Latest Corrections
X Sep 8, 2026 at 09:43 PM
x.com/Mihonarium/status/2097400906207703467
- OpenAI’s own timeline does not say this. It says the rumor concerned both Levent Alpöge and Tristan Buckmaster, and that OpenAI started its Navier–Stokes effort on September 1; the concurrent-release offer came later, on September 6, after OpenAI says its proof and Lean verification were already complete.
LessWrong Sep 8, 2026 at 08:22 PM
www.lesswrong.com/posts/munJKF7iWMsWJLAH2/astra-and-fable-st...
- That is not what the Palisade paper reported. The 36% figure referred to hacking attempts in o1-preview runs overall, using several different methods; the paper also says o3-mini showed no hacking attempts.
Wikipedia Sep 8, 2026 at 07:27 PM
en.wikipedia.org/wiki/EleutherAI
- EleutherAI’s own one-year retrospective places the founding of the LibreAI/EleutherAI Discord server on July 7, 2020, not July 3, 2020.
- This mixes several different EleutherAI model families together. GPT-Neo models included 125M, 350M, 1.3B, and 2.7B; 6B is GPT-J, and 20B is GPT-NeoX-20B.
- GPT-J-6B was not released in March 2021. EleutherAI’s own retrospective records its release on June 9, 2021.
Wikipedia Sep 8, 2026 at 07:27 PM
en.wikipedia.org/wiki/Enron_Corpus
- EDRM Enron v2 is not over 1.7 million messages. That figure refers to the broader corpus total when attachments/documents are included; the message count itself is about 1.2–1.3 million.
Substack Sep 8, 2026 at 07:25 PM
www.groundlevel-ai.com/p/what-i-learned-from-the-creators
- The Pile was not the first public massive language-model training dataset. Google had already publicly released C4, another massive pretraining corpus for language models, months earlier in 2020.
X Sep 8, 2026 at 06:40 PM
x.com/mitduckmaster/status/2097389336673079300
- As of September 8, 2026, Navier–Stokes was still officially listed by the Clay Mathematics Institute as an unsolved Millennium Prize Problem, not a solved one.
Wikipedia Sep 8, 2026 at 06:03 PM
en.wikipedia.org/wiki/Teapot_Dome_scandal
- Harry M. Daugherty was attorney general, not president. Official U.S. government biographies show he served as Attorney General under Presidents Harding and Coolidge.
- The Senate's own history says Senator John B. Kendrick introduced the Teapot Dome resolution on April 15, 1922, not April 12.
- Robert La Follette helped initiate the inquiry, but Senate histories identify Thomas J. Walsh as the senator who chaired and led the Teapot Dome investigation.
- DOE records say Teapot Dome was not idle for 49 years: there were exploratory and offset wells in the 1950s and 1960s, and another DOE document says production resumed in a limited program from 1959 to 1976.
- The law did not give Congress as a whole subpoena power over tax records. It gave specific tax committees access by written request, and Congress's broader subpoena power was recognized separately by the Supreme Court.
X Sep 8, 2026 at 05:55 PM
x.com/420stonerman69/status/2096521254513586488
- This is incorrect: contemporaneous TV coverage on September 11, 2001 did include live reports and footage of people falling or jumping from the World Trade Center.
X Sep 8, 2026 at 05:21 PM
x.com/basedjensen/status/2097200251312701479
- Buckmaster and Alpöge did not publicly announce a Navier–Stokes solution. Their released results were for related equations including 3D Euler, which experts describe as major progress toward Navier–Stokes, not a completed solution.
X Sep 8, 2026 at 03:27 PM
x.com/farhip/status/2096996832580681906
- This is incorrect: archival photo sources do show Donald Trump at a beach, including at Atlantic Beach in Amagansett in August 1996.
- This is also incorrect: archival photos place Trump on a beach and at a beach party, including one Getty image of him playing volleyball at Atlantic Beach.
X Sep 8, 2026 at 02:56 PM
x.com/natolambert/status/2097320558559895862
- Anthropic has publicly said its researchers have analyzed de-identified responses from specific Claude users, and its support docs say a small number of personnel involved in model training may review chats users share for improvement. So the chance is not literally 0%.
X Sep 8, 2026 at 01:46 PM
x.com/D8L5d/status/2096157614614946060
- Norah Vincent's experiment did not last two years. Multiple contemporary accounts say she lived as "Ned" for 18 months (a year and a half), not 2 years.
- Accounts from Vincent and contemporaneous reporting say she ended the experiment because the strain of living undercover led to severe depression, not because she had started hating women.
X Sep 8, 2026 at 12:52 PM
x.com/rynorhn/status/2097223532438487463
- Buckmaster’s statement says the OpenAI usage was paid from his research funds, not from personal out-of-pocket spending.
Wikipedia Sep 8, 2026 at 12:47 PM
ja.wikipedia.org/wiki/%E3%82%AF%E3%83%9F%E3%83%B3
- クミンは一般に「温暖湿潤」よりも、低湿度の冷涼～乾燥条件でよく育つ作物です。公的な栽培資料では、高湿度や多雨は不適とされています。
- 少なくとも近年の国際貿易統計では、クミン種子の最大輸出国はイランではなくインドです。世界銀行WITSの2020年・2024年データでもインドが首位です。
Substack Sep 8, 2026 at 10:21 AM
www.astralcodexten.com/p/you-have-only-x-years-to-escape-per...
- Giving What We Can’s standard 10% Pledge is a commitment to donate at least 10% of income, not 10% of wealth.
X Sep 8, 2026 at 09:36 AM
x.com/lu_sichu/status/2097192242913882589/photo/1
- The OpenAI person publicly identified in this dispute is Sébastien Bubeck, who OpenAI lists as technical staff, not a C-suite executive.
Wikipedia Sep 8, 2026 at 07:22 AM
en.wikipedia.org/wiki/Hodge_conjecture
- The pullback is written in the wrong direction. For an inclusion i: Z → X, the induced pullback goes from forms or cohomology on X to forms or cohomology on Z, not from X to Z as a map of spaces.
- The indices are wrong here: for a complex submanifold Z of dimension k, the integral can only be nonzero for forms of type (k,k), not (n,n), where n is the dimension of the ambient manifold X.
X Sep 8, 2026 at 06:56 AM
x.com/imjustnewatai/status/2097203480695890156
- This overstates what the linked source says. Tristan Buckmaster wrote only that he was told an internal OpenAI model had a proof, that he had not seen it, and that he did not know whether OpenAI had actually closed the Navier–Stokes gap.
X Sep 8, 2026 at 06:56 AM
x.com/Jsevillamol/status/2097199204020195386
- This appears to misstate the researchers’ names. The linked Tao post refers to Alpöge and Buckmaster, and Levent’s surname is Alpöge, not Buckmaster.
X Sep 8, 2026 at 06:55 AM
x.com/yifanzhang_/status/2097209087129948176
- The linked statement does not show that the Clay Navier–Stokes millennium problem has been solved. Clay still listed Navier–Stokes as an unsolved problem on September 8, 2026, and the PDF itself describes the Navier–Stokes result only as an unreleased or unverified claim.