Scores by kind of claim
Each kind of claim is scored on its own, out of 100: the share that held up, with the range it likely falls in. Fewer resolved claims give a wider range.
-
Factual claims
Statements about the present or the past, checked against the record: did they hold up?
76 69 82 -
Predictions
Statements about events outside the person's control, checked once they fell due: did they come true?
56 42 70 -
Promises
Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?
Insufficient record: 4 promises resolved, 5 needed for a score.
All types combined: 72 (likely 66 to 78, high confidence), from 93 claims: 76 factual claims, 17 predictions. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.
Dario Amodei's claims
28 claims
“The things we emphasized were different.”
Dario Amodei, Anthropic CEO, interviewed 28 February 2025 Hard Fork - Anthropic's C.E.O. Dario Amodei on Surviving the A.I. Endgame Transcript (Podscripts (Hard Fork))
Anthropic's approach to reasoning models in Sonnet 3.7 emphasized different things from competitors' reasoning models.
Resolved true The record shows it happened or was so.
Said for an organisation they lead or work for; it counts on both records.
What would count as true or false
Judge as of 2025-02-28. True: Anthropic's public materials for Claude 3.7 Sonnet (the launch announcement and system card) show emphases that clearly differ from competing reasoning models such as OpenAI o1/o3-mini and DeepSeek R1. Examples include a hybrid single model with a toggleable extended-thinking mode, a visible thought process, a user-controllable thinking budget, and a focus on real-world or coding tasks rather than maths or competition benchmarks. Credible third-party coverage should also recognise at least one such difference. Mostly true: the differences are real but small or only partly distinctive, for example only one minor feature differs. Mixed: the claimed emphases are mostly shared with competitors, with only modest differences. Mostly false: the emphases are essentially the same as competitors' and the differences are cosmetic. False: the evidence shows Anthropic's approach matched competitors' emphases in substance.
Evidence
- Claude 3.7 Sonnet and Claude Code Anthropic · 24 February 2025 Own record
“Third, in developing our reasoning models, we’ve optimized somewhat less for math and computer science competition problems, and instead shifted focus towards real-world tasks that better reflect how businesses actually use LLMs.”
- Claude 3.7 Sonnet is here—here's what we know BdTechTalks (Substack) · 25 February 2025
“While other AI labs separate their general-purpose LLMs from their LRMs, Anthropic has decided to ship both into a single model to provide a “seamless experience for users.””
- Claude 3.7 Sonnet is here—here's what we know BdTechTalks (Substack) · 25 February 2025
“When using the model, users can turn on the “extended thinking mode,” which causes Claude 3.7 to generate a chain-of-thought (CoT) sequence before giving the final answer, which is equivalent to what LRMs such as o1 and R1 do.”
- Claude 3.7 Sonnet is here—here's what we know BdTechTalks (Substack) · 25 February 2025
“This is in contrast to Grok 3 and o3, which show a summarized version of the CoT to prevent competitors from copying their models.”
Reasoning
Anthropic's launch post says its reasoning work focused less on maths and competition problems and more on real-world business tasks (E1). Independent coverage from the same week says Anthropic put general and reasoning abilities into one hybrid model, while other labs keep them separate (E2). The same coverage says Claude shows its raw reasoning, while Grok 3 and o3 show only summaries (E4). E3 notes that the underlying chain-of-thought mechanism is equivalent to o1 and R1. However, the criteria name the hybrid design, the visible thinking and the real-world focus as distinguishing emphases. Anthropic's launch post documents the shift toward real-world tasks rather than competition problems [E1], and independent coverage documents the single-model design [E2] and contrasts Claude 3.7 Sonnet with Grok 3 and o3, which show a summarized version of their reasoning [E4].
Cite this claim
Dario Amodei, “Anthropic's approach to reasoning models in Sonnet 3.7 emphasized different things from competitors' reasoning models.” (said 28 February 2025). Verdict: resolved true. Reliability API, claim c_eefeb1b95e38, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_eefeb1b95e38
Claims without a verdict
- Unresolvable 157
- Not yet due 40
- Own records only 43
- Not researched 134
- Not checkable 864
Is a verdict here wrong, or do you speak for Dario Amodei? Reply on the claim's own page, with the form at its foot (how replies work). Verdicts are made automatically; how verdicts and scores are made.