METR: AI R&D Capabilities Rising Fast, Increasingly Hard to Measure

METR's November 2026 update finds that AI R&D-related capabilities up to Opus 5.5 and GPT-6 Astra are improving rapidly but exceeding its time-horizon benchmark's measurable range, with multi-agent swarms overestimating accuracy by up to 50x.

2026-11-21 ~ 2026-11-21 · 4 related posts

1 near-duplicate retellings: najoungkim