• GenderNeutralBro@lemmy.sdf.org
      link
      fedilink
      English
      arrow-up
      3
      ·
      8 hours ago

      They say this was done with Codex and OpenCode.

      I admit I’m behind the time with AI coding agents. Is it actually possible to do that effectively in such a short time? Or is this going to be riddled with sloppy bugs?

      • WhiteRice@lemmy.ml
        link
        fedilink
        English
        arrow-up
        2
        ·
        edit-2
        6 hours ago

        I would say recompiling code with discrete units tests is probably a good use case. It allows for a tight feedback loop with the agent.

        You will probably get the mean deliverable. Not industrial unless you think to bake in the requirements.

    • Canuck@sh.itjust.worksOP
      link
      fedilink
      English
      arrow-up
      2
      ·
      edit-2
      9 hours ago

      Yes, I posted that one and also another one in my original description if you’re interested; it has a full breakdown of the original source code