Why it matters
METR agrees with Anthropic's bottom-line assessment that catastrophic risk from Claude Opus 4.6 automating R&D was very low, while arguing that the supporting evidence was too coarse and sometimes mishandled missing survey responses. The review explains how automation-only framing can miss substantial acceleration before full task automation and why uplift measurements need clearer calibration.
My takeaway: For consequential capability claims, publish granular measurements, sample and missing-data treatment, uncertainty, and calibrated thresholds. Track partial task substitution and workflow acceleration as leading indicators instead of waiting for end-to-end automation, and invite an external reviewer to challenge the argument as well as the benchmark score.