• Flamekebab@piefed.social
      link
      fedilink
      English
      arrow-up
      2
      ·
      2 days ago

      If so, is it any good?

      I imagine if one writes janky spaghetti then it’s easy to think that LLMs will result in redundancy. My experience is that they’re like having an over-enthusiastic junior who doesn’t learn. Useful when one can’t be bothered to write something with very limited scope but is quickly out of their depth on anything involved.

      • rmrf@lemmy.ml
        link
        fedilink
        English
        arrow-up
        1
        ·
        2 days ago

        Yeah. AI 100% makes me more productive and by a good bit. I used it to write a section of code today to export all the information my program gathers about a system to a json file. Woulda taken me 20 minutes but chatgpt did it in seconds.

        It also, in the same snippet, introduced a breaking change I didn’t ask for in the original code. I only copied the json part; I just happened to notice the change in the code it wrong. It added a fork bomb lol

        • Flamekebab@piefed.social
          link
          fedilink
          English
          arrow-up
          3
          ·
          2 days ago

          That’s the thing though - I find that what it saves me in time writing it costs me in reviewing. I hate reviewing.

          • rmrf@lemmy.ml
            link
            fedilink
            English
            arrow-up
            1
            ·
            23 hours ago

            I’ve cut down on that by only giving it tasks that are reviewed automatically.

            “Spit this into a json”

            It gives me the code, I continue with my program knowing exactly what is where in the json and if I get a parsing error or something I know exactly where to look. TBH, though, for things like that I must have an error rate lower than 5%

            Ask it to configure a reverse proxy for a cors sensitive application, though, and I think I’d rather die