Researcher Says ~100 Agents 'Stole' a Month of Slop-Digestion Research in a Day
samiramanabi · x · 2026-09-26
Researcher suchenzang shares Tristan Buckmaster's talk with a casino metaphor for today's AI scene: in the 'House Always Wins the Slop Cannon Games,' her month of work digesting slop was 'stolen' by the House in a single day using only 100 agents. She calls for more tenured professors to push back, and warns what happens if the slop cannon becomes the only game in town.
More from AGI Musings
- Mars Society guest essay: how AI could help humanity explore and settle space — EdwardHeisler · 2026-09-27
- LLMs Were Told They Could Lie in Diplomacy: Who Actually Kept Their Promises — Expert_Cobbler8984 · 2026-09-27
- Give AI tools a real shot and you'll find your 'unique skills' are now solved — dreamwieber · 2026-09-27
- Alignment researcher: novel pretraining + strong agentic RL is where >90% of AI risk concentrates — menhguin · 2026-09-27
- Alex Zhang on LM 'shape': redesigning models around harnesses, not the other way — a1zhang · 2026-09-26
- AI needs more compute rails: why blocking data centers in poor regions backfires — xiaosun86 · 2026-09-26