A blog about making culture. Since 1999.
  • 6️⃣9️⃣4️⃣2️⃣0️⃣@lemmy.worldEnglish
    5·
    1 hour ago

    This is a good read and makes a lot of great points. I think everyone in tech needs to understand the arguments here. The biggest thing for me is that LLMs are incredibly useful tools, but not in the way they are advertised. They are great for learning how existing code works, but shit at writing anything novel or innovative. From the article:

    The past is a prison when you’re inventing the future.

    In my opinion, if you’re using LLMs to do anything but help you learn from the past, you’re doing it wrong. LLMs cannot move you forward, and I think that may be the point.

    • rozodru@pie.andmc.caEnglish
      3·
      25 minutes ago

      I’ve always said LLMs are fantastic rubber ducks. But taking said rubber duck and telling it to build something end to end or hitting tab without verifying anything is going to lead you to a world of hurt. the person doing it won’t know the world of hurt is coming because they simply don’t know any better. the company won’t know the world of hurt is being built around them because like the vibe coder they dont’ know any better.

      then suddenly it’s finished, pushed to production, and there’s your world of hurt.

      but hey I get paid to fix said world for them so keep on trucking I guess. at this rate I’ll be retiring by the end of the next year.