METR's 1.9x AI R&D speedup at Anthropic, and why it ranges from 1.1x to 2.9x
Comments on “METR's 1.9x AI R&D speedup at Anthropic, and why it ranges from 1.1x to 2.9x”
Claim-checked. Scored by all three judges, and its checkable claims were compared with the sources it cites. How far stories are checked
METR, an independent evaluator, published an 11-week assessment concluding that capability progress at Anthropic has run about 1.9x faster since October 2025 (95% interval 1.5x to 2.8x), measured with an internal version of Epoch's capabilities index. The figure depends on a modeling assumption: alternative link functions give 1.1x to 2.9x, and the data fits most about equally well. METR's roughly 80% attribution to AI is a judgment, and Anthropic redacted much of the key evidence, so outside checks are limited.
Why it ranked
No. 12 of 107 with a score of 0.67. The panel rated it highest on new information (0.93) and reader interest (0.82), lowest on practical relevance (0.54). Placed No. 12, No. 19 by score alone: the fourth AI & machine learning story of the day (−0.020). The placement terms move stories; they are not part of the score.
- AI judgeJudge 2claude-haiku-5-5
- Major AI policy-relevant evaluation with real methodology detail and strong caveats; significant to AI builders, heavy redaction limits verifiability, and some sections are speculative judgment rather than measurement.
- AI judgeJudge 3gpt-5.6-luna
- Important, timely AI capability-growth assessment with substantial methodological detail and caveats. Its relevance is broad, but evidence is partly inaccessible, redacted, judgment-based, and not independently reproducible.
Each judge wrote its line after answering the rubric, with the site and the author still hidden. Every answer behind this is in the table under Score.
Score
0.67total under rubric v2.3
- Significance0.73
- Reader interest0.82
- Audience breadth0.66
- Practical relevance0.54
- Substantiated0.67
- Technical depth0.73
- New information0.93
- Counts against: Promotional0.05
- Counts against: Overclaim0.29
Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).
| Question | Judge 1 | Judge 2 | Judge 3 | Mean | Spread | Weight | Contribution (weight times mean) |
|---|---|---|---|---|---|---|---|
| SignificanceHow much does this change what people in tech know or can do? | 0.63 | 0.72 | 0.84 | 0.73 | 0.21 | +0.24 | +0.175 |
| Reader interestWould a well-informed senior engineer or tech founder want to read this today, and likely pass it on? | 0.67 | 0.85 | 0.93 | 0.82 | 0.26 | +0.26 | +0.212 |
| Audience breadthHow many people who build technology does it concern? | 0.48 | 0.70 | 0.80 | 0.66 | 0.32 | +0.13 | +0.086 |
| Practical relevanceHow useful is it to someone who builds software or tech products? | 0.43 | 0.50 | 0.68 | 0.54 | 0.25 | +0.22 | +0.118 |
| SubstantiatedHow well is it backed? Consider whether it is the primary source, and whether its main claims come with data, code, a demo, a reproducible method or citations. | 0.62 | 0.68 | 0.72 | 0.67 | 0.10 | +0.05 | +0.034 |
| Technical depthHow technically substantive is it? | 0.79 | 0.72 | 0.70 | 0.73 | 0.09 | +0.05 | +0.037 |
| New informationDoes it contain information absent from the stories covered in the last 7 days (list provided)? | 0.85 | 0.95 | 0.98 | 0.93 | 0.13 | +0.05 | +0.046 |
| PromotionalIs it mainly marketing with little substance (a press release, ad or sponsored post with no real information)? A launch post that carries real news is not promotional. | 0.08 | 0.03 | 0.03 | 0.05 | 0.05 | −0.19 | −0.009 |
| OverclaimDo the headline or conclusions claim more than the body supports? | 0.43 | 0.25 | 0.18 | 0.29 | 0.25 | −0.11 | −0.032 |
| Sum of contributions | +0.668 | ||||||
| Claim-check penalty | 0.000 | ||||||
| Total, as published | 0.668 | ||||||
Spread is the highest minus the lowest judge answer. marks a spread above 0.40: the judges disagree, and the story is marked contested.
The badge for this ranking, for the author's page or anyone's: one line of HTML, and it links back here.
<a href="https://rubric.hrih.si/editions/2026-11-07/12/"><img src="https://rubric.hrih.si/badge/2026-11-07/12.svg" alt="Ranked #12 on Rubric, 7 Nov 2026" width="238" height="28"></a>Claim check
Claim check: 6 matches source · 2 couldn’t check. Each claim is compared with the source the article itself cites; a flag is published only when both checkers agree.
Matches sourceNumber
METR found Anthropic's capability trend has grown about 1.9x faster since October 2025, from about 15 to about 29 index points a year, with a 95% interval of 1.5x to 2.8x.
The cited source says the same thing.
The article says “Anthropic's capability trend has grown about 1.9x faster since October 2025, from about 15 to about 29 index points a year. The 95% interval is 1.5x to 2.8x”
The source says “have accelerated 1.9x [95% confidence interval: 1.5–2.8x] since October 2025”
Source cited by the article: metr.org/blog/2026-11-05-anthropic-ai-rd-acceleration-assessment
Matches sourceDate
METR ran an 11-week review from August 12 to October 29, 2026, with about 2.5 full-time staff and data Anthropic chose to share.
The cited source says the same thing.
The article says “METR ran an 11-week review, August 12 to October 29, 2026, with about 2.5 full-time staff and data Anthropic chose to share”
The source says “From August 12 to October 29, 2026, METR staff performed a high-access evaluation at Anthropic”
Source cited by the article: metr.org/blog/2026-11-05-anthropic-ai-rd-acceleration-assessment
Matches sourceNumber
Swapping the link function changes the estimated Model 4 acceleration from 1.1x (complementary log-log) to 2.9x (log-log), with most curves fitting the data about equally well at around 86 to 87% R squared.
The cited source says the same thing.
The article says “METR reports that most of these curves fit the data about equally well out of sample, around 86 to 87% R squared, so the data cannot pick a winner”
The source says “Complementary log-log (heavy left tailed) | Benchmarks have a larger tail of only easy problems (for example, if there are many easily guessable tasks). | 1.1x [0.8–1.5] | 87.1%”
Source cited by the article: metr.org/blog/2026-11-05-anthropic-ai-rd-acceleration-assessment
Couldn’t checkBenchmark result
The slope-change model had an error of 1.9 points against 3.9 for the one-time-jump model and won in 81% of bootstrap draws.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “the slope-change model's error was 1.9 points against 3.9, and it won in 81% of bootstrap draws”
The source says No passage could be compared.
Source cited by the article: metr.org/blog/2026-11-05-anthropic-ai-rd-acceleration-assessment
Matches sourceNumber
Across all frontier developers the same method gives about 1.4x, and Anthropic's Haiku 5.5 system card reported 1.7x, with the gap mostly due to METR including internal-only models.
The cited source says the same thing.
The article says “across all frontier developers, not just Anthropic, the same method gives about 1.4x, and that Anthropic's own Haiku 5.5 system card reported 1.7x”
The source says “The most important difference between the latest Anthropic model release, the Haiku system card (1.7x acceleration), and our estimate (1.9x acceleration) is the inclusion of Anthropic’s recent internal-only models”
Source cited by the article: metr.org/blog/2026-11-05-anthropic-ai-rd-acceleration-assessment
Couldn’t checkNumber
Anthropic staff Slack polls on speedup from the latest model rose from about 1.2x for Claude 3.7 Sonnet in January 2025 to about 12x for Model 4 in October 2026.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “Those polls rose from about 1.2x for Claude 3.7 Sonnet in January 2025 to about 12x for Model 4 in October 2026”
The source says No passage could be compared.
Source cited by the article: metr.org/blog/2026-11-05-anthropic-ai-rd-acceleration-assessment
Matches sourceNumber
Most METR staff on the project shared an estimate that about 80% of the October 2026 speedup comes from AI, with large uncertainty.
The cited source says the same thing.
The article says “METR says most staff on the project shared an estimate of about 80% for October 2026, with large uncertainty”
The source says “We guess that a majority (perhaps ≈80%) of the current (October 2026) acceleration is due to AI”
Source cited by the article: metr.org/blog/2026-11-05-anthropic-ai-rd-acceleration-assessment
Matches sourceCausal conclusion
Anthropic had unilateral redaction rights and redacted most of the evidence METR found most informative, and a planned compute-split check was cancelled midway at Anthropic's request.
The cited source says the same thing.
The article says “Anthropic had unilateral redaction rights and redacted most of the evidence METR found most informative”
The source says “Anthropic redacted a majority of the evidence we found most informative for these conclusions”
Source cited by the article: metr.org/blog/2026-11-05-anthropic-ai-rd-acceleration-assessment
Also covered by
Elsewhere
Elsewhere: HN 2 pts, not on front page
Recorded for context only, checked 7 Nov 2026, 05:47 UTC. It played no part in the ranking.