William Merrill

AI

StateBench's team says GPT-6 Sol and Claude Opus 5.5 scored 13% accuracy or lower on its hardest, longest problems.

3801234567890123456789

OP

WM

KW

5 Sources, 15h ago, updated 15h ago

AI

A researcher says frameworks for deriving batch-size scaling rules are less general than they claim.

5801234567890123456789

CL

WC

SM

6 Sources, 19d ago, updated 19d ago

AI

A user points readers interested in the form/meaning question around 'stochastic parrots' to a linked work they call a classic.

1501234567890123456789

WM

1 Source, 36d ago, first seen 36d ago

AI

Jay Hack coins term amid debate on autoregressive LLMs for code with tools.

185012345678901234567890123456789

JS

SB

AC

17 Sources, 92d ago, first seen 92d ago