• irmadlad@lemmy.world
      link
      fedilink
      English
      arrow-up
      15
      ·
      24 小时前

      I’m not so much worried about ‘vibe coding’ as long as the dev actually knows the validity of the code presented in the LLM. At that point, the LLM becomes the assistant, not the dev itself. However, if I were to speculate, this dev team didn’t, got called on it, didn’t know how to respond or validate the code, so they closed up shop.

      • chicken@lemmy.dbzer0.com
        link
        fedilink
        English
        arrow-up
        11
        ·
        15 小时前

        The term ‘vibe coding’ I think was originally about generating and using code without understanding it

        • irmadlad@lemmy.world
          link
          fedilink
          English
          arrow-up
          3
          ·
          8 小时前

          Well, I have used LLM to do code for in house type stuff. Very simple. In that scenario, it’s fairly good. I’ve found that LLM are good at compose files for instance. But that’s much different than producing a piece of software for thousands of people to use with confidence. Especially when dealing with anything 'arr and the mitigation that takes place to operate that in a secure, private, and anonymous manner.

        • orgrinrt@lemmy.world
          link
          fedilink
          English
          arrow-up
          1
          ·
          edit-2
          14 小时前

          Yeah doesn’t sound like vibe coding if you actually go through the necessary dances anyway, i.e double-check the produced code and validate it and actually understand it and the domain.

          Edit: But almost nobody does. Because then you’d rather just write it yourself and save time, money and energy…