Conversation

Notices

1. Embed this notice
   iced depresso (icedquinn@blob.cat)'s status on Monday, 11-May-2026 16:23:46 JST iced depresso

   in reply to

   • Peter
   @peter people run on way worse than 95% all the time. though when i tinkered with using GLM for code it was more like 40-50% correct :cat_sad:
   • Embed this notice
     Peter (peter@thepit.social)'s status on Monday, 11-May-2026 16:23:47 JST Peter

     in reply to

     that "mostly" is doing a lot of work. since ChatGPT rolled out to the public in 2022, that has always been the problem with these stochastic systems. 95% right is not good enough for important tasks because then you have to verify every output, so there's no point.

   • Embed this notice
     Peter (peter@thepit.social)'s status on Monday, 11-May-2026 16:23:49 JST Peter

     in reply to

     anyway, this is **very** funny if you're into this kind of post. guy deployed an LLM-powered app as a wedding concierge. it didn't really work! for very predictable reasons. https://www.reddit.com/r/LLMDevs/comments/1t9jhgb/i_deployed_an_llm_agent_as_a_guest_concierge_for/

   • Embed this notice
     Peter (peter@thepit.social)'s status on Monday, 11-May-2026 16:23:51 JST Peter

     in reply to

     "sounds like a redditor, must be a bot"

   • Embed this notice
     Peter (peter@thepit.social)'s status on Monday, 11-May-2026 16:23:52 JST Peter

     my favorite thing about the LLM dev subreddits now is everyone accusing everyone else of being a bot