Claims checked against evidence by Gary Marcus, page 2 of 15
Deadlines are judged as of 2026-10-29. About Gary Marcus and the sources covered.
Claims checked against evidence by Gary Marcus (1471)
Each of these claims was checked against dated evidence.
Page 2 of 15: claims 101 to 200 of 1471, ordered by the date first said, newest first. To find a claim by its words, use search.
Assessment: MixedSaid
Observation
“They had a lead but they squandered it.”
Our reading: OpenAI had a lead in AI but squandered it.
No horizon Evidence as of
Limitation: Different evaluations measure different capabilities and preferences.
Assessment: SupportedSaid
Observation
“And, breaking news, costs are rapidly rising too:”
Our reading: The costs of developing AI data centers, including OpenAI's Stargate, are rising rapidly.
No horizon Evidence as of
Limitation: The Stargate-specific costs Marcus pointed to were not directly checked.
Source: An entire industry is being propped up by math that is insane.
Assessment: UnresolvedSaid
Forecast
“That is absolutely not going to happen.”
Our reading: SpaceX will not return 2,538x its IPO investment as Amazon did, which would imply a market cap of about $4,442 trillion.
No deadline stated Evidence as of
Limitation: The evidence covers only the June 12, 2026 debut day. It does not show SpaceX's share price or market value at any later date, including the evaluation date.
Source: An entire industry is being propped up by math that is insane.
Assessment: UnresolvedSaid
Observation
“The one possible major productivity story (coding) also happens to be the one burning the most cash.”
Our reading: AI coding, the one possible major productivity success, is also the use case burning the most cash.
No horizon Evidence as of
Limitation: The target quarter and cash-burn definition are evaluator interpretations because Marcus did not specify them.
Source: Slop, productivity, and why the AI-fueled world is going nowhere mighty fast
Assessment: UnresolvedSaid
Conditional forecast
“As I noted on X, nobody is going to trust an American AI company that is partly owned by the US Government.”
Our reading: If an American AI company is partly owned by the US government, nobody is going to trust it.
No deadline stated Evidence as of
Only if: if an American AI company is partly owned by the US government
Source: AI’s Black Friday
Assessment: UnresolvedSaid
Causal claim
“None of that is particularly great for Nvidia, CoreWeave, Oracle, Nebius, Micron, and a variety of other hardware and semiconductor providers”
Our reading: The developments described (SpaceX leasing compute to Google and Anthropic, and firms buying hardware before there was demand for it) are not particularly great for Nvidia, CoreWeave, Oracle, Nebius, Micron, and a variety of other hardware and semiconductor providers.
No horizon Evidence as of
Limitation: The customer accounts do not establish what purchases or contracts would otherwise have occurred [e1, e2].
Source: AI’s Black Friday
Assessment: UnresolvedSaid
Observation
“AGI: machine can do anything human can do, autonomously. That has not achieved]”
Our reading: AGI, a machine that can autonomously do anything a human can do, has not been achieved.
No horizon Evidence as of
Limitation: This is an evidence limitation, not a finding that AGI had been achieved.
Source: No need to panic about Anthropic’s new blog, and some more good news
Assessment: UnresolvedSaid
Forecast
“Getting to AGI will require new ideas, not just new code optimizations.”
Our reading: Reaching AGI will require new ideas, not merely new code optimizations.
No deadline stated Evidence as of
Limitation: The meaning of new ideas remains partly judgmental.
Source: No need to panic about Anthropic’s new blog, and some more good news
Assessment: SupportedSaid
Observation
“Which is funny, because Sir Demis Hassabis [let’s call him Sir Demis Hassabis #2] , just a few months earlier in January 2026 at Davos, had something rather different to say, anticipating a somewhat slower timeline of 2031-2036:”
Our reading: Hassabis predicted AGI around 2030 in mid-2026 but gave a slower 2031-2036 timeline at Davos in January 2026.
No horizon Evidence as of
Limitation: The Davos evidence is a republication of an edited interview; Stanford evidence is reporting rather than a directly retrieved recording transcript.
Assessment: UnresolvedSaid
Forecast
“We won’t see AGI this decade, and we not even close, exactly for the reasons that he laid out.”
Our reading: AGI, defined as a system with all human cognitive capabilities, will not be achieved this decade (by the end of 2029).
Not yet due Evidence as of
Limitation: The endorsed definition is demanding but lacks a specified evaluation protocol.
Assessment: SupportedSaid
Observation
“Today, most definitely not on my bingo card, President Trump signed an executive order that was very much along on these lines:”
Our reading: On 2026-06-02 President Trump signed an executive order closely resembling Marcus's proposal for preflight checks on high-impact AI models.
No horizon Evidence as of
Limitation: A reading that requires mandatory FDA-style approval would not be supported.
Assessment: UnresolvedSaid
Conditional forecast
“If there is no moat, nobody is going to take 90% of the market.”
Our reading: If there is no moat, no AI company will capture 90% of the AI market.
No deadline stated Evidence as of
Only if: if there is no moat
Assessment: UnresolvedSaid
Forecast
“Prediction: Nobody knows when this will all collapse, but 2026 will be remembered in hindsight as the year in which retail investors and index funds were left holding the bag.”
Our reading: 2026 will be remembered as the year retail investors and index funds were left holding the bag on AI investments, though the timing of the collapse is unknown.
Not yet due Evidence as of
Limitation: The sole cited investment example is contextual, not a measure of the full claimed investor outcome.
Assessment: UnresolvedSaid
Causal claim
“And companies like Anthropic are ending their all you can eat buffet, which will make many customers think twice.”
Our reading: Anthropic and similar companies are ending all-you-can-eat pricing, which will make many customers think twice about using them.
No horizon Evidence as of
Limitation: The policy evidence concerns particular Anthropic products and uses.
Assessment: UnresolvedSaid
Forecast
“Same will happen with Anthropic and OpenAI.”
Our reading: As with SpaceX, Anthropic and OpenAI will be pushed into indices and index trackers and retail investors will suffer losses when reality catches up.
No deadline stated Evidence as of
Limitation: Anthropic's preliminary prospectus does not establish a completed offering, index entry or investor losses, and provides no evidence of OpenAI's status.
Assessment: UnresolvedSaid
Conditional forecast
“With no clear winners, nobody can charge monopoly prices; instead, you get price wars and commodity pricing.”
Our reading: If there are no clear winners among AI competitors, nobody will be able to charge monopoly prices, and instead there will be price wars and commodity pricing.
No deadline stated Evidence as of
Only if: with no clear winners
Assessment: UnresolvedSaid
Causal claim
“But everybody is building essentially the same technical solution with essentially the same data, so there is no moat.”
Our reading: Because everybody competing to build AI is building essentially the same technical solution with essentially the same data, there is no moat for any of them.
No horizon Evidence as of
Limitation: The observations precede June 2, 2026; the Stanford report became available on June 29.
Assessment: SupportedSaid
Observation
“Dawkins seems to imagine that since LLMs say things people do, they must be like people, and that simply does not follow.”
Our reading: It does not follow that LLMs are like people just because they say things people say.
No horizon Evidence as of
Limitation: The evidence concerns selected tasks in one model and comes from its developer.
Source: The Pope appears to understand AI better than Geoffrey Hinton does.
Assessment: MixedSaid
Observation
“And the differences are immense; one (the LLM) effectively memorizes the entire internet; the other (the human) builds a mental model through experience with world.”
Our reading: LLMs differ immensely from humans: LLMs effectively memorize the internet, while humans build mental models through experience of the world.
No horizon Evidence as of
Limitation: The result depends on reading the sentence as a contrast between modes of learning, rather than merely saying humans and LLMs differ.
Source: The Pope appears to understand AI better than Geoffrey Hinton does.
Assessment: UnresolvedSaid
Observation
“They are creating interactive fiction that is trained to predict the language of actual beings.”
Our reading: LLMs are interactive fiction trained to predict the language of actual beings.
No horizon Evidence as of
Limitation: On a purely behavioral reading, the role-play characterization has support.
Source: The Pope appears to understand AI better than Geoffrey Hinton does.
Assessment: UnresolvedSaid
Observation
“Within hours of when I posted yesterday’s essay on the decline of tokenmaxxing, I saw further converging evidence, such as this:”
Our reading: There is further converging evidence, as of late May 2026, that tokenmaxxing (the push to maximize AI token usage) is declining.
No horizon Evidence as of
Limitation: Amazon's precise leaderboard shutdown date is unspecified (e1).
Source: What happens next, after the decline of tokenmaxxing?
Assessment: UnresolvedSaid
Observation
“Consistent with the Axios quote, and the Uber concern’s I expressed a couple days ago, I am seeing more and more reports online like this:”
Our reading: Marcus is seeing more and more online reports like the quoted one, about companies burning money on tokens without significant ROI.
No horizon Evidence as of
Limitation: The item provides no comparable observation windows or complete record of Marcus's encounters [e1].
Source: Breaking: bad news for three of the biggest IPOs in history
Assessment: UnresolvedSaid
Causal claim
“One big reason why Anthropic is having a blowout quarter (aside from the massive one time subsidy that they got from SpaceX) is that companies have been “tokenmaxxing” for the last few months, encouraging employees to use GenAI as much possible, without much regard to payoff.”
Our reading: A major reason for Anthropic's blowout quarter, besides a one-time SpaceX subsidy, is companies tokenmaxxing: pushing employees to use GenAI as much as possible without much regard to payoff.
No horizon Evidence as of
Limitation: The filing covers a broader period than the quarter-to-date claim.
Source: Breaking: bad news for three of the biggest IPOs in history
Assessment: UnresolvedSaid
Causal claim
“That means lots of short-term revenue to Anthropic and OpenAI, but it never seemed like a sustainable idea to me:”
Our reading: Tokenmaxxing means lots of short-term revenue for Anthropic and OpenAI, though it never seemed to Marcus like a sustainable idea.
No horizon Evidence as of
Limitation: Ramp's adoption and spending figures do not isolate tokenmaxxing's revenue contribution to each provider [e1].
Source: Breaking: bad news for three of the biggest IPOs in history
Assessment: UnresolvedSaid
Causal claim
“which means less revenue for Anthropic and others. Or lower prices and even more challenges in being profitable.”
Our reading: The reported enterprise swing away from tokenmaxxing means less revenue for Anthropic and others, or else lower prices and even more challenges in being profitable.
No horizon Evidence as of
Limitation: The report does not connect customer usage changes to attributable financial effects at Anthropic or other providers [e1].
Source: Breaking: bad news for three of the biggest IPOs in history
Assessment: UnresolvedSaid
Conditional forecast
“In which cases those stocks will eventually fall.”
Our reading: If enough customers have second thoughts, so that none of the three IPO'ing companies ever hits its long-term projections, those companies' stocks will eventually fall.
No deadline stated Evidence as of
Only if: if enough customers have second thoughts
Source: If enough other companies report the same, the bubble pops. 🫧
Assessment: ContradictedSaid
Causal claim
“The wildest stat I saw this morning was this: Nvidia is spending so much on circular financing its cash flow is headed towards zero.”
Our reading: Nvidia is spending so much on circular financing that its cash flow is headed towards zero.
No horizon Evidence as of
Limitation: This conclusion uses the ordinary meaning of circular financing as funding customers or partners.
Source: Checking the math behind OpenAI and Anthropic’s latest headlines
Assessment: SupportedSaid
Observation
“The other big news per a scoop from Berber Jin at the WSJ, is that Anthropic is projecting its first (slightly) profitable quarter ever.”
Our reading: Anthropic is projecting its first, slightly profitable quarter ever, per the WSJ.
No horizon Evidence as of
Limitation: The assessment relies on a full-text secondary account because the original WSJ report was inaccessible.
Source: Checking the math behind OpenAI and Anthropic’s latest headlines
Assessment: SupportedSaid
Forecast
“In my Politico “Black Swan” prediction for 2026 year, from January I made the seemingly wild prediction that,“By the end of 2026, President Trump will have begun to distance himself from the aggressively pro-AI industry policies that characterized his AI strategy in 2025””
Our reading: By the end of 2026, President Trump will have begun distancing himself from the aggressively pro-AI-industry policies of his 2025 AI strategy.
Settled before its deadline Evidence as of
Limitation: The outcome depends on treating administration implementation under a presidential order as presidential distancing.
Assessment: SupportedSaid
Observation
“As I am sure you have heard by now, the backlash against AI i keep nattering on about is growing fast.”
Our reading: Public backlash against AI is growing fast as of May 2026.
No horizon Evidence as of
Limitation: The polls establish year-over-year changes in US sentiment, not a worldwide trend or a precise growth rate during May. [e1, e2]
Assessment: UnresolvedSaid
Observation
“Every time I think about the numbers and lack of RoI for most pilot studies and the reliability problems I think so.”
Our reading: Most AI pilot studies show no return on investment, and AI has ongoing reliability problems.
No horizon Evidence as of
Limitation: No cited study provides representative pilot-level counts with comparable financial costs and returns for January–March 2026. [e1, e2, e3]
Assessment: SupportedSaid
Observation
“The US has 1,200 AI bills (roughly 150 enacted into law), but nothing that feels like a coherent AI policy.”
Our reading: The US has 1,200 AI bills, roughly 150 of them enacted into law, but nothing that feels like a coherent AI policy.
No horizon Evidence as of
Limitation: The verdict depends on reading the abbreviated post with its linked context.
Source: US AI policy is a clumsy mess. Here’s what to do about it.
Assessment: SupportedSaid
Observation
“Additionally, in the new study 89.4% of agents showed drift relative to their goals after about 30 steps in their process, and 94% of agents with some form of memory-augmentation were vulnerable to poisoning attacks.”
Our reading: In the new study, 89.4% of agents drifted from goals after about 30 steps, and 94% of memory-augmented agents were vulnerable to poisoning.
No horizon Evidence as of
Limitation: The primary paper was not found. The assessment rests on news reports.
Assessment: SupportedSaid
Observation
“A new study from researchers from Stanford, MIT CSAIL, Carnegie Mellon, ITU Copenhagen, and NVIDIA and Elloe AI Labs, examining 847 autonomous agent deployments drawn from healthcare, finance, customer service and code-generation, showed that 91% were vulnerable to subtle but dangerous tool-chaining attacks (see diagram below),”
Our reading: A new study found 91% of 847 autonomous agent deployments were vulnerable to tool-chaining attacks.
No horizon Evidence as of
Limitation: The primary paper was not found online. The assessment rests on news reports.
Assessment: SupportedSaid
Observation
“The new paper very much amplifies what a team of AWS and Berkeley researchers showed in February, documenting similar vulnerabilities:”
Our reading: The new findings on agent vulnerabilities reinforce similar vulnerabilities documented by AWS and Berkeley researchers in February 2026.
No horizon Evidence as of
Limitation: Marcus does not name the February paper; it is identified as STAC by inference.
Assessment: SupportedSaid
Observation
“It’s already splitting both parties.”
Our reading: Frustration with AI is already splitting both parties.
No horizon Evidence as of
Limitation: These examples do not establish party-wide realignment, formal fractures, or electoral consequences.
Source: The growing AI backlash
Assessment: UnresolvedSaid
Observation
“Practically every day I see new signs that the backlash against AI is growing.”
Our reading: Public backlash against AI is growing, with new signs appearing nearly every day.
No horizon Evidence as of
Limitation: The result depends partly on how much weight is given to practically every day.
Source: The growing AI backlash
Assessment: UnresolvedSaid
Causal claim
“Some of this is technical (LLMs aren’t reliable), some of it is political/economic (such as the utter lack of responsible regulation).”
Our reading: Some of generative AI's downsides have technical causes, such as LLMs being unreliable, and some have political or economic causes, such as the utter lack of responsible regulation.
No horizon Evidence as of
Limitation: One adjudicated incident cannot measure the general reliability of all LLMs.
Source: The growing AI backlash
Assessment: UnresolvedSaid
Causal claim
“GenAI has been undermining secondary and college education, opening up mass surveillance, increasing disinformation, delusions, impersonation, phishing, and other forms of cybercrime, nonconsensual deep fake porn, bias in employment and other domains, and economic disparity, drowning the world in slop and unwanted, over-leveraged environment-damaging data centers that risk causing a recession.”
Our reading: Generative AI has been undermining secondary and college education, opening up mass surveillance, increasing disinformation, delusions, impersonation, phishing and other cybercrime, nonconsensual deepfake porn, bias in employment and other domains, and economic disparity, and drowning the world in slop and in unwanted, over-leveraged, environment-damaging data centers that risk causing a recession.
No horizon Evidence as of
Limitation: The search and verified evidence do not cover every listed harm adequately.
Source: The growing AI backlash
Assessment: SupportedSaid
Observation
“The problem of course is that LLMs are mimics, and what they say isn’t always true.”
Our reading: LLMs are mimics and what they say is not always true.
No horizon Evidence as of
Limitation: The support for 'mimics' uses the ordinary behavioral reading specified in the contract.
Assessment: SupportedSaid
Observation
“Realism re AI coding is knowing that next-word prediction gets us a surprisingly long way in writing code, but less far in making sure that code is robust.”
Our reading: Next-word prediction gets us a surprisingly long way in writing code, but less far in making sure that code is robust.
No horizon Evidence as of
Limitation: The April results come from a security vendor’s own research announcement; commercial interests warrant caution, and the packet provides no independent replication [e1].
Assessment: SupportedSaid
Observation
“But that’s exactly why we still need software engineers — and why what Amodei recently said about software engineers disappearing is so absurd.”
Our reading: Software engineers are still needed, and Amodei's claim that they will disappear is absurd.
No horizon Evidence as of
Limitation: The decisive occupational count is secondary evidence.
Source: Dario Amodei, hype, AI safety, and the explosion of vibe-coded AI disasters
Assessment: SupportedSaid
Observation
“AI coding tools are genuinely revolutionizing the software industry, but more than almost anything I have ever seen, they fall into the category of things you maybe shouldn’t try at home without significant prior experience:”
Our reading: AI coding tools are genuinely revolutionizing the software industry.
No horizon Evidence as of
Limitation: The reporting period precedes the post by several months.
Source: Dario Amodei, hype, AI safety, and the explosion of vibe-coded AI disasters
Assessment: UnresolvedSaid
Forecast
“And I am sure we will see many more catastrophic failures soon enough, involving things like privacy, security and data loss. Already there have been many.”
Our reading: Many more catastrophic vibe-coding failures involving privacy, security and data loss will occur soon.
No deadline stated Evidence as of
Limitation: Public incident reporting is incomplete.
Source: Dario Amodei, hype, AI safety, and the explosion of vibe-coded AI disasters
Assessment: UnresolvedSaid
Observation
“Here is a further example of my last point that ChatGPT’s exciting new image system has trouble with functional understanding, sent to me by a reader:”
Our reading: ChatGPT's exciting new image system has trouble with functional understanding.
No horizon Evidence as of
Limitation: The assessment concerns observable functional depictions, not whether the model possesses human-like understanding.
Assessment: UnresolvedSaid
Observation
“But if you look closely, there are several errors, and those errors are revealing.”
Our reading: ChatGPT's new image engine's labeled bike diagram contains several revealing labeling errors.
No horizon Evidence as of
Limitation: This is an evidence-access limitation, not a finding that the labels are correct.
Assessment: UnresolvedSaid
Observation
“But anybody knowledgeable about bikes (racers, mechanics, designers etc) would immediately see numerous problems.”
Our reading: ChatGPT's tandem bike image contains numerous problems that bike experts would immediately spot.
No horizon Evidence as of
Limitation: The packet contains no inspectable copy of the tandem image or qualified independent assessment.
Assessment: SupportedSaid
Observation
“That’s still true – and it very much applies in medicine.”
Our reading: Large language models are still frequently wrong yet confident, producing authoritative bullshit that shouldn't be trusted, and this applies strongly in medicine.
No horizon Evidence as of
Limitation: Support depends on the contextual evidence-period reading.
Assessment: SupportedSaid
Observation
“All of this – the hallucinations, mistakes, and overconfidence – is entirely typical of LLMs, and entirely problematic in medicine.”
Our reading: Hallucinations, mistakes and overconfidence are entirely typical of LLMs and entirely problematic in medicine.
No horizon Evidence as of
Limitation: The decisive audit uses misinformation-oriented prompts and older model versions.
Assessment: SupportedSaid
Observation
“Because a large fraction of the population has begun to turn to chatbots for medical advice.”
Our reading: A large fraction of the population has begun using chatbots for medical advice.
No horizon Evidence as of
Limitation: The U.S.-adult population is an interpretation, not an explicit commitment by Marcus.
Assessment: SupportedSaid
Causal claim
“we should not take the apparent “confidence” of large language models — itself an illusion of how they are trained — to mean”
Our reading: The apparent confidence of LLMs is an illusion that comes from how they are trained.
No horizon Evidence as of
Limitation: The result depends on reading 'illusion' as misleading certainty rather than complete absence of information.
Assessment: UnresolvedSaid
Observation
“And all this from a company thought (or claimed) that GPT-5 was going to be tantamount to AGI (spoiler alert: it wasn’t)”
Our reading: GPT-5 was not tantamount to AGI, although OpenAI thought or claimed that it would be.
No horizon Evidence as of
Limitation: Support for one clause is insufficient to support the whole claim.
Source: Peak absurdity, Part II
Assessment: SupportedSaid
Observation
“Today, even after a very strong week, Oracle stands at just $170.”
Our reading: As of the post date, even after a very strong week, Oracle stock stands at just $170, compared with about $308 on September 11.
No horizon Evidence as of
Limitation: Marcus gives no quotation timestamp; the assessment uses daily closing prices.
Source: Peak absurdity, Part II
Assessment: SupportedSaid
Observation
“On the darker side, it really does arm attackers to a greater degree than Mythos’s predecessors.”
Our reading: Claude Mythos Preview gives cyber attackers more capability than earlier Claude models did.
No horizon Evidence as of
Limitation: The finding concerns controlled offensive testing, not observed criminal deployment.
Source: Claude Mythos, evaluated
Assessment: UnresolvedSaid
Observation
“On the happy side, in its current form, Myth is nowhere near as scary as Tom Fridman (who worries about schoolchildren accidentally taking down power grids) and others made it out to be.”
Our reading: In its current form, Claude Mythos Preview is nowhere near as dangerous as Thomas Friedman and others portrayed it.
No horizon Evidence as of
Limitation: Simulated network compromise and individual exploits do not directly measure disruption of a real power grid.
Source: Claude Mythos, evaluated
Assessment: SupportedSaid
Observation
“Second, they show that a neurosymbolic hybrid model, which integrates both a neural networks (for pattern recognition) and a symbolic planner generalizes far better.”
Our reading: A neurosymbolic hybrid with a symbolic planner generalizes far better than VLAs on the Tower of Hanoi task.
No horizon Evidence as of
Limitation: This establishes the reported result for one simulated task and the tested configurations. It does not establish a general advantage across robotics.
Source: Even more good news for the future of neurosymbolic AI
Assessment: SupportedSaid
Observation
“LLMs have persistent weaknesses around planning, reasoning and generalization.”
Our reading: LLMs have persistent weaknesses in planning, reasoning and generalization.
No horizon Evidence as of
Limitation: Generalization experiments use restricted settings.
Source: Even more good news for the future of neurosymbolic AI
Assessment: UnresolvedSaid
Observation
“LLMs are an efficient way to pattern recognition where perfect results are not required, but an inefficient way to reason a plan.”
Our reading: LLMs are efficient for imperfect pattern recognition but inefficient for reasoning out a plan.
No horizon Evidence as of
Limitation: Different efficiency baselines could produce different judgments.
Source: Even more good news for the future of neurosymbolic AI
Assessment: SupportedSaid
Observation
“Third, they show that the neurosymbolic hybrid was vastly more efficent in terms of energy use - by nearly two orders of magnitude.”
Our reading: The neurosymbolic hybrid used nearly two orders of magnitude less energy than the VLAs in the Tufts study.
No horizon Evidence as of
Limitation: The outcome depends on interpreting the statement as a training comparison.
Source: Even more good news for the future of neurosymbolic AI
Assessment: UnresolvedSaid
Observation
“You can read the full paper by Timothy Duggan, Pierrick Lorang, Hong Lu, and Matthias Scheutz, which very much vindicates the Apple reasoning paper, here.”
Our reading: The Tufts paper strongly vindicates the Apple reasoning paper's findings.
No horizon Evidence as of
Limitation: A weaker reading of vindicates as merely consistent with would receive more support.
Source: Even more good news for the future of neurosymbolic AI
Assessment: UnresolvedSaid
Forecast
“But anybody who thinks LLMs are a direct route to the sort AGI that could fundamentally transform society for the good is kidding themselves.”
Our reading: LLMs are not a direct route to the sort of AGI that could fundamentally transform society for the good, and anyone who thinks they are is kidding themselves.
No deadline stated Evidence as of
Limitation: The forecast's central terms remain underspecified.
Source: Even more good news for the future of neurosymbolic AI
Assessment: SupportedSaid
Observation
“Claude Code (in particular version 4.6), which does seem genuinely impressive to me and a number of friends I trust, still makes plenty of mistakes, and still can’t be trusted.”
Our reading: Claude Code (in particular version 4.6), which seems genuinely impressive to Marcus and a number of friends he trusts, still makes plenty of mistakes and still can't be trusted.
No horizon Evidence as of
Limitation: The controlled evaluation covers one programming language and a particular task suite.
Source: Even more good news for the future of neurosymbolic AI
Assessment: UnresolvedSaid
Observation
“It’s obviously not AGI.”
Our reading: Claude Code is not AGI.
No horizon Evidence as of
Limitation: This is an evidentiary finding, not a finding that Claude Code was AGI.
Source: Even more good news for the future of neurosymbolic AI
Assessment: SupportedSaid
Observation
“Putting things differently, Anthropic, when push came to shove, went exactly where I have said for 25 years that the field needed to go: to Neurosymbolic AI.”
Our reading: Anthropic adopted neurosymbolic AI in building Claude Code.
No horizon Evidence as of
Limitation: The result depends on Marcus's broad contextual definition of neurosymbolic AI.
Assessment: UnresolvedSaid
Observation
“The source code leak proves it. Tucked away at its center is a 3,167 line kernel called print.ts.”
Our reading: The leaked Claude Code source shows a 3,167-line kernel called print.ts at its center, proving it is not a pure LLM.
No horizon Evidence as of
Limitation: A longer file alone does not disprove a shorter function inside it.
Assessment: SupportedSaid
Observation
“And the thing is, Claude Code is NOT a pure LLM. And it’s not pure deep learning. Not even close.”
Our reading: Claude Code is not a pure LLM and not pure deep learning.
No horizon Evidence as of
Limitation: The independent study is a preprint.
Assessment: UnresolvedSaid
Forecast
“The first part has already come to pass. In time, other three will, too.”
Our reading: The other three parts of Marcus's 2020 agenda (knowledge, reasoning, world models) will eventually be adopted, as neurosymbolic AI already has been.
No deadline stated Evidence as of
Limitation: The terms are vague, and current industry uses of 'reasoning' and 'world models' may not match the paper's meaning.
Assessment: UnresolvedSaid
Observation
“The demo was definitely proof of concept that we need to get our regulatory and technical house in order, but not the immediate threat the media and public was lead to believe.”
Our reading: The Mythos demo was not the immediate threat the media and public were led to believe.
No horizon Evidence as of
Limitation: The principal technical evidence is Anthropic's own disclosure.
Source: Three reasons to think that the Claude Mythos announcement from Anthropic was overblown
Assessment: UnresolvedSaid
Observation
“The model itself is incrementally better than previous recent models, but certainly not an off-the-chart breakthrough:”
Our reading: Mythos is incrementally better than recent previous models and certainly not an off-the-chart breakthrough.
No horizon Evidence as of
Limitation: The original Mythos system card could not be read through the required tool.
Source: Three reasons to think that the Claude Mythos announcement from Anthropic was overblown
Assessment: UnresolvedSaid
Observation
“ChatGPT can’t reliably run a timer but it has still been implicated in delusions, suicides, cognitive surrender, mass disinformation, and so much more.”
Our reading: ChatGPT cannot reliably run a timer, yet it has been implicated in delusions, suicides, cognitive surrender and mass disinformation.
No horizon Evidence as of
Limitation: Legal allegations establish implication, not proven causation.
Source: What should we take from Anthropic’s (possibly) terrifying new report on Mythos?
Assessment: SupportedSaid
Observation
“Right now he is in trouble. People are increasingly concerned about the economics of his company,”
Our reading: As of April 2026, Sam Altman is in trouble, with people increasingly concerned about OpenAI's economics.
No horizon Evidence as of
Limitation: The result depends on reading in trouble as facing pressure and scrutiny.
Assessment: SupportedSaid
Observation
“Third, there is the well-known tendency of generative AI to be sycophantic; to tell users that their ideas were the greatest ever, even when they are not.”
Our reading: Generative AI tends to be sycophantic, telling users their ideas are excellent even when they are not.
No horizon Evidence as of
Limitation: The decisive March evidence is an independent benchmark, not a peer-reviewed representative survey.
Source: In the Iran war, it looks like AI helped with operations, not strategy
Assessment: UnresolvedSaid
Causal claim
“First, strategy, more than e.g., writing requisition memos, requires a broad, deep understanding of the world, but generative AI has never had robust models of the world.”
Our reading: Generative AI has never had robust world models, so it is poorly suited to strategy, which requires a broad, deep understanding of the world.
No horizon Evidence as of
Limitation: The research concerns restricted experimental domains.
Source: In the Iran war, it looks like AI helped with operations, not strategy
Assessment: UnresolvedSaid
Observation
“AI is fine for writing memos; nobody should rely on it to plan a war or to guess its likely outcomes.”
Our reading: AI is adequate for writing memos.
No horizon Evidence as of
Limitation: The selected latest-month period prevents older studies from serving as decisive outcome evidence.
Source: In the Iran war, it looks like AI helped with operations, not strategy
Assessment: SupportedSaid
Observation
“This study reinforces what Anh Totti Nguyen has been saying for a long time, in a series of underappreciated papers like Vision Language Models are Blind that I keep trying to draw attention to.”
Our reading: The new Stanford study reinforces what Anh Totti Nguyen has long been saying in a series of underappreciated papers, such as 'Vision Language Models are Blind'.
No horizon Evidence as of
Limitation: The studies test different model versions, tasks, and failure mechanisms.
Source: The mirage of visual understanding in current frontier models
Assessment: UnresolvedSaid
Observation
“AGI this stuff ain’t.”
Our reading: The frontier models described in the study are not AGI.
No horizon Evidence as of
Limitation: The result depends on how broadly competence must extend and what shortfalls remain compatible with AGI.
Source: The mirage of visual understanding in current frontier models
Assessment: UnresolvedSaid
Observation
“Another dramatic sign of changing times: Sam Altman, who ridiculed my 2022 critique of LLMs that argued that scaling would not bring us to AGI and that we would need architectures has just argued that”
Our reading: Sam Altman, who ridiculed Marcus's 2022 critique arguing that scaling would not bring us to AGI, has just said he bets there is another new architecture to find that would be as big a gain as transformers were over LSTMs, which Marcus calls a dramatic sign of changing times.
No horizon Evidence as of
Limitation: The original post does not identify Marcus or his particular 2022 critique; the attribution rests on secondary commentary [e1, e2].
Source: BREAKING: Sam Altman concedes that we need major breakthroughs beyond mere scaling to get to AGI
Assessment: MixedSaid
Observation
“Hassabis is no longer on board either, and nor are Sutskever and LeCun. Nadella and Pichai have also hinted at skepticism around scaling.”
Our reading: Hassabis, Sutskever and LeCun no longer believe in scaling as the route to AGI, and Nadella and Pichai have hinted at skepticism about it.
No horizon Evidence as of
Limitation: The record does not establish earlier endorsement and subsequent rejection of scaling alone for everyone described as having changed position.
Source: BREAKING: Sam Altman concedes that we need major breakthroughs beyond mere scaling to get to AGI
Assessment: UnresolvedSaid
Observation
“Note that in this talk Altman didn’t claim to have found such an architecture. That represents a significant retrenchment from his claim fourteen months ago that “We now know how to build AGI as it’s usually understood.””
Our reading: In this talk Altman did not claim to have found such an architecture, which Marcus says is a significant retrenchment from Altman's claim fourteen months earlier that 'We now know how to build AGI as it's usually understood.'
No horizon Evidence as of
Limitation: The contemporaneous news report is not a complete transcript.
Source: BREAKING: Sam Altman concedes that we need major breakthroughs beyond mere scaling to get to AGI
Assessment: UnresolvedSaid
Observation
“AI is supposed to change all that. But so far it hasn’t.”
Our reading: So far, AI has not changed the toll that cancer takes on people.
No horizon Evidence as of
Limitation: The claim gives no magnitude, geography, or precise outcome.
Source: F Cancer
Assessment: UnresolvedSaid
Causal claim
“Over and over we discover drugs with AI that work on trials in animals but not safely enough in humans; we just don’t understand biology well enough.”
Our reading: Over and over, drugs discovered with AI work in animal trials but are not safe enough in humans, because we just don't understand biology well enough.
No horizon Evidence as of
Limitation: The packet lacks candidate-level records establishing the complete claimed sequence and its biological cause.
Source: F Cancer
Assessment: SupportedSaid
Observation
“In my 2020 article, The Next Decade in AI, I foresaw all this, and urged the field to start focusing on world (cognitive) models and neurosymbolic AI.”
Our reading: In his 2020 article 'The Next Decade in AI', Marcus foresaw these developments and urged the field to start focusing on world (cognitive) models and neurosymbolic AI.
No horizon Evidence as of
Limitation: Support depends on interpreting 'foresaw all this' as a general warning, rather than a specific prediction of the 2026 Meta and xAI developments.
Source: BREAKING: Expensive new evidence that scaling is not all you need
Assessment: SupportedSaid
Observation
“the fact that LLMs hallucinate (patently true)”
Our reading: LLMs hallucinate.
No horizon Evidence as of
Limitation: The evidence is a manufacturer's internal evaluation.
Assessment: SupportedSaid
Observation
“It is a misapprehension of LLMs to confuse the things they say with what is true.”
Our reading: What LLMs say should not be taken as true just because they say it; treating their statements as truth misunderstands them.
No horizon Evidence as of
Limitation: The benchmark uses one person's biography and that person's manual judgments; it cannot establish a general error rate.
Assessment: SupportedSaid
Observation
“Essentially the same argument applies to Claude’s constitution, which amounts to a set of built-in guardrails. OpenAI has those, too!”
Our reading: Claude's constitution is basically a set of built-in guardrails, and OpenAI's models have guardrails too.
No horizon Evidence as of
Limitation: The outcome depends on reading guardrails broadly and amounts to as shorthand.
Assessment: UnresolvedSaid
Observation
“Apocalyptic talk is a special case of hype, Sam and Dario both play that card over and over, basically amounting to saying “our tech is so scary it might ruin society, so give us more money so we can make that tech ever faster.””
Our reading: Altman and Amodei both repeatedly use apocalyptic talk, a special case of hype, which basically amounts to saying their tech is so scary it might ruin society so they should be given more money to build it ever faster.
No horizon Evidence as of
Limitation: Hype is an opinion rather than a separately measurable fact.
Assessment: SupportedSaid
Observation
“How about copyright? OpenAI notoriously trained on practically anything that is “publicly available” whether public domain or not.”
Our reading: OpenAI trained on practically anything publicly available, whether public domain or not.
No horizon Evidence as of
Limitation: Support applies to broad use including copyrighted material, not a numerical interpretation of “practically anything.”
Assessment: UnresolvedSaid
Forecast
“AI can help in the search for candidate drugs, but we still need to test them. There is zero chance that we double human life span in the next decade.”
Our reading: There is zero chance that human life span is doubled in the next decade.
Not yet due Evidence as of
Limitation: The lifespan measure, population and March 2026 baseline remain unspecified.
Assessment: UnresolvedSaid
Observation
“Both, for example, have endlessly hinted that AGI is near, and just quietly moved the deadline back when they didn’t deliver.”
Our reading: Amodei and Altman have repeatedly hinted AGI is near and quietly moved the deadline back when it didn't arrive.
No horizon Evidence as of
Limitation: Evidence for individual components does not settle the compound claim.
Assessment: UnresolvedSaid
Observation
“Millions of people, literally, are angry; many feel betrayed.”
Our reading: Millions of people are angry and many feel betrayed by Altman/OpenAI.
No horizon Evidence as of
Limitation: Anger and betrayal are not measured directly. Boycott sign-ups and uninstalls are proxies that may reflect other motives.
Source: BREAKING: Sam Altman’s greed and dishonesty are finally catching up to him
Assessment: SupportedSaid
Observation
“The public anger on social media is growing fast. A major boycott is afoot:”
Our reading: Public anger at OpenAI on social media is growing fast and a major boycott is underway.
No horizon Evidence as of
Limitation: No platform-wide time series establishes the precise growth rate of angry social-media activity.
Source: BREAKING: Sam Altman’s greed and dishonesty are finally catching up to him
Assessment: SupportedSaid
Observation
“Using a system you shouldn’t trust with your taxes to blow people up. 2026 in a nutshell.”
Our reading: Generative AI, a system that people shouldn't trust with their taxes, is being used by certain parts of Washington to blow people up.
No horizon Evidence as of
Limitation: The result depends on reading the military phrase as operational support rather than autonomous firing.
Source: Don’t trust Generative AI to do your taxes — and don’t trust it with people’s lives
Assessment: UnresolvedSaid
Observation
“Essentially anyone who uses a chatbot is at risk.”
Our reading: Essentially anyone who uses a chatbot is at risk of having their beliefs distorted by sycophantic AI.
No horizon Evidence as of
Limitation: Treating essentially anyone as broad population coverage materially affects the assessment; a weaker claim that ordinary users can be affected is better supported.
Source: Breaking: “sycophantic AI distorts belief, manufacturing certainty where there should be doubt”
Assessment: UnresolvedSaid
Causal claim
“It has made dodgy companies that gave lip service to AI safety tremendously wealthy —on the promise, not yet delivered—that they could create AGI.”
Our reading: Doomer hype has made dodgy companies that gave lip service to AI safety tremendously wealthy on the not-yet-delivered promise that they could create AGI.
No horizon Evidence as of
Limitation: Financing and valuation are distinct measures and are used only as context.
Assessment: UnresolvedSaid
Causal claim
“By proclaiming that LLMs would change the world, and that they would do so imminently, doomers contributed to the growth of the LLM companies (essentially free publicity) and to the White House thinking LLMs were critical to the survival of the nation.”
Our reading: By proclaiming that LLMs would change the world imminently, doomers contributed to the growth of LLM companies and to the White House thinking LLMs were critical to the nation's survival.
No horizon Evidence as of
Limitation: Public explanations may omit influences present in private deliberations.
Assessment: UnresolvedSaid
Causal claim
“the point being that doomers, especially those who have screamed that AGI is nigh, have wanted to slow AI acceleration (at least until we could make AI safe) and instead have only accelerated it.”
Our reading: Doomers, especially those who claimed AGI is nigh, wanted to slow AI acceleration (at least until AI could be made safe) but instead have only accelerated it.
No horizon Evidence as of
Limitation: Safety agreements are not measurements of development speed.
Assessment: SupportedSaid
Observation
“And of course Generative AI still has problems with common sense, as an endless of stream of examples have shown, this one involving circulating on X today:”
Our reading: Generative AI still has problems with common sense.
No horizon Evidence as of
Limitation: The main systematic test comes from a commercial vendor.
Assessment: UnresolvedSaid
Observation
“Worse, they did it in favor of someone else who took pretty similar terms and happened to have given more campaign contributions.”
Our reading: The government acted in favor of someone else (the OpenAI/Altman side) who took pretty similar terms to Anthropic's and happened to have given more campaign contributions.
No horizon Evidence as of
Limitation: Executives' personal donations are not corporate contributions.
Source: The whole thing was a scam
Assessment: SupportedSaid
Observation
“It’s one thing for the government to reject Anthropic’s terms—and entirely another to banish them permanently and, absurdly and punitively declare them a supply chain risk.”
Our reading: The US government permanently banished Anthropic and absurdly and punitively declared it a supply chain risk, going beyond merely rejecting its terms.
No horizon Evidence as of
Limitation: Formal designation followed in March.
Source: The whole thing was a scam
Assessment: SupportedSaid
Observation
“The company ripped off a lot of writer’s work (per the $1.5B settlement), and recently walked back its core safety pledge.”
Our reading: Anthropic ripped off a lot of writers' work, as shown by its $1.5B settlement.
No horizon Evidence as of
Limitation: The assessment reads 'ripped off' as unauthorized acquisition, not unlawful training.
Source: The whole thing was a scam