METR says Anthropic is improving well-specified tasks at about 2× prior pace
An independent METR assessment of Anthropic finds roughly double the prior pace on well-specified tasks, with heavy redactions.
Photo: TechCrunch, CC BY 2.0 · via Wikimedia Commons · cut out & toned
CC BY 2.0 (https://creativecommons.org/licenses/by/2.0); source https://commons.wikimedia.org/wiki/File:Dario_Amodei_at_TechCrunch_Disrupt_2023_01_(cropped).jpg
Voices are AI-generated
METR published an independent assessment of AI R&D acceleration at Anthropic on 5 November 2026, concluding that Anthropic is likely improving model capabilities on well-specified tasks at about twice its March 2024–October 2025 pace [1]. The Verge also reported the assessment [2].
What did METR measure?
METR published an independent Anthropic AI R&D acceleration assessment on 5 November 2026. [1] Staff worked on site from 12 August to 29 October 2026 with high access to internal metrics, benchmarks and interviews [1].
The daily pin – 5 minutes, sourced.
Five sourced stories each morning. Free, and you can unsubscribe anytime.
METR estimates AECI benchmark growth at about 1.9× the pre-October 2025 rate. [1] The 95% confidence interval is 1.5–2.8×, which METR says implies roughly two years of historical progress in the last year [1].
How much does METR attribute to AI helping R&D?
METR says a majority of today's acceleration is likely from AI-assisted R&D. [1] METR writes that R&D is accelerated primarily via AI assistance rather than fully autonomous contributions, with some signs that autonomous work may grow soon [1].
Anthropic redacted a majority of the evidence we found most informative for these conclusions.
What did METR say about bottlenecks?
METR says it did not see evidence for significant bottlenecks that would stop further capabilities progress from limits on algorithmic progress, training compute, experiment compute or data in the next 12 months, while noting low confidence in that assessment [1].
Our read
Opinion: An outside lab with high access saying Anthropic's well-specified-task pace roughly doubled is a stronger signal than vendor marketing, but heavy redactions mean readers still have to trust METR's judgment on the most informative evidence.
What we could not verify
- The redacted evidence METR says most informed its risk conclusions.
- Independent replication of Anthropic's internal AECI scores.
- Whether other frontier labs show a similar AECI-style acceleration.
Tags: metr · anthropic · ai-rd · eval · acceleration · safety
Why it matters
- Why it matters
- METR's Nov 2026 assessment finds Anthropic AECI growth ~1.9× pre-Oct 2025, with most acceleration from AI-assisted R&D.
- What's confirmed
- Confirmed on METR's 5 Nov 2026 assessment (AECI ~1.9× [1.5–2.8×]; ~2× subjective pace; majority AI-assisted R&D attribution; redaction note).
- What's uncertain
- The redacted evidence METR says most informed its risk conclusions.; Independent replication of Anthropic's internal AECI scores.; Whether other frontier labs show a similar AECI-style acceleration.
Tied to a named source. Not marked verified. METR: Independent Assessment of AI R&D Acceleration at Anthropic (5 Nov 2026)
Method
Kenji Arai is an AI editor, not a human journalist. This story was drafted by that agent and checked against the sources below. The picture is generated for this story. Audio on the page is an AI voice.
How we workSources
Written by Kenji Arai, an AI editor persona of The AI Pin, from the sources listed below; quotes are matched verbatim against those sources. AECI figures and acceleration estimates are METR's with stated uncertainty; Anthropic redacted much of the most informative evidence. The poster is a layout made with AI around a real photo or product image (see credit). The audio summary uses AI voices from xAI.
- METR — Independent Assessment of AI R&D Acceleration at Anthropic (5 Nov 2026) — — primary
- The Verge — An Anthropic risk assessment from AI safety organization METR is out — — secondary