I really enjoy programming with AI agents. This is the closest I’ve ever come to having magical superpowers. Today’s AI is already smart, capable, resourceful and agentic enough to bring me joy.

This post is about the flip side of that: for most people I know, this fun turns into a death march. We have these powerful tools, but we feel more exhausted and overworked than ever. We ship more, but enjoy it less.

Source: https://frantic.im/bottleneck/

Sure thing you got disappointed - you chose not to train your mind, your confidence, your attitude… to stay self-sufficient and actually treasure achievements to be proud of. No. You chose to outsource, to depend, and to atrophy yourself and your history, where no names exist but tokens.

Congratulations. You damaged your own history.

  • REDACTED@infosec.pub
    link
    fedilink
    arrow-up
    6
    ·
    5 days ago

    I’m not part of this community and came here from the front page, but I gotta say - I do sometimes feel unreal working with codex. It solves issues in such a short time that I would not know where even to start. Then again, I’m not doing it for profit or to create stuff, but simply to solve issues and advance some DIY projects.

    Last night, before going to sleep, I got tired of trying to find and fix an issue in BeamNG with a large (1GB) mod where a car constantly failed to spawn for career. I asked it to try and fix it, took only 2 minutes, I thought to myself “bullshit”. Tested it and the fix worked.

    • Zachariah@lemmy.world
      link
      fedilink
      arrow-up
      6
      arrow-down
      1
      ·
      5 days ago

      How do you verify it didn’t break something else or introduce security or stability issues?

      • REDACTED@infosec.pub
        link
        fedilink
        arrow-up
        2
        ·
        5 days ago

        Trial and error. So far, the only issues I’ve encountered are small oversights, not critical errors.

        I did try working with codex (I think I used ChatGPT back then?) some 16 months ago, and half of the things I wanted to do ended up in a mess. Astra is the first one that seems to be actually somewhat usable (edit: but it’s also like 10-20x more expensive, I run into limits under 20€ subscription quite fast)

        On related note - It makes me uncomfortable to admit this, but the progress blows my mind away and it makes me uncomfortable, by now I truly believe it’s there to stay and will make so many jobs obsolete. I was part of the AI hating club 1 month ago, and cognitive dissonance is real

      • Nordon@lemmy.world
        link
        fedilink
        arrow-up
        3
        arrow-down
        1
        ·
        5 days ago

        You can read the code that it changed for starters. A lot of AI hate comes from not actually reviewing their work.

        2nd if it was a config issue. LLMs are fantastic at parsing an error log and tracking it back to a config issue. If it actually had to write code then it depends on if you actually read what it did.

    • hendrik@palaver.p3x.de
      link
      fedilink
      English
      arrow-up
      3
      ·
      5 days ago

      Yes, certainly feels unreal. I really don’t know what to feel. I mean on the one hand it generates lots of actually working code and boilerplate stuff. On the flipside it’s often messy and convoluted and I struggle to stay at the top of it. And I think the people who compare it to a slot machine have a point. I’ll end up in situations where it’ll either accomplish astonishing stuff. Or it’ll go on some crazy wild goose chase with me for like 2h straight. You never know.