Migrating to the new meshcore.io site
  • atzanteol@sh.itjust.worksEnglish
    0·
    18 days ago

    We ran a poll recently, and asked in the MeshCore Discord about AI and trust, and these are the results

    I don’t quite understand the “trust” issue with AI code. Well - I kinda do but it’s overblown.

    Trust is not about having people vs. robots. It’s about having good code review and transparency. It’s a process thing not a “who wrote this” thing.

    If you’re running an open source project you need to be putting those guardrails in place anyway, unless you want to end up with an XZ Utils-like backdoor in your project (which required no AI to do).

    • hendrik@palaver.p3x.deEnglish
      0·
      18 days ago

      I have a hunch how that might just not happen. If it’s a singular developer pumping out all kinds of features, websites, software… There’s a good chance there’s little oversight nor any proper guardrails in place.

      So hypothetically, sure?! In practice I also see how some things sometimes go hand in hand. And you better check who writes and maintains your software, if you’re planing to use it long term. At least that’s what I’ve seen. Once a project is mostly vibe-coded, there’s automatically a high chance there’s weird things going on. Or I’ll write something to the issue tracker and there’s nobody there. Or maintenance stops after a week because they vibe-code the next thing… So I’m a bit wary. I guess because in reality lots of people don’t use these tools responsibly.