the archive 16 drops

← back to the feed

reading path

month

filtering by

A plausible model explanation is not a root cause

Quantization damage hides in the flips, not the average

Perfect accuracy cannot reveal the algorithm

ARC-AGI-3's perfect score belongs to the harness

The agent turf war happened in the lab, not the wild

Guardrail benchmarks are graded before the attacker moves

Model leaderboards can't see the harness

Temperature zero is not a determinism guarantee

An isolated sandbox needs evidence for its exceptions

The AI safety index grades disclosure, not safety

Quantized retrieval needs slice-level evals

Agent swarms need aggregation evals

page 1 of 2