Login
You're viewing the mstdn.social public feed.
  • BorgBackupborgbackup@fosstodon.org
    Jun 17, 2026, 8:29 PM

    @scy Guess some people in this thread who are making assumptions on how badly AI affects SW quality didn't try it recently or didn't use it in a good way.

    Maybe some have also the idea that all (human) developers are geniuses and write great, secure, safe, well-documented, bug-free code all the time.

    I can tell you from experience that the reality looks different.

    You also shouldn't just assume that AI is used irresponsibly if you don't have proof.

    💬 3🔄 1⭐ 3

Replies

  • BorgBackupborgbackup@fosstodon.org
    Jun 17, 2026, 8:35 PM

    E.g. did you notice how many bugs (sometimes very severe bugs) get fixed since recently, since they use AI to hunt bugs?

    These bugs are often old and were undiscovered many years long, even in open source projects.

    And these bugs were usually introduced by human developers, who are not perfect, obviously.

    So, better judge code quality by looking at the code, not by wild assumptions about whoever/whatever wrote the code.

    💬 2🔄 0⭐ 4
  • BorgBackupborgbackup@fosstodon.org
    Jun 17, 2026, 8:40 PM

    Also, don't assume that just because claude (or other AI) contributed to a commit means that it is purely AI generated and nobody looked at it.

    Usually, there is a human in the loop and the code gets reviewed and judged like any contribution.

    Also, it is usually not a one-shot, but a longer planning phase, followed by an implementation phase, running automated tests - all with human supervision and direction.

    💬 0🔄 0⭐ 1
  • scyscy@chaos.social
    Jun 17, 2026, 8:42 PM

    @borgbackup You're assuming that this is solely, or at least primarily, about code quality. But for a lot of people criticizing LLM usage, the main issues are moral and societal ones.

    💬 1🔄 0⭐ 15
  • BorgBackupborgbackup@fosstodon.org
    Jun 17, 2026, 8:51 PM

    @scy I totally see these issues.

    But barking at understaffed opensource projects is not the right way to get moral / societal / environmental issues fixed.

    These have to be addressed by government, law makers and also by prosecution if AI companies are doing illegal stuff.

    💬 1🔄 0⭐ 2
  • scyscy@chaos.social
    Jun 17, 2026, 8:54 PM

    @borgbackup I wasn't barking at you. I intentionally wasn't even tagging you. I was expressing disappointment at the state of things. The one replying with (unsolicited, btw) walls of text to defend themselves against accusations I didn't even make is you. 🤷‍♂️

    💬 1🔄 0⭐ 11
  • BorgBackupborgbackup@fosstodon.org
    Jun 17, 2026, 9:02 PM

    @scy I didn't say you were barking at me.

    This is just my general impression about how some people react to AI usage.

    Some even run around and advertise anti AI policies to projects they have never contributed to.

    I am all for activism if it is for a good cause, but one should know who is the right party to address.

    💬 1🔄 0⭐ 1
  • Faolanfaolan@tech.lgbt
    Jun 17, 2026, 9:58 PM

    @borgbackup the multiple unsolicited replies towards scy has also ensured I won’t look at or consider BorgBackup either.

    Reconsider your approach when trying to defend LLM usage as the current approach is not working. In my mind, usage of these in any form without putting the ethical concerns first is why I just move on to other non LLM touched options. OSS without ethics doesn’t make sense to me.

    💬 1🔄 0⭐ 11
  • 💬 0🔄 0⭐ 0
  • 💬 1🔄 0⭐ 5
  • BorgBackupborgbackup@fosstodon.org
    Jun 17, 2026, 8:46 PM

    @johnnythan That's a good point. 🙂

    I don't know how, but they somehow managed it that it usually does not output shitty code.

    Maybe rule compliance is better in AI than in humans, like writing and using tests, documentation, etc.

    Also, it can just try or look at some stuff much faster than humans can.

    Sometimes it is also running a wrong direction though, like a junior developer.

    That's why someone experienced has to look at the code and the design.

    💬 2🔄 0⭐ 0
  • Johnny Thanjohnnythan@tuebingen.network
    Jun 17, 2026, 8:54 PM

    @borgbackup Me personally, I like coding. I like writing line after line and see what happens.

    I don't like (never did) to debug someone else's code.

    So that is why I rather program myself, instead of thoroughly checking code outputs from a machine. That is just no fun (for me).

    And: Main part is not writing code. I think most of the times that's the easy and fun part. Having a good structure, a future-proof architecture, that is where the real thinking lies.

    For me.

    💬 0🔄 0⭐ 3
  • Daniel Bohrerdaniel_bohrer@chaos.social
    Jun 18, 2026, 7:16 AM

    @borgbackup @johnnythan A thought: if people cannot always successfully distinguish "bad" code from "good" code when writing it themselves, often while working with those codebases for multiple years, can it be possible for them to distinguish "bad" code from "good" code when an AI wrote it?

    💬 1🔄 0⭐ 5
  • BorgBackupborgbackup@fosstodon.org
    Jun 18, 2026, 4:08 PM

    @daniel_bohrer sure it is the same problem.

    also, if some contributor contributes code in a PR, you can not necessarily know whether AI was used or not.

    but for code review, it does not matter, you have to review it anyway to ensure that code quality is ok.

    overlooking something in review happens of course. AI can be useful as a 2nd or 3rd reviewer here, sometimes it finds stuff human reviewers have overlooked.

    💬 0🔄 0⭐ 0
  • Stefan Ihringercompfu@mograph.social
    Jun 18, 2026, 8:17 PM

    @borgbackup @scy I have an honest question because I don't know enough about Claude: why isn't an LLM coding assistant used like a text editor/IDE helper and is just not mentioned in commits? I don't read "Emacs contributed to this repo".

    What's the worst that could happen? You hurt the AI's feelings after Anthropic monetized what it stole from all of open source.

    💬 1🔄 0⭐ 0
  • Natanox 🇺🇦🇵🇸Natanox@chaos.social
    Jun 22, 2026, 2:28 PM

    @compfu @borgbackup @scy If I remember correctly Claude and others will usually add themselves to the PR if it's done by the LLM as well. For maintainers this can be beneficial (with some languages LLMs just seriously suck so much ass you can almost guarantee there's some nonsense in it), but I assume the AI corpos set it up that way as some kind of PR / marketing move (to show off their market penetration perhaps, or to "counter" only competitors showing up).

    💬 0🔄 0⭐ 1