AI is dramatically accelerating AI research right now

Anthropic has likely already crossed its Responsible Scaling Policy threshold for “dramatically accelerated progress due to AI,” according to a METR deep-access report finding that capabilities progress at the company ran roughly twice as fast between October 2025 and October 2026 as in the prior twelve months, as measured by an internal version of the Epoch Capabilities Index. METR attributes most of the sustained acceleration to AI automating AI R&D rather than to compute or other inputs, though it redacted what it called the majority of its most informative evidence. Because overall capabilities combine compute multipliers from algorithmic work with training compute, a twofold acceleration in total progress implies human researchers are being sped up by roughly 2.5–3x, the report’s modeling suggests. What to watch next is the unredacted supporting evidence and whether Anthropic formally declares the threshold met under its RSP.