Summary
This video breaks down METR’s newly published report, “Independent Assessment of AI R&D Acceleration at Anthropic in October 2026,” released on November 5. METR, the non-profit known for its charts tracking how long a task AI agents can complete autonomously, spent from August 12 to October 29 working inside Anthropic’s systems. The team reviewed internal benchmark scores, training data, compute figures, forecasts, and staff surveys, and interviewed around 15 researchers.
The report’s headline findings are that Anthropic is likely improving its models at roughly twice the rate seen between March 2024 and October 2025, that around 80% of that speed-up is probably attributable to AI itself, and that METR found no convincing bottlenecks likely to bite within the next 12 months. The video also covers the unreleased internal models METR calls Model 2, Model 3, and Model 4. Model 4 finished training in September to October and reportedly scored above 50% on all 15 benchmarks tested, with some results near 90%.
The presenter discusses the concerns raised by researchers, including the heavy redaction of the evidence METR considered most informative, comments from Dario Amodei about earlier speed-ups of 15 to 20%, and whether Anthropic may have crossed the 2x acceleration threshold in its Responsible Scaling Policy. Community speculation about a future Claude release is flagged as unconfirmed.
📺 Source: TheAIGRID · Published November 06, 2026
🏷️ Format: News Analysis
