Researchers argue the Bitter Lesson was misread: more compute makes algorithm choice more critical

DimitrisPapail · x · 2026-11-08

MillionInt argues a whole generation of scientists got reward-hacked by misreading the Bitter Lesson and the brief era when scaling transformers did most of the legwork.
Core claim: the more compute you pour into training, the more important your algorithm is — some algorithms scale, others don't, and much energy went into dead ends while entire scaling dimensions remain unexplored.
@arohan adds this will upset young scientists who stopped developing their own insights believing scaling alone was sufficient, and felt defeated by the "Bitter Lesson."

More from AGI Musings