AI Now Drives Most of Its Own Rapid Progress
METR finds AI accounts for about 80% of Anthropic's recent speed-up, a research-stage finding with no release timeline announced.
Updated on Nov. 5, 2026 in Artificial Intelligence
Live Poll
Do you think AI should be strictly regulated now, before it advances further?
METR, a research group, found that Anthropic's AI models have improved 1.9 times faster since October 2025 than previously. The findings, published Thursday, are based on an 11-week assessment and are research-stage findings rather than a shipped capability.
Why it matters
METR estimated AI accounts for perhaps 80% of the current acceleration, pointing to a feedback loop in which AI plays a growing role in developing more powerful versions of itself. METR ruled out a faster scaling of training compute as the cause, which places the speed-up in the development process itself.
METR's assessment reports a 1.9x speed-up in Anthropic's capability improvement since October 2025, with about 80% of the acceleration attributed to AI itself. The figures come from confidential research and unreleased model data, so the specific benchmarks and the pre-October baseline are not public in the source.
The players
METR
An independent AI evaluation research group that assesses frontier models' capabilities and the pace of AI development.
Anthropic
An AI company developing the Claude family of large language models, whose model improvement rate METR assessed.
The details
The acceleration appears to come from AI taking a larger role in building its successors, a feedback loop in which models contribute to the research and engineering that produces the next generation. METR ruled out faster growth in training compute as the driver, so the gain is attributed to the development process rather than to more hardware being applied. AI is not yet capable of fully autonomous, highly significant contributions to the most important stages of model training, so human researchers remain central to that work.
Timeline
October 2025: the period of accelerated Anthropic model improvement begins.
Thursday, November 5, 2026: METR publishes its findings.
Next year: METR finds no convincing evidence of major bottlenecks that could slow progress.
The Tech Race
METR's estimate that AI drives roughly 80% of the current acceleration places Anthropic's progress at the center of a race in which models increasingly shape their successors. The finding suggests the competitive pace among frontier labs may now depend on how much each lab's models can contribute to research.
If the acceleration holds, the time between successive frontier model releases could shorten, bringing more capable AI tools to developers and users sooner. The findings are research-stage and no release timeline or pricing change has been announced.
The takeaway
Watch whether METR's projection of no major bottlenecks over the next year holds in upcoming Anthropic model releases. A follow-up METR assessment or a published benchmark comparison would confirm or reject the 1.9x trajectory.
What happens next
METR's assessment points to no convincing evidence of major bottlenecks over the next year, a projection that later model releases and METR follow-up reports can test.
Further reading
For broader context on how AI systems are reshaping model development, see the Artificial Intelligence section.
Source note: This article includes information reported by Reuters.
Live Poll
Do you think AI should be strictly regulated now, before it advances further?
