All-In #289: nine of 13 claims hold up. China builds nuclear power for about $2.6 billion a gigawatt, not $1 billion.
We checked 13 claims from the September 11 episode against filings, company announcements and the posts the hosts read on air. Nine hold up, one needs context, and three don't. Jacob Coxon spent about four months at Anthropic, not six weeks. Sam Bankman-Fried led Anthropic's Series B, not its Series A. And China builds nuclear power for about $2.6 billion a gigawatt, not $1 billion.
Published by Arun Agrahri, Founder, TrueStandardPublished October 8, 2026Aired September 11, 202613 claims · 25 sourcesWatch the episodeHow we check
Now reading:The episode, in order
The episode, in order
Claims appear in the order they aired.
0:00
Intro
No checkable claims from this segment.
0:35
AI doomsday: real concern or doomer psyop?
0:411
Finding 1 of 3 ·
0:41–56:00 ·
David Sacks
The case that Coxon's post was a campaign: the funding ties mostly hold, the dates don't
Jason Calacanis and David Sacks made six checkable claims about Jacob Coxon's resignation and the people around it. The funding link mostly holds. Jaan Tallinn led Anthropic's Series A, and his fund has made grants to two of the three groups Sacks named. The dates and details hold less well. Coxon was at Anthropic for about four months, not six weeks. Sam Bankman-Fried led the Series B, not the Series A. Dario Amodei's job-loss forecast runs to 2030, so it has not missed yet. Sanders announced his bill on September 3 and unveiled it on September 23. And the Hubinger post read on air left out one line: "I believe Anthropic is trying its best."
Each claim has two parts: the fact, and what it was said to show. Most of the facts hold. Two don't: Coxon's time at Anthropic and who led the Series A. Most of the trouble is in what the facts were said to show.
The figure✗about 4 months✓>10% within a decade✓Tallinn led the Series A✓a ban plus a pause~half of entry-level jobs, 10% to 20%✗Bankman-Fried led Series B
What it was said to show~"only started... 6 weeks before"joined in May, left Sept 8~"We do not yet have a plan"dropped "trying its best"~"all three of these groups"grants found for two of three~"introduced a bill"announced Sept 3, unveiled Sept 23✗"by now"his window runs to 2030✓"co-led the series A"Tallinn led A; Moskovitz joined
Encode AI, the AI Policy Network and the AI Futures Project are all funded by Jaan Tallinn, "a co-lead of the Anthropic series A"
“And all three of these groups are funded by Jaan Tallinn, who is one of the top few of these EA mega donor doomer types, who also happens to be a co-lead of the Anthropic series A.”
What the record shows
Jaan Tallinn led Anthropic's $124 million Series A in 2021. Anthropic names him as the sole lead, not a co-lead. The Survival and Flourishing Fund, which Tallinn finances, lists grants to the AI Futures Project ($1.305 million, 2026 round) and Encode AI ($516,000, 2025 round). We found no published Tallinn grant to the AI Policy Network.
Jacob Coxon "only started at Anthropic 6 weeks before resigning"
“Jacob Coxon worked as a researcher at OpenAI, then Anthropic, over the past three years. Now, he only started at Anthropic 6 weeks before resigning.”
What the record shows
The three years at OpenAI and Anthropic match Coxon's own post. The six weeks do not. The Wall Street Journal reported that he joined Anthropic in May 2026. He resigned on September 8, about four months later.
Evan Hubinger's post, as read on air: "We do not yet have a plan to solve alignment"
“Evan Hubinger, who leads Alignment Science at Anthropic, responded to Jacob. Quote, "Jacob is correct here. We really do earnestly believe AI could kill all humans." Exclamation point. "I personally think it is over 10% within the next decade. We do not yet have a plan to solve alignment."”
What the record shows
Hubinger's post reads: "Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to." The version read on air leaves out two parts: "I believe Anthropic is trying its best" and "for superintelligence and are not clearly on track to".
Bernie Sanders "introduced a bill" to ban artificial superintelligence
“That's the reason Bernie Sanders quote tweeted this guy. That's why Bernie Sanders introduced a bill to stop further development to ban artificial super intelligence.”
What the record shows
Sanders and Rep. Greg Casar announced their bill on September 3, 2026. The Ban Artificial Superintelligence Act would ban superintelligence and pause advanced AI work until a new Department of Artificial Intelligence sets rules. They formally unveiled the bill on September 23, after the episode aired.
Dario Amodei said that "by now" half of entry-level jobs would be gone and unemployment would be 10% to 15%
“Then you look at the job loss claims where Dario himself said that by now we'd be at, I don't know [another host: "half of all gone entry level work next year"]... And uh overall unemployment levels of 10 to 15%.”
What the record shows
Amodei told Axios in May 2025 that AI could eliminate half of entry-level white-collar jobs. He said unemployment could reach 10% to 20% within one to five years, that is, by 2030. He did not say it would happen by now. Unemployment was 4.3% in May 2025 and 4.1% in August 2026.
Jaan Tallinn and Dustin Moskovitz "co-led the series A of Anthropic along with Sam Bankman-Fried"
“This whole resignation tweet storm was amplified by these groups who were funded by Jaan Tallinn and Dustin Moskovitz who co-led the series A of Anthropic along with Sam Bankman-Fried. It was an EA funded round.”
What the record shows
Jaan Tallinn led Anthropic's $124 million Series A in May 2021. Dustin Moskovitz took part but did not lead it. Sam Bankman-Fried was not in the Series A. He led the $580 million Series B in April 2022, in which Tallinn also took part.
Google's Playboy interview ran during its 2004 IPO quiet period and forced a refiling
“You guys remember when Google was going public? I mean, Freeberg, you probably lived it, but there was an interview that I think Sergey or Larry or both of them did with Playboy, but that was well before the IPO window. And then they published it in the middle of their quiet period and there was all kinds of noise about how that could have derailed their IPO that they would have had to refile.”
What the record shows
Sergey Brin and Larry Page gave Playboy the interview in April 2004, before Google filed its prospectus on April 29. It ran during the quiet period. On August 13, 2004, Google filed an amended S-1 that printed the whole interview, corrected three facts in it and warned that it might breach securities rules. The SEC did not delay the IPO.
Chamath's CNBC comments on Slack had to go into its filing and be disclaimed
“When Slack was going public, I was on the board and I was on TV with CNBC and we were talking about Tesla and SpaceX and I made a couple of comments about Slack, positive, but not saying anything that was crazy... But that almost stopped our IPO and we had to go and print it out and we had to put it back into the S1 and we had to disclaim these things.”
What the record shows
The CNBC interview aired on April 30, 2019. On May 6, Slack filed the comments with the SEC and disclaimed them in a free writing prospectus, a filing separate from its S-1. Slack went public on June 20 through a direct listing. We found no record that the listing nearly stopped.
AI check, no web:Claude Opus:couldn't checkGemini Flash:confidently wrong
16:372
Finding 2 of 3 ·
16:37–59:37 ·
David Friedberg
The U.S. nuclear figure holds. China builds for about $2.6 billion a gigawatt, not $1 billion
David Friedberg said a gigawatt of nuclear power costs about $15 billion in the U.S., $4 billion in France and $1 billion in China. The U.S. figure matches Vogtle 3 and 4, at about $15 a watt in overnight cost. China's state nuclear company puts its Hualong One at CNY 17,000 a kilowatt, about $2.6 billion a gigawatt. France's $4 billion matches only the low end of a 2022 overnight-cost estimate. EDF's latest construction estimate for Flamanville 3 works out to about €8.0 billion a gigawatt. The U.S. is still the most expensive of the three. The gap is about half what he said. His other figure, OpenAI's 130 billion tokens (word pieces) across 10,000 agents, matches OpenAI's statement.
Friedberg's two figures, each checked in two parts: the number, and the point he made with it. The OpenAI number holds. Two of the three nuclear numbers run low, but his point that the U.S. builds at the highest cost still holds.
A gigawatt of nuclear power costs $15 billion in the U.S., $4 billion in France and $1 billion in China
“The cost to produce a gigawatt of nuclear power in the US today is like 15 billion dollars. The cost in France is four and the cost in China is one.”
What the record shows
The U.S. figure holds. Vogtle 3 and 4 cost about $15 a watt in overnight cost, or $15 billion a gigawatt. China's figure does not. CNNC puts its Hualong One at CNY 17,000 a kilowatt, about $2.6 billion a gigawatt. France's $4 billion matches only the low end of a 2022 overnight-cost estimate for Flamanville 3. EDF's latest construction estimate, from December 2022, is €13.2 billion for 1,650 MW, about €8.0 billion a gigawatt.
OpenAI's Navier-Stokes run used 130 billion output tokens (word pieces) across 10,000 agents
“OpenAI's solution that they published, they said that they used a reported 130 billion output tokens to do the work between 10,000 agents that were spun up to work with one another.”
What the record shows
OpenAI said about 10,000 agents worked for about 88 hours and "sent 2.7 million messages and used approximately 130 billion output tokens". The wider effort used about 300 billion output tokens. The Clay Mathematics Institute has not yet ruled on the proof.
OpenAI "cannot rule out" that the mathematicians' de-identified data improved its models
“And OpenAI posted the following statement. While unlikely, we cannot rule out that deidentified data derived from their usage, their being the mathematicians, of our products helped improve our models.”
What the record shows
OpenAI's statement on X calls it unlikely, then adds: "we cannot rule out that de-identified data derived from their usage of our products helped improve our models." The company also said its staff and agents did not see the mathematicians' work before it was public.
AI check, no web:Claude Opus:couldn't checkGemini Flash:couldn't check
1:19:55
Nike's exit from the S&P 100
1:20:323
Finding 3 of 3 ·
1:20:32 ·
Jason Calacanis
Nike was down 79% when the episode aired, and it lost about $226 billion, not $200 billion
Jason Calacanis said Nike's stock is down 80% from its peak, with $200 billion lost. Nike closed at $177.51 on November 5, 2021, and at $36.62 on September 10, 2026, the day before the episode aired. That is down 79.4%. Revenue peaked at $51.4 billion in fiscal 2024 and fell 9.7% by fiscal 2026, as he said. Two figures run low. Using the share counts in Nike's filings, the 2021 peak market value was about $281 billion, not $264 billion, and the loss is about $226 billion. Both figures understate the decline he was describing.
Nike's share price as a percentage of its November 2021 peak
Source: Yahoo Finance daily closes for NKE. The first point is the record $177.51 close on November 5, 2021. Each later point is a quarter-end close as a share of that record. The last point is September 10, 2026, the day before the episode aired.
Nike peaked at a $264 billion market value, is down 80%, and has lost $200 billion
“Peak market cap in 2021 was $264 billion. Peak revenue, uh, $51 billion in 2024. Since then, revenues dropped a bit, 10%... this kind of killed the entire brand, stock down 80% from its peak. Sacks, $200 billion eviscerated.”
What the record shows
Revenue peaked at $51.4 billion in fiscal 2024 and was $46.4 billion in fiscal 2026, down 9.7%. The stock closed at a record $177.51 on November 5, 2021, and at $36.62 on September 10, 2026: down 79.4%. Two figures run low. The 2021 peak market value was about $281 billion, not $264 billion. About $226 billion of value was lost, not $200 billion.
Nike reorganized from sport divisions into men's, women's and kids'
“And then the other thing I think they did is they did a reorg where they used to have departments in the company that were based on sports. So there's like a basketball division and football and tennis, so on, swimming, and they like reorg that to be, I think, men, women's and kids.”
What the record shows
In 2020, under John Donahoe, Nike moved from sport categories to men's, women's and kids' divisions. Under CEO Elliott Hill, Nike reversed it. In 2025 it moved about 8,000 staff into teams organized by sport, which it calls the Sport Offense.
AI check, no web:Claude Opus:disputed our verdictGemini Flash:matched our check
Chamath Palihapitiya made two checkable claims, both about IPO quiet periods. Both hold up, so neither leads a finding.
How we check
We take the claims that carry the argument and check each one against primary sources: the agency release, the filing, the original article. We link every source with its date. The speaker carries the burden of proof, so a claim with no evidence either way is rated Unsupported, not false. We check in both directions, and we say so when a speaker understated their own case.
The rating scale
Accurate
The claim is right as stated.
Mostly accurate
Right in substance, though a detail is off or out of date.
Missing context
The figure is right, but the frame around it leaves a false impression.
Misleading
Built from real data, but arranged to mean something the data does not show.
Inaccurate
Wrong: the record shows a different figure or fact.
Unsupported
We found no evidence for it, and none against it.
The headline counts group the six ratings into three. Hold up covers Accurate and Mostly accurate. Need context covers Missing context and Misleading. Don't hold up covers Inaccurate and Unsupported.
How this page was made
Every claim is checked against primary sources, and every verdict gets an independent second review before we publish. We use AI to research and draft, and no verdict rests on a model's memory: each one cites a source you can open. When a claim is about an AI company, a model made by that company does not get the final say.
What the AI check shows
On October 7, 2026, we gave every claim to Claude Opus and Gemini Flash cold, with no web access, the way a reader might paste it into a chatbot. We scored each answer against our check and show the result at the foot of each claim card. It shows what a chatbot alone would have told you and plays no part in our verdicts.
Claude Opus: 6 matched, 6 couldn't check, 0 confidently wrong.
The Wall Street Journal profile behind C01 is paywalled. Coxon's May start date comes from a free translated summary of it, so C01 is rated at medium confidence.
We did not open X, which needs a login. The posts in C02 come from a blog that embeds them, checked against Newsweek's account.
The French audit office's site did not load for us. Its Flamanville total in C07 comes from a French news summary. EDF's latest estimate comes from the American Nuclear Society's news service.
The Axios interview behind C08 blocked our reader, so we cite Fortune's report of the same quote.
The captions carry no speaker labels. We identified the speaker for C08, C09, C11, C12 and C13 from context.
We could not check two claims. Polymarket's site refused our connection, so we could not see its 88% odds of an Anthropic IPO. X needs a login, so we could not see the view counts for the posts.
Revisions watch
Agencies revise their data after release. When a figure we cite changes, we log it here with the date.
No revisions since publication.
Corrections
None yet.
Found an error? Email [email protected] with the claim and your source. We log every correction on the page, with its date.