• saltesc@lemmy.world
    link
    fedilink
    arrow-up
    16
    ·
    11 months ago

    It’s like handing a job to the new guy.

    I use it for snippets. Always double check though and always need to remind it of methods that will and will not work. Often it’s just inspiration or for doing lengthy subquery style stuff that takes ages to type out. But even then, 10 mins of battling it when I didn’t want to take 5 mins to just manually do it is sadly common.

    I do like asking it to optimise and it can’t find anything that could be better, though. We’re friends then.

    • karmiclychee @sh.itjust.works
      link
      fedilink
      arrow-up
      2
      ·
      11 months ago

      Boilerplates, or solutions to really weird one off problems that require a bunch of synthesis across different, unfamiliar domains (ie, as a web dev, how do I install imagemagick on an internal, stripped down rootless RHEL container with microdnf)

      • Pennomi@lemmy.world
        link
        fedilink
        English
        arrow-up
        1
        ·
        11 months ago

        Yeah, or implementing well-defined single functions that would otherwise just be a hassle. I’ll just stub out the function signature and tell it to finish the rest.

  • namelivia@lemmy.world
    link
    fedilink
    arrow-up
    10
    ·
    11 months ago

    I find it actually useful to write code you understand and you could have written yourself, you are just lazy to type.

    But I cannot imagine developing using ChatGPT without actually knowing what is in the output or what specifically ask for.

  • fmstrat@lemmy.nowsci.com
    link
    fedilink
    English
    arrow-up
    9
    arrow-down
    1
    ·
    11 months ago

    In a year, ChatGPT will be the #1 contributor to code rot and things that need to be refactored.

  • NounsAndWords@lemmy.world
    link
    fedilink
    arrow-up
    6
    ·
    11 months ago

    If it can tell me what the code I’m looking at (that I wrote yesterday) is supposed to do, that’s enough for me.

    • Tixanou@lemmy.world
      link
      fedilink
      Español
      arrow-up
      2
      ·
      11 months ago

      I’m sorry, but as a large AI language model, I don’t have the ability to debug this code.

  • karmiclychee @sh.itjust.works
    link
    fedilink
    arrow-up
    3
    ·
    11 months ago

    I used it a bit during the first hype wave. It was neat, but immediately I found I hadn’t actually learned anything. When I went back to the code, there wasn’t a matching trail of purple links though Google or stack overflow for me to follow and retrace my steps. That probably reinforced for me that programming is about design choices, and provenance of ideas, more than anything else.

    • TrickDacy@lemmy.world
      link
      fedilink
      arrow-up
      1
      ·
      11 months ago

      You’re not supposed to let ai write code you don’t understand. It’s meant to help with syntax you don’t remember (but would recognize upon seeing), reduce typing, and get creative juices flowing by presenting a way of doing the things.