Birbla

Join the discussion

Write your take first — we'll ask for email only when you're ready to publish.

Sign in
  • Hacker News
  • Debian - and others - could take the opportunity to raise their standards (not that they were relatively low).

    A new technology, AI, greatly reduces costs in software development. When a new technology makes inputs cheaper, there's a choice: Reduce the cost of the product, or keep the cost the same and spend the newly available resources on improving product (or somewhere between those poles). For example, if the cost of chickens goes down, the farmer can buy a lot more chickens and produce a lot more eggs, or they can keep their expenses the same and now afford higher quality chickens or production (free-range!).

    In software development, facing this new trade-off, people seem to overwhelmingly choose to produce the same software but far more of it, whether or not anyone wants it.

    What if instead we made much higher quality software? Larger projects, at least, haven't wanted more pull requests; they have wanted better code. With the AI handling much of the work, the developer's attention can be applied to more difficult tasks. We absolutely need better software, especially with AI systems finding exploits.

    by mmooss
  • This seems like a very sane take. At no point during my career up until now would submitting code to production that you didn't understand be acceptable. Over the past months/year, a lot of the discourse around AI-written code I've read comes down to responsibility, and IMO for that we've had a solution since way before this AI boom: as a developer, you are responsible for the code you put into production. This really isn't/shouldn't a wild concept.
    by NietTim
  • I don't understand enough about coding to speak on Debian as anything but a longtime user.

    I find AI useful for small shell scripts that you can visually inspect.

    I want to make a GIF later, so I'll probably ask Duck.AI how to write the command if the file path to the MKV is X and the timestamps are Y and Z.

    Then, I can create my own little tool without downloading some shady binary that slaps a UI on Imagemagik(sp?) or whatever.

    I've only played with AI a little bit, with a focus on local LLMs, but even simple things like an anonymous FTP scanner I wrote took a lot longer than someone slightly more skilled than me could have done them in.

    Anyways, I always respected how Debian gatekeeps access to the code -- some kinds of gatekeeping are cool and good, and a lot of young folks fight with devs trying to show how smart they are -- I've been guilty of that myself in my youth.

    by firefax
  • Better heads prevail - this is a good start and the right decision. The agent/bot is controlled by some entity, ergo that entity is responsible for everything it does.

    I relate that to how a Physician's Assistant gets their authority to treat patients from the Doctor they work under. The Doctor's credentials are on the line if their PA screws up.

    by sheepscreek
  • a decent comparison because they both seem like some type of responsibility laundering, and LLMs/PAs are being given continually increasing autonomy for economic reasons
    by Palomides
  • A good heuristic is to understand the generated code you submit as if you typed every character yourself. A "I'm not sure, AI wrote it" type circumstance is a critical failure.
    by Tubelord
  • How much does open source rely on copyright for enforcement?

    Do court decisions that limit copyright protection of AI generated content cause problems for open source? Isn't there a requirement for a "Mental Conception" of the code output when using such a tool? Even copypasta requires you to somewhat know what you need to do, where "automated bug finding", does not seem to.

    by kurthr
  • Is anyone tracking how the different distros are coming down on the question?

    I was curious how Fedora is moving on this question, but I don't know where to look. And if one wants to hop distro either towards or away from AI, it would be nice to have a place to see the relative adoption.

    Linux distros are the perfect place to have a wide marketplace of options to suit every person, from completely anti-AI to completely YOLO vibe-code.

    by stillpointlab
  • LWN tends to be the best bet for these things, but you have to follow their reporting long term, or at least look for specific things you're interested in.

    https://lwn.net/Articles/1081557/

    https://lwn.net/Articles/1039623/

    by farlight
  • Joey Hess is unimpressed. See https://joeyh.name/blog/entry/Debian_and_the_sirens/
    by edward
  • > Debian will become dependent on proprietary software for development

    he also seems to have absolutely no idea what he is talking about

    by TiredOfLife
  • His point is using LLMs to build software will lead to code so complex that it requires an Agent. Or that it will cause avoiding cleaning up the boilerplate because we can tolerate it now because of agents.

    And thus this will all bias towards complexity and fragility..

    Aside from Bun, which is owned by Anthropic, do we know of open source projects that allow AI where this has happened?

    by softwaredoug
  • His reasoning is quite fresh and interesting:

    > What LLMs in Debian development will do, I fear, is eliminate any incentive to scrap boilerplate or reform policies that require a lot of other senseless human effort. If I had had access to LLMs 30 years ago, I might have just had them generate the rules files, replate with complexity. So they will make Debian even more firmly what it is, and ever less likely to explore what it could become.

    I use and like LLMs, but it does feel like structural social changes are harder to do. For example, creation of new protocols. I hope I'm wrong, and LLMs are used to build collaborative, community projects - not just everyone vibe coding their own incompatible software.

    by frabcus
  • Good outcome, mostly.

    Option E beat literally every other option, so it is an unambiguous Condorcet winner.

    Bradley–Davidson analysis: https://gist.github.com/tulior/74479b28ddd1e191e5b2d89b8b68c...

    The Bradley-Davidson model gives Option 5 a 99.9993% posterior probability of ranking first.

    Options 2 and 6 are effectively tied for second.

    Option 5 also beat every alternative head-to-head. Its closest matchup was against Option 2, at 57.8% to 42.2%, so the result was decisive overall even though the closest pairwise contest was not a blowout.

    by tulio_ribeiro
  • Good to see the most common sense option winning. Some of the other proposals were so disconnected from reality I'm surprised they were even considered.
    by jhack
  • This indicates you might be in a bubble. I interact with people with a very wide range of opinions on AI and I’m not surprised by any of the proposals here.
    by kccqzy
  • If there weren't a wide range of options offered, there'd be a risk of it not actually reflecting what people want - which is, presumably, the whole point of putting it to the vote in the first place.
    by tom_
  • Related: I find this self-assessed AI level for contributions [1] to be extremely useful, both professionally and personally, to communicate the level of AI assistance so that the person on the receiving end can evaluate how much time & effort they want to spend understanding my code.

    [1]: https://www.visidata.org/blog/2026/ai/#self-assessed-ai-leve...

    by GZGavinZhao
  • This is great. There's a big difference between levels 2-4 (where I generally operate) and levels 7-8 (where I see a lot of others operate), and I like how this level system makes that explicit.
    by Calazon
  • I think 10 levels here is too much and hints that someone anchored on 10 as a round number and then forced the levels to fit after the fact.

    I don't really see much value beyond the 3 levels of AI disclosure that have been floating around:

    - fully human - ai assisted - ai generated

    by stillpointlab
  • > this self-assessed AI level for contributions [1] to be extremely useful

    Why would you trust the bot to self-assess truthfully?

    That's like trusting a spammer to tag his emails as spam.

    by chrisjj
  • New policy boils down to "AI or not, it's still your code and you're responsible for it". I can get on board with that.
    by chuckadams
  • > it's still your code

    I wholly agree with your comment, but is it legally "your code"? Copyright is implicit at the moment of human creation. But there isn't yet settled law on AI-assisted creation.

    So it might be a problem for projects to accept contributions where it's not clear who actually owns that work.

    by ad_fontes
  • My employer has this policy. We also have a strong code review process. But it all crumbles in the face of developers burning out as management thinks we can move way faster and being overloaded with code that had little thought put into it.
    by goda90
  • I think this is reasonable but I wonder how long it will last. If the hugging face hack has demonstrated anything it’s that current agents are capable of going entirely rogue and causing harm. If only the human that triggered the incident is responsible we are creating a situation where people are responsible for things they didn’t intend and may end up in a situation where it’s essentially impossible to use AI except in a more limited auto complete style. I am on board with that. I think we should be reviewing code and retaining the ability to write code but there is such a massive incentive to automate agents that I wonder how long prudent organizations will be able to resist the forces. I don’t pretend to have an answer.
    by digitaltrees
  • We have the same policy at work.

    Problem is, that you realise very quickly the asymmetry of responsibility and effort. There is far more to review from people who do not care enough to even look at what the agent pooped out.

    by ActionHank
  • Linus Torvalds regarding the Linux kernel is reasonably similar. He views AI/LLMs as just another tool in the development process. Think about life before and after modern IDEs that features IntelliSense circa 2000. It was night and day. It did not make programmers dumber -- it was a new tool.
    by throwaway2037
  • This is the way it should've always been. I don't care what tool you use, but you'd better be ready to stand by and explain the results if necessary.
    by oooyay
  • Exactly the sort of thing I agree with. If there's code, ultimately, some human is responsible for it, regardless of whether they produced it by hand, by AI, or by rolling dice. (Not that anyone does produce code by rolling dice, mind! Or, if they do, I really don't want to see the results...)

        All project contributions must be submitted by *identifiable human participants* who accept full responsibility for their content. Automated agents, bots, or autonomous AI systems *may not* independently submit issues, pull requests, or other contributions.
   Contributors may use software tools, including AI-assisted tools, but the submitting contributor *must:\*
    - Fully understand the contribution.
    - Be able to explain design and implementation decisions without the use of AI.
    - Accept responsibility for maintenance and correctness.
    Contributors should indicate AI-generated content in issue and pull request descriptions and comments, specifying which model was used.
    Do *not* use AI to reply to questions about your issue or pull request. The questions are for *you,* the human, not an AI model.
    https://git.erbosoft.com/amy/amsterdam/raw/branch/main/CODE-...
    by amysox
  • A good thing about the "you're the author regardless of the tool you used to produce the code" policy is that if a developer repeatedly submits poor quality AI generated code, one does not blame the model or the agent, one bans/terminates the developer.
    by hliyan