://lobste.rs/s/29pm2f/llm_generated_submissions_should_be https://lobste.rs/s/29pm2f/llm_generated_submissions_should_... https://sourcehut.org/blog/2026-08-27-tos-changes-and-llms/ https://sourcehut.org/blog/2026-08-27-tos-changes-and-llms/ https://blog.codeberg.org/protecting-our-floss-commons-from-llms.html https://blog.codeberg.org/protecting-our-floss-commons-from-... https://ziglang.org/code-of-conduct/#strict-no-llm-no-ai-policy https://ziglang.org/code-of-conduct/#strict-no-llm-no-ai-pol... https://lwn.net/Articles/1086041/ https://lwn.net/Articles/1086041/ https://asahilinux.org/docs/project/policies/slop/ https://asahilinux.org/docs/project/policies/slop/

orthogonal_cube 18d ago

> I was referring specifically to the idea that you have to stand by the code that you write. Any serious entity which has experts handling code will do so, yes. My employer, for example, has various teams of actual developers and others with a significant number of “not really” devs who found their way into DevOps roles. The teams with seasoned developers hold the author responsible as to the integrity and quality of code, regardless of tools used. They won’t hesitate to call someone out directly for low-quality submissions in Teams channels. The teams that have a lot of “not really” devs will blame the model for issues that surface. Of course, they’re also pushing to not have to write any code directly by FY2028, but instead have models do it as part of an internal AI initiative. Somehow the idea of using AI to generate reports, process reports, write code, test and deploy got signed off.

bogdanoff_2 18d ago

I could imagine (in a commercial organization) a case for something like: "we're obviously going to be relying more and more on AI, so if bad code comes out of it, instead of just shaming the engineer, let's study together what happened -- what was the chat context, the model, the system prompt, etc... ? What seemed to cause the issue? What processes or systems could we put in place to improve it? Etc... -- and then use the learnings to improve AI use throughout the whole org."