LRM reasoning breaks down down past an unfamiliarity threshold, not "complexity"

5 points| k1m | 8 months ago |twitter.com