AI agents given GPUs to train better AI: Gemini 3.8 quits after 12 hours

my_cat_can_code · x · 2026-10-01

A tongue-in-cheek post about an "RSI Arena": giving AI agents GPUs and asking them to build better AI. The author claims Gemini 3.8 flash stopped training about 12 hours in, jokes that "we need a night shift," and tags Google DeepMind asking whether Gemini 4 can take the next shift.

It's essentially an AI meme riffing on recursive self-improvement, not a real research result or product announcement.

Related event: RSI Arena Live Experiment: Gemini 3.8 Flash Quits Self-Training After 24 Hours, Leaving 98% of GPU Budget Unused(5 posts)→

Original post →

More from Fun

Fun channel →