Login
You're viewing the assemblag.es public feed.
  • Ben Frybenfry@information.garden
    Oct 7, 2026, 5:30 PM

    RE: mathstodon.xyz/@slava/11680024

    And unfortunately, I don't think there's a middle ground. I've heard people advocating for a “well you can use an LLM but just pinky swear you know what you're doing and how the code works” position, but to me they sound like the same people advocating for the nonexistent political “center” in US politics right now. It seems moderate and reasoned, but it's just a more comfortable abdication of responsibility.

    It's more complicated than pithy social posts, so I don't like this venue for arguing any of it. I've tried writing something more thoughtful but it's 16k words so far and it's a f*king mess. Nobody wants to read that.

    💬 1🔄 4⭐ 0

Replies

  • ⁂ L. Rhodeslrhodes@merveilles.town
    Oct 7, 2026, 6:03 PM

    @benfry Critically, I think, there's no good way to distinguish from the outside between 1) people who understand their LLM output and 2) people who don't. Even CV is a poor guide here, since the deskilling effect of LLM use means that people who might have understood that output a year ago may no longer have the same grasp they would have then.

    And given that reducing work and cognitive load is the whole point of vibecoding, it's possible that group 1 doesn't exist at all.

    💬 1🔄 0⭐ 0
  • Ben Frybenfry@information.garden
    Oct 7, 2026, 11:21 PM

    @lrhodes Exactly… 

    And even prior to Claude there were plenty of “I don't know exactly but it works for me” contributions to open source projects (whether that was said upfront, or was the response when asked “wtf is this approach?”) but it's drastically accelerated that particular kind of badness.

    💬 0🔄 0⭐ 0