• @[email protected]
    link
    fedilink
    44 hours ago

    You all realize pretty soon no human is going to update an existing code-base?

    Who wants to spend their time understanding 10 year old legacy code? They’ll just feed it into an AI and tell it to add or fix a feature, then generate tests, and file a PR.

    If it ends up having an airplane do a loop on take-off or sending your paycheck to Antarctica 🤷🏻‍♂️

  • @[email protected]
    link
    fedilink
    14 hours ago

    How are people making apps with LLMs? I can’t even make the fuckin thing to write a me a short story

  • @[email protected]
    link
    fedilink
    9
    edit-2
    13 hours ago

    Yes, but bad example pics are used bcs the right panel example is also built by humans and is very efficient for the intent.

    So both pics are great examples for the left panel - the AI would produce a clusterfuck of convoluted multilevel railways that would never see irl implementations in the railway business.

  • @[email protected]
    link
    fedilink
    2618 hours ago

    Whenever i ser apps built with ai assitance i just think:“wait until an exploit pops up and no one knows how to fix it. Since it was hallucinated by ai from 2009 code”

    • @[email protected]
      link
      fedilink
      English
      2517 hours ago

      sometimes it feels like they’re doing it on purpose.

      yesterday, i asked for it to create a basic rest api authentication function/method and it gave me something back that explicitly disabled ssl and when i asked why; it simply said “oops”.

        • @[email protected]
          link
          fedilink
          10
          edit-2
          14 hours ago

          i think its dumb enough to put exploits in because someone on stackexchange thought it made it more difficult 9 years ago.