• kadu@scribe.disroot.org
    link
    fedilink
    arrow-up
    86
    ·
    edit-2
    2 days ago

    I have a visceral “AI” sensor that triggers when I see these:

    “Rust Implementation (v2)”

    “Performance Benchmarks (Validated)”

    Human beings don’t self-validate explicitly like that. AI loves doing it.

    You generate code, there’s a bug, you ask for a fix, your AI of choice will always output with:

    *** Fix build issue ***

    *** End fix ***

    and then call it “Version 2 (Validated)”.

    Sometimes it’s more subtle, but you can feel it, it loves adding “confirmed”, “working”, “validated”.

    • FishFace@piefed.social
      link
      fedilink
      English
      arrow-up
      53
      arrow-down
      2
      ·
      2 days ago

      My sensor is much simpler. If I see emoji in headings or bulleted lists, I assume it’s shit. It might be AI slop, or it might just be kids getting overexcited with the little pictures, but both deserve suspicion and scrutiny.

      If a bunch of the emoji don’t even make sense it can get in the bin.

      • Grey Cat@lemmy.world
        link
        fedilink
        English
        arrow-up
        5
        ·
        9 hours ago

        Ahhh idk, I saw a lot of genuine repos do emojis, at least for headings. Even before LLMs.

        I like them 'cause with the right amount, it makes a README easier to parse when quickly scrolling over it.

        • FishFace@piefed.social
          link
          fedilink
          English
          arrow-up
          1
          ·
          7 hours ago

          As an ancient husk of a person, it all looks crack-addled to me. I don’t really see how you can parse out headings from emoji because their usage isn’t consistent.

      • PlexSheep@infosec.pub
        link
        fedilink
        arrow-up
        1
        ·
        7 hours ago

        I like putting the little pictures in my readmes sometimes. In my biologically generated repositories. Please don’t discriminate against neat little pictures you can just put in text 🐑.

    • MajinBlayze@lemmy.world
      link
      fedilink
      arrow-up
      16
      arrow-down
      2
      ·
      edit-2
      2 days ago

      I have a project with a bunch of compose files that define the services I self host. I “deploy” the project by sshing into my server and doing “git pull” which means I’m often making changes that don’t get tested before committing to source control. As a result I have long chains of commits like:

      • refactor the sproingy widget
      • refactor the sproingy widget v2
      • refactor the sproingy widget working
      • maybe the sproingy widget works this time?
      • ok finally found the issue with refactor sproingy widget
      • fix formatting of sproingy widget

      And now I’m wondering if I’ve been an llm this whole time

      • kkj@lemmy.dbzer0.com
        link
        fedilink
        English
        arrow-up
        2
        ·
        19 hours ago

        Why not just edit the YAML directly on the server via a command-line text editor or SSHFS and then push from there when it works?

      • exu@feditown.com
        link
        fedilink
        English
        arrow-up
        14
        arrow-down
        1
        ·
        2 days ago

        Make your changes in a new branch and rebase/squash when you push it to main.

        • IronKrill@lemmy.ca
          link
          fedilink
          arrow-up
          6
          ·
          2 days ago

          This also means modifying your git pull command to pull the correct branch. A small change perhaps, but may be harder than just committing to main lol.

          I had a similar problem with GitHub actions, it was hard to test without messing up the main repo history.