Thinking Machines News

Dwarkesh Patel hosts AI researchers on recursive self-improvement

Episode features John Schulman, Beren Millidge and Charlie O'Neill discussing what's happening at the frontier and what comes next. The first segment, running to 18:39, steelmans the case against recursive self-improvement.

RLVR with task expertise beats human on text-to-SQL

A model trained with RLVR on Tinker achieved state-of-the-art on BIRD, scoring above the human mark of 92.96%. The work, a guest post by UIUC and Bridgewater researchers, folds expert judgment into every part of RLVR.

AI safety research grants up to $50k in Tinker API credits