🔍 Read the full analysis: AI Building AI: What METR Found And What It Couldn’t Share on ThorstenMeyerAI.com
Get smart everyday buys delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
METR’s November 5 assessment estimates that Anthropic’s progress on measured AI capabilities accelerated about 1.9 times after October 2025, with a 95% confidence interval of 1.5x to 2.8x. METR attributes perhaps 80% of the acceleration to AI use, but says Anthropic redacted most of the evidence it found most informative, limiting public scrutiny.
METR, a nonprofit that evaluates frontier AI systems, estimated on November 5 that Anthropic’s measured capability progress accelerated about 1.9 times after October 2025, compared with its earlier pace. The assessment, based on 11 weeks of access to internal company data, says AI assistance likely accounts for most of the increase—but METR also reports that Anthropic redacted a majority of the evidence it considered most informative.
METR’s central estimate compares the pace of Anthropic’s capability gains before and after a change point it places at October 7, 2025. Its measure, AECI, combines results across benchmarks into a single capability index. METR estimates the index rose by about 15 points per year before the change point and 29 points per year afterward. In METR’s framing, that is roughly two years of historical progress in one year.
The estimate has a 95% confidence interval of 1.5x to 2.8x. METR says it would not be surprised by an acceleration between 1.3x and 3x, and reports that alternative assumptions about task difficulty move the result from 1.1x to 2.9x. A comparison of frontier models from all developers, rather than Anthropic models alone, produces a smaller estimate: 1.4x, from 17 to 24 index points per year.
METR estimates that perhaps 80% of the acceleration comes from AI use and says it is confident the change was not caused by faster growth in training compute. The report draws on unreleased-model benchmark scores, training metrics, staff surveys and about 15 interviews. METR says roughly 2.5 full-time-equivalent staff worked on the assessment using Anthropic devices; the company supplied laptops and free tokens, and retained the right to redact non-public information. METR says it received no payment.
Infographic 01 · How fast is AI speeding up AI?
About twice the old pace, most likely because of AI
An outside evaluator's estimate from internal Anthropic data, plus Anthropic's own measurements. Wide uncertainty, much of the evidence redacted.
Acceleration of Anthropic’s benchmark progress (AECI) since October 2025, 95% confidence interval 1.5x to 2.8x. METR’s overall judgment: around twice the old rate; it “would not be surprised” by 1.3x or 3x.
Capability points per year
AECI trend of Anthropic models, break point 7 Oct 2025
All labs before October 2025: +17 a year, so 1.4x. Different task-difficulty assumptions: 1.1x to 2.9x.
Sources: METR assessment, 5 Nov 2026 · METR risk report review · Anthropic Institute measurements.
Powered by Thorsten Meyer AI
Why Faster AI Research Matters
If AI tools are helping researchers and engineers develop stronger AI systems faster, the development could affect how quickly model capabilities advance and how much time companies and outside evaluators have to identify risks. METR’s estimate is evidence about one company’s measured progress, not proof that AI systems can independently build their own successors or that the same pace applies across the industry.
The finding is especially relevant to Anthropic’s Responsible Scaling Policy, which includes a threshold for automated research and development. The policy treats a doubling of progress pace as a potential trigger when the increase is plausibly substantially attributable to automated research or engineering. METR says the threshold may plausibly have been crossed by July 15, but that is its assessment, not a confirmed determination by Anthropic. Anthropic’s November 2 SEC filing said the company “may imminently meet” its definition of dramatically accelerated progress.
The redactions also shape the significance of the report. The public can see METR’s estimates and methods in outline, but cannot inspect much of the underlying evidence the evaluators considered most useful. That makes it harder for independent readers to judge the strength of the estimate or assess the explanations offered for it.
Popular right now
De'Longhi La Specialista Touch Semi-Automatic Espresso Machine with Grinder
Dyson Ball Animal 3 Upright Vacuum – Corded, 290 AW, De‑tangling Motorbar Cleaner Head, 3 Suction Modes
Bose QuietComfort Headphones (2nd Gen), Wireless Over-Ear Noise-Cancelling Headphones, Black
As an affiliate, we earn on qualifying purchases.
Anthropic’s Acceleration Threshold
Anthropic’s Responsible Scaling Policy, version 3.4, took effect on July 8, 2026. It identifies two possible triggers for automated research and development: a model able to replace the company’s research staff at competitive cost, or progress doubling under a definition that connects the increase plausibly to automated research or engineering. The policy describes the doubling threshold as achieving in one year the progress otherwise expected over two years.
Anthropic has described recursive self-improvement as an AI system fully autonomously designing and developing its successor. The source material distinguishes that prospect from the current situation, in which AI assists human researchers and engineers who still choose goals. Anthropic has said, “We are not there yet, and recursive self-improvement is not inevitable.” METR’s report measures acceleration in research and development; it does not establish that full autonomy has been reached.
METR’s work differs from a typical outside evaluation of a publicly released model. Its staff examined internal data over 11 weeks, from August 12 to October 29, 2026, while working on Anthropic devices. The assessment included company data and interviews, but Anthropic’s right to redact non-public material limits how much of that record is available for public review.
“A 2x acceleration means that 2 years of historical progress is made in 1 year.”
— METR
Evidence Behind the Estimate
The estimate is not a direct measure of every part of AI development. AECI combines benchmark scores, and METR’s results vary with assumptions about task difficulty. The all-developer comparison is lower than the Anthropic-only estimate. METR offers two possible explanations: Anthropic may have been behind the frontier in 2025, which could make its subsequent acceleration look larger, or earlier models may have been less tuned to benchmarks, which could make the acceleration look smaller.
Most importantly, METR says Anthropic redacted a majority of the evidence it found most informative. The source material also says a planned check of how Anthropic divides compute between safety work and capability development was dropped. The public account therefore does not allow readers to independently examine much of the underlying evidence or that proposed comparison.
It remains unclear whether the company considers its policy threshold crossed, how it reached that conclusion, or how much the reported acceleration reflects changes that would carry over to other labs. METR’s report is an outside assessment, but the available information does not resolve those questions.
Anthropic’s Policy Decision Ahead
The next key development is whether Anthropic determines that it has met the accelerated-progress trigger in its Responsible Scaling Policy and what safeguards, if any, that determination requires. The November 2 SEC filing said the company may imminently meet its definition; it did not announce that the threshold had been reached. No specific decision date is given in the source material.
Further clarity would depend on what Anthropic makes public about its policy assessment and whether additional evidence or independent evaluation becomes available. Until then, METR’s figures should be read as estimates based on its analysis of company-provided internal data, with substantial underlying evidence withheld from public view.
Key Questions
What did METR estimate?
METR estimated that Anthropic’s measured capability progress accelerated about 1.9x after October 2025 compared with its earlier pace. Its 95% confidence interval is 1.5x to 2.8x.
Does the report say AI can build its own successor autonomously?
No. The report concerns AI assisting human research and engineering. The supplied material says humans still choose goals and distinguishes this from fully autonomous recursive self-improvement.
How much of the acceleration did METR attribute to AI?
METR said perhaps 80% was due to AI use. That is an estimate, not a directly measured share. METR also said it was confident the acceleration did not result from faster growth in training compute.
Why can’t the public fully check METR’s findings?
METR says Anthropic redacted a majority of the evidence it found most informative. Anthropic supplied internal data access and retained the right to redact non-public information, limiting what readers can independently inspect.
Has Anthropic confirmed it crossed its acceleration threshold?
The material does not report a confirmation. An SEC filing on November 2 said Anthropic “may imminently meet” its definition of dramatically accelerated progress; METR separately judged that the threshold may plausibly have been crossed by July 15.
Source: ThorstenMeyerAI.com
