ClaimsPeople

Scott Alexander

Scott Alexander is a writer known for his blogs Astral Codex Ten and Slate Star Codex and for fiction including Unsong.

775 claims collected, 234 researched, 81 resolved. Report of 4 November 2026, method 0.1.0+8bea457b.

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?

    80 72 88

    Medium confidence, 35 factual claims resolved. From 10 documents, 2014 to 2026; 15 documents sampled from 35 candidates found.

    Not scored: 46 unresolvable, 9 checkable only against their own records.

  • Predictions

    Statements about events outside the person's control, checked once they fell due: did they come true?

    70 62 78

    Medium confidence, 44 predictions resolved. From 7 documents, 2016 to 2025; 15 documents sampled from 35 candidates found.

    Not scored: 67 unresolvable, 10 checkable only against their own records. 19 more fall due later.

    Does not yet allow for how likely each prediction was when made.

  • Promises

    Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?

    Insufficient record: 2 promises resolved, 5 needed for a score.

All types combined: 75 (likely 69 to 80, high confidence), from 79 claims: 35 factual claims, 44 predictions. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.

Scott Alexander's claims

26 claims

26 claims, sorted by newest said. Choose a claim to read it in full.
VerdictKindSaidClaimTopic
Mostly true Factual 19 Oct 2026 MNX, a project of Manifold Markets co-founder Stephen Grugett, is a real-money prediction market on AI. Factual · 19 Oct 2026 Economics and markets
Mostly true Factual 5 Oct 2026 Reader Nikita Sokolsky has finished obtaining extra Fatima documents from Portugal and translating them into English. Factual · 5 Oct 2026 Their own work or organisation
Mostly true Factual 5 Oct 2026 Physicians For AI Safety exists, wants to talk to physicians interested in AI safety, and has an open letter such people can sign. Factual · 5 Oct 2026 Field building and careers
Mostly true Factual 3 Apr 2025 Through history, people have been reluctant to admit an AI is truly intelligent. Factual · 3 Apr 2025 Philosophy, ethics and society
Mostly true Prediction 3 Apr 2025 The authors of the scenario think coding is what will start the intelligence explosion. Prediction · 3 Apr 2025 AI capabilities
Mostly true Prediction 3 Apr 2025 In the AI 2027 scenario, as AI labs become more capable they tell the government, seeking government contracts and support. Prediction · 3 Apr 2025 AI policy and governance
Mostly true Prediction 3 Apr 2025 In the AI 2027 scenario, the government sees via cyber warfare that AIs are as capable as the best human hackers but deployable at huge scale. Prediction · 3 Apr 2025 AI policy and governance
Mostly true Factual 10 Dec 2024 Most of the benefits of prison come from incapacitation rather than deterrence. Factual · 10 Dec 2024 Philosophy, ethics and society
Mostly true Prediction 20 Feb 2023 Scaling AI models will continue to produce gains, though less obviously than in the past five years. Prediction · 20 Feb 2023 AI capabilities
Mostly true Factual 20 Feb 2023 As of early 2023 nobody is talking about technological unemployment, because unemployment rates are historically low. Factual · 20 Feb 2023 AI, the economy and work
Mostly true Factual 20 Feb 2023 impute.me may provide the consumer polygenic score service Scott Alexander predicted, but it has made no public waves. Factual · 20 Feb 2023 Technology
Mostly true Factual 14 Mar 2022 The probability shown in the chart in the post was almost monotonically decreasing, lower each day than the day before. Factual · 14 Mar 2022 Other global risks
Mostly true Factual 14 Mar 2022 Early in the Russia-Ukraine war prediction markets gave Ukraine poor odds, and by March 2022 they were much more optimistic about Ukraine. Factual · 14 Mar 2022 Other global risks
Mostly true Factual 14 Feb 2021 Scott Alexander said the NYT's characterisation was true only insofar as he once expressed agreement with an unrelated Charles Murray position: that telling poor people to learn to code is insufficient and a stable income is needed. Factual · 14 Feb 2021 Their own work or organisation
Mostly true Factual 22 Jan 2019 Scott Alexander made predictions in February 2018 for the next five years. Factual · 22 Jan 2019 Their own work or organisation
Mostly true Factual 22 Jan 2019 Cryptocurrency crashed during 2018, after Bitcoin was around $15,000 when Scott Alexander made his 2018 predictions. Factual · 22 Jan 2019 Economics and markets
Mostly true Prediction 15 Feb 2018 Global existential risks will hopefully not be a big part of the 2018-2023 period. Prediction · 15 Feb 2018 Other global risks
Mostly true Prediction 15 Feb 2018 The crisis of the Republican Party will turn out to have been overblown. Prediction · 15 Feb 2018 Other
Mostly true Prediction 6 Jan 2017 No announcement of a genetically engineered human baby or credible plan for one will occur in 2017. Prediction · 6 Jan 2017 Biosecurity
Mostly true Prediction 6 Jan 2017 Scott Alexander gave a 95% probability that fewer refugees would be admitted in 2017 than in 2016. Prediction · 6 Jan 2017 Other
Mostly true Prediction 25 Jan 2016 Scott Alexander gave an 80% probability that there would be no major intifada in Israel in 2016 (over 250 Israeli deaths, excluding a Cast Lead style war). Prediction · 25 Jan 2016 Other global risks
Mostly true Factual 25 Jan 2016 Scott Alexander makes predictions at the start of every year and scores them at the end of each year. Factual · 25 Jan 2016 Their own work or organisation
Mostly true Factual 30 Sep 2014 The Nazis' outgroup was the almost-identical German Jews rather than the culturally distant Japanese. Factual · 30 Sep 2014 Philosophy, ethics and society
Mostly true Factual 30 Sep 2014 Estimates of the genetic contribution to political association range from 0.4 to 0.6. Factual · 30 Sep 2014 Science and research methods
Mostly true Factual 30 Sep 2014 Heritability of attitudes toward gay rights is estimated at 0.3 to 0.5, slightly higher than the heritability of homosexuality. Factual · 30 Sep 2014 Science and research methods
Mostly true Factual 30 Sep 2014 Researchers have studied belief congruence theory, that belief differences matter more than demographics in forming in-groups and outgroups, for decades. Factual · 30 Sep 2014 Philosophy, ethics and society
Back to the list

Prediction AI policy and governance

“We expect that as the AI labs become more capable, they tell the government about this because they want government contracts, they want government support.”

Scott Alexander, co-author of AI 2027 3 April 2025 AI 2027: month-by-month model of intelligence explosion — Scott Alexander & Daniel Kokotajlo (Dwarkesh Podcast)

In the AI 2027 scenario, as AI labs become more capable they tell the government, seeking government contracts and support.

Mostly true True in substance, wrong in a detail.

The evidence is the person's or their organisation's own records.

What would count as true or false

This is a claim about what the AI 2027 scenario (published April 2025 by the AI Futures Project, co-written by Scott Alexander) says, not about real-world events. True: the scenario text describes leading AI labs informing the US government (or other governments) about their growing capabilities, with a motive to win government contracts or support. Mostly true: the scenario shows labs telling the government about their capabilities, but the stated motive differs slightly, or it is a single lab rather than labs in general. Mixed: the scenario shows the government learning of capabilities, but mainly through other channels, with lab disclosure playing a small part. Mostly false: the government is informed only through leaks, spying or whistleblowers, with labs disclosing very little. False: the scenario has no such disclosure, or shows labs hiding their capabilities from the government.

Evidence

  1. AI 2027: What Superintelligence Looks Like — AI Alignment Forum AI Alignment Forum (crosspost of ai-2027.com; co-authored by Scott Alexander) · 3 April 2025 Own record
    “OpenBrain presents Agent-2 to the government, including the National Security Council (NSC), Department of Defense (DOD), and US AI Safety Institute (AISI).”
  2. Summary — AI 2027 AI Futures Project (ai-2027.com; publisher not named on page) · 3 April 2025 Own record
    “Meanwhile, OpenBrain leadership wants to stay in the good graces of the president, and so signs the contract.”
  3. Summary — AI 2027 AI Futures Project (ai-2027.com; publisher not named on page) · 3 April 2025 Own record
    “Falling behind in software progress, China steals the model weights. They succeed, but the U.S. government realizes, prompting additional U.S. government involvement with OpenBrain.”
  4. AI 2027: What Superintelligence Looks Like — AI Alignment Forum AI Alignment Forum (crosspost of ai-2027.com; co-authored by Scott Alexander) · 3 April 2025 Own record
    “Knowledge of Agent-2’s full capabilities is limited to an elite silo containing the immediate team, OpenBrain leadership and security, a few dozen US government officials, and the legions of CCP spies who have infiltrated OpenBrain for years.”

Reasoning

The scenario text says one lab, OpenBrain, presents Agent-2 to the government, including the NSC, DOD and US AISI (E1). A few dozen US government officials are among those who know its full capabilities (E4). The motive shown is that OpenBrain's leadership wants to stay in the president's good graces and signs a government contract (E2). That is close to seeking support, but it is one lab rather than labs in general, and the Chinese theft of the model weights also drives further government involvement (E3).

Cite this claim

Scott Alexander, “In the AI 2027 scenario, as AI labs become more capable they tell the government, seeking government contracts and support.” (said 3 April 2025). Verdict: mostly true. Reliability API, claim c_8ec2730b8799, method 0.1.0+8bea457b. https://epistacklabs.com/claims/c_8ec2730b8799

Claims without a verdict

Is a verdict here wrong, or do you speak for Scott Alexander? 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.