AI News Page

← 2026-10-17 edition

THE SHORT VERSION

ARC PRIZE: THE STANDARD HARNESS HALVED THE SCORE — 'A SCORE WITHOUT ITS HARNESS IS A FOOTNOTE WITH THE NOTE TORN OFF'

GOOGLE'S 97.5% HAD CHEAT CODES BUILT IN — THE SAME MODEL SCORED 49.8% ON THE REAL TEST

A tiny anxious safety inspector in a lab coat holds up a SAFETY REVIEW report card with a big red circled F, while a giant suited hand wields a red pen poised to change the grade — satirizing Reuters' report that Anthropic and OpenAI sought veto power over what independent AI evaluators can publish.
Machine Age Cartoon · Zimbo