Breaking news, and how the end might begin
We recorded 24 claims from this post. 11 were checked against evidence. 13 can’t be checked. 13 more are held. Read the post at its source.
Claims from this post (24)
In the order they appear in the post. Each quotation is the speaker’s words in this post. A label such as “Supported” or “Can’t be checked” is the claim’s current assessment; its page has the evidence and the reasons.
Assessment: Can’t be checkedSaid
“I think it’s a low probability of the bet coming through, but it’s not zero.”
What we tested: The giant GenAI investment bet has a low but nonzero probability of paying off.
Assessment: UnresolvedSaid
“Open AI, I think, has a problem right now. They are burning money the fastest.”
What we tested: OpenAI has a problem and is burning money faster than its AI competitors.
Assessment: MixedSaid
“They had a lead but they squandered it.”
What we tested: OpenAI had a lead in AI but squandered it.
What held: GPT-4 ranked first in the reported April 24–May 8 voting window. This supports the earlier-lead component, without proving superiority on every task. (LMSYS, )
What did not: The evaluator reports GPT-5.5 leading its Intelligence Index by three points and leading several individual evaluations. This contradicts an unqualified reading that OpenAI had ceased to hold a technical lead. (Artificial Analysis, )
Assessment: UnresolvedSaid
“And so at some point, I think OpenAI is gonna have a problem where they can’t really meet their obligations.”
What we tested: At some point OpenAI will be unable to meet its financial obligations.
Assessment: SupportedSaid
“Anthropic has taken a lot of the business market and OpenAI has lost a lot of it.”
What we tested: Anthropic has taken a large share of the business market while OpenAI has lost much of it.
What happened: Menlo reports Anthropic’s enterprise LLM API share rising from 12% in 2023 to 40% in 2025, while OpenAI’s fell from 50% to 27%. Both movements are substantial under the ordinary share-based reading. (Menlo Ventures, distributed by GlobeNewswire, )
Assessment: Can’t be checkedSaid
“It’s pretty clear he should not be trusted.”
What we tested: It is pretty clear that Sam Altman should not be trusted.
Assessment: Can’t be checkedSaid
“they’re clearly afraid of anthropic.”
What we tested: OpenAI, in rushing its IPO ahead of Anthropic, is clearly afraid of Anthropic.
Assessment: Can’t be checkedSaid
“I tend to think that OpenAI is the one that goes down first.”
What we tested: OpenAI is likely the first major AI company to fail.
Assessment: Can’t be checkedSaid
“There are others, but that that is the most likely scenario to me is that openAI at some point can’t really make ends meet.”
What we tested: The most likely scenario for the AI boom breaking is that OpenAI at some point can't make ends meet.
Assessment: SupportedSaid
“And the reason that those tend to be more expensive is that they tend (with the help of some good old fashioned AI that nobody wants to talk about) to run multiple large language models in the background many times until they get the answers that by some criteria seem to correct.”
What we tested: AI agents tend to be more expensive because they tend to run multiple large language models in the background many times until they get answers that by some criteria seem correct.
What happened: Provider data shows agents, and especially multi-agent systems, use several times more tokens than chat, which drives cost. (Anthropic, )
Assessment: UnresolvedSaid
“First of all, domino one: tokenmaxxing has indeed clearly died, just as Eisman and I had anticipated a few weeks ago.”
What we tested: Tokenmaxxing has clearly died, as Marcus and Eisman anticipated.
Assessment: Can’t be checkedSaid
“SoftBank took an immediate hit; dropping the ask reeks of desperation, and its a terrible vote of nonconfidence that the banks still said no:”
What we tested: SoftBank took an immediate hit, its reduced margin-loan ask reeks of desperation, and the banks still refusing it is a terrible vote of no confidence.
Assessment: Can’t be checkedSaid
“That’s domino 2: we can infer that the banks don’t really think that OpenAI is worth its current valuation. They don’t want OpenAI stock as collateral.”
What we tested: Banks evidently do not think OpenAI is worth its current valuation and don't want its stock as collateral.
Assessment: Can’t be checkedSaid
“So maybe domino 3 is this: OpenAI may well wind up as the WeWork of AI.”
What we tested: OpenAI may well end up as the WeWork of AI.
Assessment: SupportedSaid
“And OpenAI still has yet to turn a profit, and it’s still not clear they ever will.”
What we tested: OpenAI has not yet turned a profit and it is unclear whether it ever will.
What happened: In the quarter just before the post, OpenAI had a $9.3 billion operating loss and a $21.3 billion net loss. That fits 'has yet to turn a profit' under either definition. (The Decoder (citing The Information), )
Assessment: SupportedSaid
“Core problems of reliability and factuality haven’t been solved,.”
What we tested: The core problems of reliability and factuality in LLMs have not been solved.
What happened: An independent test from April 2026 shows the three leading frontier models still giving confident wrong answers instead of abstaining, at rates of 36% to 86% of non-correct responses. That is not a solved problem. (Artificial Analysis, )
Assessment: UnresolvedSaid
“Two years later, competitors have caught up. OpenAI has had to cut prices.”
What we tested: Over the past two years competitors have caught up to OpenAI and it has had to cut prices.
Assessment: SupportedSaid
“Opensource models (though perhaps not Meta’s) are only a few months behind.”
What we tested: Open-source models, perhaps excluding Meta's, are only a few months behind frontier models.
What happened: An average four-month lag over Jan to May 2026 fits 'only a few months behind'. (Epoch AI, )
Assessment: Can’t be checkedSaid
“Every bad sign is worse now. The big difference is that people are finally noticing.”
What we tested: Every warning sign about OpenAI is worse now than two years ago, and people are finally noticing.
Assessment: Can’t be checkedSaid
“now the entire world is worried about whether AI is a bubble. And rightly so.”
What we tested: The entire world (hyperbolically) is now worried about whether AI is a bubble, and rightly so.
Assessment: Can’t be checkedSaid
“That Wile E Coyote moment could turn out to be soon.”
What we tested: The moment when the AI boom collapses, as everyone realizes they got it wrong, could come soon.
Assessment: Can’t be checkedSaid
“If the market gets squirrelly on OpenAI’s IPO, which after today seems like a real possibility, OpenAI might be forced to withdraw its IPO (Domino 3), or to cut their aspirations sharply.”
What we tested: If the market turns skittish on OpenAI's IPO (a real possibility), OpenAI might be forced to withdraw its IPO or sharply cut its aspirations.
Assessment: UnresolvedSaid
“If they can’t meet their obligations, I expect Nvidia and Oracle and many others would take a big hit as collateral damage (Domino 5).”
What we tested: If OpenAI can't meet its obligations, Nvidia, Oracle and many others would take a big hit as collateral damage.
Assessment: Can’t be checkedSaid
“And the damage might affect the banking system, retirement plans and more (too many dominoes to count).”
What we tested: If OpenAI can't meet its obligations, the damage might spread to the banking system, retirement plans and more.
Held from this post (13)
These records did not pass every check, so nothing about them is published. Held is not a finding.
- The two reviews read the claim differently
- The two reviews read the claim differently
- The record does not meet the current rules. Our reading, not assessed: The latest models (other than GPT-5) are costly to run and cost is a growing issue.
- The record does not meet the current rules. Our reading, not assessed: If OpenAI is forced to withdraw its IPO or sharply cut its aspirations, Marcus does not see how it could meet its immense (about $600B) obligations.
- The record does not meet the current rules. Our reading, not assessed: The end of tokenmaxxing means less revenue for LLM companies, which will hurt their IPOs.
- The record does not meet the current rules. Our reading, not assessed: LLM providers' margins on serving tokens are very slim or negative in many cases.
- The criteria are ambiguous. Our reading, not assessed: Compared with the rival AI providers he named (Anthropic, Google, maybe Amazon), OpenAI is spending the most money with the fewest assets and has made the most commitments.
- The record does not meet the current rules. Our reading, not assessed: When the financial numbers are fully disclosed, they will show Anthropic is more efficient than OpenAI.
- The criteria are ambiguous. Our reading, not assessed: LLM providers no longer want to offer all-you-can-eat pricing and customers no longer want to pay per-token a la carte.
- The two reviews read the claim differently
- The record does not meet the current rules. Our reading, not assessed: LLM providers have been running their services at a loss, and agents make the loss even bigger.
- The two reviews read the claim differently
- The criteria are ambiguous. Our reading, not assessed: OpenAI will find it hard to IPO at an even higher valuation given the failed SoftBank loan.