Claims that can’t be checked by Gary Marcus, page 6 of 24
Deadlines are judged as of 2026-10-29. About Gary Marcus and the sources covered.
Claims that can’t be checked by Gary Marcus (2362)
Public evidence cannot settle these claims, so they are recorded and attributed, not scored. Most were screened out: both AI providers read the claim on its own and found it a value judgment, rhetoric, hedged, too vague to test, or about something no public evidence could show. The rest were found to be recommendations or too open to test in a full review.
Page 6 of 24: claims 501 to 600 of 2362, ordered by the date first said, newest first. To find a claim by its words, use search.
Assessment: Can’t be checkedSaid
Forecast
“I doubt their runway would last much more than a year.”
Our reading: OpenAI's runway is unlikely to last much more than a year.
No horizon
Qualifier: if venture capital dries up (implied)
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Hopefully the US government will have the good sense not to bite; Google and Anthropic are doing just fine, without government handouts.”
Our reading: The US government should not bail out OpenAI, since Google and Anthropic are doing fine without handouts.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“That can’t be good. A zillionaire hellbent on winning the AI race and no longer interested in AI safety is not what the world needs.”
Our reading: A zillionaire, Musk at xAI, who is hellbent on winning the AI race and no longer interested in AI safety is not what the world needs, and that can't be good.
No horizon
Scope: referring to Musk after xAI's reported gutting of its safety team
Assessment: Can’t be checkedSaid
Rhetoric
“It would indeed be one the great ironies of history if one of the biggest IP thieves in history was taken down by mass IP theft.”
Our reading: It would be one of the great ironies of history if OpenAI, one of the biggest IP thieves in history, were taken down by mass IP theft.
No horizon
Qualifier: if one of the biggest IP thieves in history were taken down by mass IP theft
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“It would be nice if places like FT would refuse to play along, providing more context (about conflicts of interest and past prediction track records). It would also help if they would more regularly seek independent outside opinions.”
Our reading: Media outlets like the FT should give context on conflicts of interest and track records when reporting AI executives' predictions, and seek independent opinions.
No horizon
Source: Promises are cheap
Assessment: Can’t be checkedSaid
Causal claim
“Tesla trades at nearly 400 times earnings not because they have ever made all that much money, but because Elon is a master of hype.”
Our reading: Tesla's valuation of nearly 400 times earnings is due to Musk's hype rather than profits.
No horizon Evidence as of
Limitation: The P/E figure comes from one financial press article (e1), not from audited data.
Source: Promises are cheap
Assessment: Can’t be checkedSaid
Rhetoric
“Promises are cheap. And all the big tech CEOs have come to realize that making them costs them nothing.”
Our reading: All the big tech CEOs have come to realize that making promises costs them nothing.
No horizon
Scope: all the big tech CEOs
Source: Promises are cheap
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“The public has not been well served by this practice.”
Our reading: The public has not been well served by media companies' practice of platforming predictions from people who stand to gain from their narratives with too little skepticism.
No horizon
Source: Promises are cheap
Assessment: Can’t be checkedSaid
Conditional forecast
“If the whole thing collapses, as many people fear, the careless repetition of hype by editors who love the “everything is about to change” narrative (always good for clicks) may well turn out to have been an important contributing factor.”
Our reading: If the AI boom collapses, careless media repetition of hype may turn out to have been an important contributing factor.
No horizon
Only if: if the whole thing (AI boom) collapses
Source: Promises are cheap
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Instead, too often media companies platform predictions from people who stand to gain immensely from their narratives, with too little skepticism.”
Our reading: Too often, media companies platform predictions from people who stand to gain immensely from their narratives, with too little skepticism.
No horizon
Qualifier: too often
Source: Promises are cheap
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“A lot of people may have taken his post seriously, but they shouldn’t have.”
Our reading: People should not have taken Shumer's viral post about AI seriously.
No horizon
Source: About that Matt Shumer post that has nearly 50 million views
Assessment: Can’t be checkedSaid
Rhetoric
“The picture he sells just isn’t realistic, however much people might wish it were true.”
Our reading: The picture Shumer presents isn't realistic, however much people might wish it were true.
No horizon
Qualifier: however much people might wish it were true
Source: About that Matt Shumer post that has nearly 50 million views
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Silicon Valley can do one of two things. It can continue to dismiss critics and criticism, praying for magical solutions, or it can face reality and start focusing on alternatives to LLMs.”
Our reading: The AI industry should face reality and start focusing on alternatives to LLMs instead of dismissing critics.
No horizon
Source: BREAKING: LLM “reasoning” continues to be deeply flawed
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“If you want a full idea of how far we have left to go (and, implicitly, why LLMs aren’t the right answer) read it.”
Our reading: The Caltech/Stanford review of LLM reasoning failures shows how far AI still has to go and, implicitly, why LLMs aren't the right answer.
No horizon
Qualifier: implicitly, as conveyed by the review
Source: BREAKING: LLM “reasoning” continues to be deeply flawed
Assessment: Can’t be checkedSaid
Forecast
“But it’s not just that a trillion dollars or more might go down the drain, but that there might be considerable collateral damage, to the rest of society, both economic and otherwise (e.g., in terms of how LLMs have undermined college education).”
Our reading: A trillion dollars or more invested in AI might be lost, with possible considerable economic and other collateral damage to society.
No horizon
Source: Some disconcerting facts about AI and banking that may have profound consequences
Assessment: Can’t be checkedSaid
Forecast
“But the unreliable statistical mimicry of Generative AI is unlikely to be the ticket to those massive productivity gain.”
Our reading: Generative AI, being unreliable statistical mimicry, is unlikely to deliver the anticipated massive productivity gains.
No horizon
Source: Some disconcerting facts about AI and banking that may have profound consequences
Assessment: Can’t be checkedSaid
Forecast
“Instead, the misguided obsession with large language models will likely cause a lot of pain over the next few years, economically and socially, until a new, far more reliable form of AI is developed.”
Our reading: The obsession with LLMs will likely cause substantial economic and social pain over the next few years, until a far more reliable form of AI is developed.
No horizon
Qualifier: until a new, far more reliable form of AI is developed
Source: Some disconcerting facts about AI and banking that may have profound consequences
Assessment: Can’t be checkedSaid
Forecast
“It was already clear by then that generative AI was probably going to lose a lot of money.”
Our reading: Generative AI is probably going to lose a lot of money.
No horizon
Source: Some disconcerting facts about AI and banking that may have profound consequences
Assessment: Can’t be checkedSaid
Conditional forecast
“If this deal collapses, possibly bringing down OpenAI, the multitrillion dollar expansion that so many were anticipating might well not happen at all.”
Our reading: If the Nvidia–OpenAI deal collapses, the anticipated multitrillion-dollar AI expansion might well not happen at all.
No horizon
Only if: if the Nvidia–OpenAI deal collapses
Assessment: Can’t be checkedSaid
Conditional forecast
“If this deal collapses, possibly bringing down OpenAI, the multitrillion dollar expansion that so many were anticipating might well not happen at all.”
Our reading: If the Nvidia–OpenAI deal collapses, it could possibly bring down OpenAI.
No horizon
Only if: if the Nvidia–OpenAI deal collapses
Assessment: Can’t be checkedSaid
Forecast
“The best news here is that with LLM mania subsiding, there might actually be a chance for newcomers to try new things.”
Our reading: With LLM mania subsiding, there might actually be a chance for newcomers to try new things.
No horizon
Qualifier: with LLM mania subsiding (stated as already happening)
Source: Sam Altman and the day Nvidia’s meteoric rise came to an end
Assessment: Can’t be checkedSaid
Observation
“That circularity is a warning sign, reflecting a market that is propped up rather than functioning well on its own accord.”
Our reading: Circular financing in AI is a warning sign that the market is propped up rather than healthy.
No horizon Evidence as of
Limitation: The single evidence item is one analyst's view, cited only as context. It neither confirms nor refutes the characterization.
Source: Sam Altman and the day Nvidia’s meteoric rise came to an end
Assessment: Can’t be checkedSaid
Observation
“It turns out this was authoritative bullshit. It sounded true and was said with confidence, but it was purely a bluff.”
Our reading: Altman's January 2025 claim that OpenAI was confident it knew how to build AGI was a bluff.
No horizon
Source: Sam Altman and the day Nvidia’s meteoric rise came to an end
Assessment: Can’t be checkedSaid
Forecast
“My guess is that these stocks – and the reputation of OpenAI – will fall further, but either way it is already clear that the rockets will not reach the altitude so many people were hoping for.”
Our reading: AI tech stocks such as Nvidia, CoreWeave and Oracle, and OpenAI's reputation, will probably fall further.
No horizon
Qualifier: 'my guess'
Source: Sam Altman and the day Nvidia’s meteoric rise came to an end
Assessment: Can’t be checkedSaid
Observation
“At last the market may be both ready—and desperate—for real AI.”
Our reading: The market may now be ready, and desperate, for AI approaches beyond LLM hype.
No horizon
Source: Sam Altman and the day Nvidia’s meteoric rise came to an end
Assessment: Can’t be checkedSaid
Causal claim
“Aside from Nvidia’s Jensen Huang, a brilliant and impressive CEO with considerable foresight, nobody was more responsible for Nvidia’s rise than Sam Altman, the relentless showman who runs OpenAI — and who repeatedly promised the moon.”
Our reading: Apart from Jensen Huang, nobody was more responsible for Nvidia's rise than Sam Altman, the showman who runs OpenAI and who repeatedly promised the moon.
No horizon
Qualifier: aside from Jensen Huang
Source: Sam Altman and the day Nvidia’s meteoric rise came to an end
Assessment: Can’t be checkedSaid
Observation
“Altman didn’t actually know how to build AGI [artificial general intelligence] then, and probably still doesn’t.”
Our reading: Altman did not know how to build AGI in January 2025 and probably still does not as of February 2026.
No horizon
Qualifier: 'probably' for the present
Source: Sam Altman and the day Nvidia’s meteoric rise came to an end
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“If I owned SpaceX shares I would feel like Elon way overpaid for his own company.”
Our reading: Marcus says that if he owned SpaceX shares he would feel that Musk way overpaid for xAI, his own company, in the merger.
No horizon
Qualifier: if I owned SpaceX shares
Assessment: Can’t be checkedSaid
Forecast
“If you even believe at all that data centers in space are really going to be a big, profitable thing in the next decade. (Count me skeptical, though the economics of such things is not my area of expertise.)”
Our reading: Marcus is skeptical that data centers in space will be a big, profitable thing in the next decade, though he says the economics is not his area of expertise.
No horizon
Qualifier: author says the economics is not his area of expertise
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“In contrast, trading xAI at a thousand times revenue when they are in an increasingly crowded field where everyone is being forced into price wars seems nuts.”
Our reading: Trading xAI at a thousand times revenue seems nuts when it is in an increasingly crowded field where everyone is being forced into price wars.
No horizon
Qualifier: seems
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Hopefully back in the real world the good folks at SpaceX aren’t about to replace their carefully-developed symbolic navigation algorithms with an LLM trained on X. Kaboom.”
Our reading: SpaceX should not replace its symbolic navigation algorithms with an LLM trained on X, which would be disastrous.
No horizon
Assessment: Can’t be checkedSaid
Rhetoric
“Where Apple iPhone applications are carefully sandboxed and appropriately isolated to minimize harm, OpenClaw is basically a weaponized aerosol, in prime position to fuck shit up, if left unfettered.”
Our reading: Unlike sandboxed iPhone apps, OpenClaw left unfettered is positioned to cause serious harm.
No horizon
Qualifier: if left unfettered
Source: OpenClaw (a.k.a. Moltbot) is everywhere all at once, and a disaster waiting to happen
Assessment: Can’t be checkedSaid
Observation
“In essay called LLMs + Coding Agents = Security Nightmare, we talked about how LLMs, which mimic human text (and even human-written code) but understand what that they produce only superficially, can easily be tricked.”
Our reading: LLMs understand their output only superficially and can easily be tricked, e.g., by prompt injection.
No horizon Evidence as of
Limitation: This disposition does not reject the separately testable claim that LLM agents can be manipulated.
Source: OpenClaw (a.k.a. Moltbot) is everywhere all at once, and a disaster waiting to happen
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“But in this case, the advice is clear and simple: if you care about the security of your device or the privacy of your data, don’t use OpenClaw. Period.”
Our reading: People who care about device security or data privacy should not use OpenClaw.
No horizon
Qualifier: if you care about device security or data privacy
Source: OpenClaw (a.k.a. Moltbot) is everywhere all at once, and a disaster waiting to happen
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“(Bonus advice: if your friend has OpenClaw installed, don’t use their machine. Any password you type there might be vulnerable, too.”
Our reading: If your friend has OpenClaw installed, you should not use their machine, because any password you type there might be vulnerable.
No horizon
Scope: if your friend has OpenClaw installed
Source: OpenClaw (a.k.a. Moltbot) is everywhere all at once, and a disaster waiting to happen
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“First, social media platforms must move away from the “whack-a-mole” approach they currently use.”
Our reading: Social media platforms must abandon episodic whack-a-mole takedowns in favor of continuous monitoring for improbable coordination.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“A defense that only reacts to yesterday’s tricks is destined to fail. We should instead proactively stress-test our defenses using agent-based simulations.”
Our reading: Purely reactive defenses will fail; defenders should proactively stress-test using agent-based simulations of AI swarms.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Finally, we need to end the era of plausible deniability with an AI Influence Observatory. Crucially, this cannot be a government-run “Ministry of Truth.” Instead, it must be a distributed ecosystem of independent academic groups and NGOs.”
Our reading: An independent, non-governmental AI Influence Observatory of academic groups and NGOs should be created to detect coordinated swarms.
No horizon
Qualifier: not government-run
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Currently, platforms restrict access to the data needed to detect these swarms, leaving independent experts blind. Legislation must guarantee vetted academic and civil society researchers free, privacy-preserving access to platform data.”
Our reading: Legislation must guarantee vetted academic and civil society researchers free, privacy-preserving access to platform data.
No horizon
Qualifier: platforms currently restrict access to the data needed to detect these swarms, leaving independent experts blind
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Third, we must make it expensive to be a fake person. Policymakers need to incentivize cryptographic attestations and reputation standards to strengthen provenance.”
Our reading: Policymakers should incentivize cryptographic attestations and reputation standards to make fake personas expensive.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“And see also my 2020, Next Decade in AI for an older but still very relevant perspective on how limits in statistical models militate for the use of neurosymbolic AI and world (aka cognitive) models.”
Our reading: The limits of statistical models argue for using neurosymbolic AI and world (cognitive) models, and his 2020 view on this is still very relevant.
No horizon
Qualifier: because of limits in statistical models
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Generative AI has its uses, but I am increasingly convinced that on balance it is not a step forward for humanity.”
Our reading: Marcus is increasingly convinced that, although generative AI has its uses, on balance it is not a step forward for humanity.
No horizon
Qualifier: increasingly convinced
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“We should especially worry about children being influenced by these systems, as the deaths of teenagers Adam Raine and Sewell Setzer III have made clear.”
Our reading: Society should especially worry about chatbots influencing children, as teen deaths have shown.
No horizon
Assessment: Can’t be checkedSaid
Observation
“How OpenAI is ever going to turn enough profit to justify its valuation and massive commitments to other companies remains a mystery.”
Our reading: There is no clear explanation of how OpenAI will earn enough profit to justify its valuation and commitments.
No horizon
Assessment: Can’t be checkedSaid
Observation
“Business Insider’s Katie Notopoulous once called Altman the ultimate “personality hire”, but his only technical vision was scaling and that was never going to be enough.”
Our reading: Altman's only technical vision was scaling, and scaling was never going to be enough.
No horizon
Assessment: Can’t be checkedSaid
Observation
“By now, though, Altman’s act is starting to wearing thin.”
Our reading: Sam Altman's act is starting to wear thin.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Between Apple going elsewhere, Anthropic taking a big chunk of the corporate market, and the disappointments of its flagship GPT-5, to say nothing of Altman’s horrendous failure to explain how ends would meet on the Gerstner podcast, it’s time to reconsider whether the boy king is really the man for the job.”
Our reading: Given OpenAI's setbacks, it is time to reconsider whether Sam Altman should lead OpenAI.
No horizon
Qualifier: Apple going elsewhere
Assessment: Can’t be checkedSaid
Observation
“Although Apple ran that 2024 pilot with OpenAI, it appears now in hindsight that they must have been disappointed.”
Our reading: Apple was apparently disappointed with its 2024 pilot with OpenAI.
No horizon
Assessment: Can’t be checkedSaid
Forecast
“The company always looked like an accident waiting to happen, to me. (I have long speculated that it might turn out to be the WeWork of AI).”
Our reading: OpenAI might turn out to be the WeWork of AI.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Trying to orient our economy and geopolitical policy around such shoddy technology — particularly on the unproven hopes that it will dramatically improve– is a mistake.”
Our reading: The economy and geopolitical policy should not be oriented around generative AI, especially on unproven hopes that it will dramatically improve.
No horizon
Qualifier: particularly on unproven hopes of dramatic improvement
Source: Let’s be honest, Generative AI isn’t going all that well
Assessment: Can’t be checkedSaid
Observation
“Scaling isn’t going all that well, anymore, and probably won’t cure these problems.”
Our reading: Scaling of LLMs is no longer going all that well.
No horizon
Source: Let’s be honest, Generative AI isn’t going all that well
Assessment: Can’t be checkedSaid
Forecast
“Scaling isn’t going all that well, anymore, and probably won’t cure these problems.”
Our reading: Scaling will probably not cure LLMs' problems of untrustworthiness, memorization, and limited value.
No horizon
Source: Let’s be honest, Generative AI isn’t going all that well
Assessment: Can’t be checkedSaid
Forecast
“The giant AI infrastructure plays (like Project Stargate) that he championed after his inauguration will look like an unprofitable and underused mistake.”
Our reading: Offered as an unlikely scenario that could nonetheless come true: by the end of 2026, the giant AI infrastructure plays (like Project Stargate) that President Trump championed after his inauguration will look like an unprofitable and underused mistake.
Not yet due Evidence as of
Limitation: This record does not show how Project Stargate or other AI infrastructure projects are doing. The deadline has not passed.
Source: The (possibly) coming AI backlash and information warfare
Assessment: Can’t be checkedSaid
Forecast
“It’s a genuine effort, to be praised, at trying to have a system represent the world, over and beyond mere statistics of how language gets used on the web, but it will still likely fall short.”
Our reading: LeCun's world-model approach will likely fall short.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Still for all that, LeCun’s new company is trying to explore something genuinely new and that’s what we in the field of AI desperately needs.”
Our reading: LeCun's new company is trying to explore something genuinely new, and that is what the AI field desperately needs.
No horizon
Assessment: Can’t be checkedSaid
Observation
“LeCun talks about world models, but (because of what appears to be an ego-related mental block against classical symbolic AI, which he has always allied himself against) I don’t think LeCun actually understands what a world model needs to be.”
Our reading: Marcus does not think LeCun actually understands what a world model needs to be, which he attributes to what appears to be an ego-related mental block against classical symbolic AI.
No horizon
Qualifier: I don't think
Assessment: Can’t be checkedSaid
Forecast
“That said, I doubt that it will succeed, for two reasons.”
Our reading: LeCun's new company pursuing a new approach to AI will probably not succeed.
No horizon
Assessment: Can’t be checkedSaid
Forecast
“In turn, the fog of war may become more terrifying as citizens lose trust in much of what they read or see, and when conflicts are started and escalated by false pretexts.”
Our reading: As a result of AI-enabled misinformation and disinformation, the fog of war may become more terrifying as citizens lose trust in much of what they read or see, and when conflicts are started and escalated by false pretexts.
No horizon
Qualifier: much of what they read or see
Source: The (possibly) coming AI backlash and information warfare
Assessment: Can’t be checkedSaid
Forecast
“This, rather than some intelligence breakthrough, may well be the legacy of generative AI.”
Our reading: Generative AI's legacy may well be its use for generating misinformation and disinformation that sows chaos, rather than some intelligence breakthrough.
No horizon
Qualifier: may well
Source: The (possibly) coming AI backlash and information warfare
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“The unreliability of LLMs in the face of novelty is one of the core reasons why businesses shouldn’t trust LLMs”
Our reading: Businesses should not trust LLMs, in large part because LLMs are unreliable in the face of novelty.
No horizon
Qualifier: because of unreliability in the face of novelty
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“But the idea of using them to plot strategy in rapidly-changing environments like war is laughable.”
Our reading: The idea of using LLMs to plot strategy in rapidly-changing environments like war is laughable.
No horizon
Scope: rapidly-changing environments like war
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Relying on systems that can barely keep up with the publicly reported news, let alone stuff where human correction may be a lot harder to come by, is hardly the way to win a war.”
Our reading: Relying on systems that can barely keep up with publicly reported news, let alone matters where human correction may be a lot harder to come by, is hardly the way to win a war.
No horizon
Qualifier: let alone stuff where human correction may be a lot harder to come by
Assessment: Can’t be checkedSaid
Forecast
“(As I wrote here recently, the far more likely outcome is a tie.)”
Our reading: The US-China generative AI race will far more likely end in a tie than in either side dominating.
No horizon
Assessment: Can’t be checkedSaid
Observation
“Pure LLMs are inevitably stuck in the past, tied to when they are trained, and deeply limited in their inherent abilities to reason, search the web, “think” critically etc.”
Our reading: Pure LLMs are inevitably stuck in the past of their training and deeply limited in inherent abilities to reason, search the web, and think critically.
No horizon Evidence as of
Limitation: The cited study concerns GPT-3 in 2020, not the capabilities of LLMs generally in January 2026 [e1].
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“What world models are, and why we desperately need them.”
Our reading: AI desperately needs world models.
No horizon
Source: AI Cartoon of The Year — and five (re)readings for 2026
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Generalization is — and always has been — the vital question.”
Our reading: Generalization is, and always has been, the vital question in AI.
No horizon
Qualifier: and always has been
Source: AI Cartoon of The Year — and five (re)readings for 2026
Assessment: Can’t be checkedSaid
Forecast
“LLMs are not like you and me—and never will be.”
Our reading: LLMs are not like humans and never will be.
No horizon
Source: AI Cartoon of The Year — and five (re)readings for 2026
Assessment: Can’t be checkedSaid
Conditional forecast
“If the above analysis is correct, the dominance of LLMs may soon start to decline — and we may start seeing newer, more sound approaches start to emerge.”
Our reading: If his analysis is correct, LLM dominance may soon begin to decline.
No horizon
Only if: if the above analysis is correct
Source: AI Cartoon of The Year — and five (re)readings for 2026
Assessment: Can’t be checkedSaid
Conditional forecast
“If the above analysis is correct, the dominance of LLMs may soon start to decline — and we may start seeing newer, more sound approaches start to emerge.”
Our reading: If his analysis is correct, newer, sounder alternatives to LLMs may soon start to emerge.
No horizon
Only if: if the above analysis is correct
Source: AI Cartoon of The Year — and five (re)readings for 2026
Assessment: Can’t be checkedSaid
Conditional forecast
“If September 2025 was peak bubble, 2026 will likely the year which it all falls apart.”
Our reading: If September 2025 was the peak of the AI bubble, 2026 will likely be the year it all falls apart.
No horizon
Only if: if September 2025 was peak bubble
Assessment: Can’t be checkedSaid
Observation
“And now, this morning, I read that even in coding, the flagship application of generate AI, trouble is brewing:”
Our reading: Trouble is brewing even in coding, the flagship application of generative AI.
No horizon
Assessment: Can’t be checkedSaid
Observation
“Needless to say, the speculative America, buying chips and building data centers on the hope that it will pay off, is at odds with the growing disappointment.”
Our reading: Speculative US investment in chips and data centers conflicts with growing disappointment in generative AI.
No horizon
Assessment: Can’t be checkedSaid
Forecast
“And, honestly, I think that is the most likely scenario.”
Our reading: Marcus thinks the most likely scenario is the worst case he described, in which too many data centers and rapidly devaluing chips lead to shouts of 'too big to fail' and calls for government bailouts.
No horizon
Qualifier: honestly, I think
Assessment: Can’t be checkedSaid
Forecast
“In the worst case, as I have warned before, we may hear shouts of “too big to fail” and calls for government bailouts.”
Our reading: In the worst case, there may be 'too big to fail' claims and calls for government bailouts of the AI industry.
No horizon
Qualifier: in the worst case
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Going all in on Generative AI, without diversifying intellectually and economically, is likely an immense mistake.”
Our reading: Going all in on generative AI without intellectual and economic diversification is likely an immense mistake for the nation.
No horizon
Qualifier: without diversifying intellectually and economically
Assessment: Can’t be checkedSaid
Forecast
“We could easily wind up with too many data centers, and a lot of chips that rapidly lose their value.”
Our reading: The US could easily end up with too many data centers and chips that rapidly lose value.
No horizon
Assessment: Can’t be checkedSaid
Forecast
“Trump may eventually distance himself from AI because of this backlash.”
Our reading: Trump may eventually distance himself from AI due to public backlash.
No horizon
Qualifier: because of backlash to AI
Source: Six (or seven) predictions for AI 2026 from a Generative AI realist
Assessment: Can’t be checkedSaid
Observation
“Many would agree we have reached a point of diminishing returns for scaling; faith in scaling as a route to AGI has dissipated.”
Our reading: Many would agree that scaling has reached a point of diminishing returns, and faith in scaling as a route to AGI has dissipated.
No horizon
Qualifier: many would agree (attributed to many, not stated as flat fact)
Source: Six (or seven) predictions for AI 2026 from a Generative AI realist
Assessment: Can’t be checkedSaid
Observation
“Astonishing how much the vibe has shifted in just a few months, especially with people like Sutskever and Sutton coming out with their own concerns.”
Our reading: Sentiment about near-term AGI has shifted markedly in the past few months, with figures like Sutskever and Sutton voicing concerns.
No horizon
Source: Six (or seven) predictions for AI 2026 from a Generative AI realist
Assessment: Can’t be checkedSaid
Causal claim
“If anything would lead to genuine progress, it would be that.”
Our reading: If anything would lead to genuine progress in AI, it would be the field's shift from a stagnant, unrealistic intellectual situation to one that is more fluid, more realistic, and more open-minded.
No horizon
Qualifier: if anything would lead to genuine progress
Source: Six (or seven) predictions for AI 2026 from a Generative AI realist
Assessment: Can’t be checkedSaid
Observation
“across the field, the intellectual situation has gone from one that was stagnant (all LLMs all the time) and unrealistic (“AGI is nigh”) to one that is more fluid, more realistic, and more open-minded.”
Our reading: The AI field's intellectual climate has shifted from LLM-only and AGI-is-nigh thinking to a more fluid, realistic, and open-minded one.
No horizon
Source: Six (or seven) predictions for AI 2026 from a Generative AI realist
Assessment: Can’t be checkedSaid
Conditional forecast
“Putting society in the hands of giant, superficial correlation machines is not going to end well.”
Our reading: If society is put in the hands of giant, superficial correlation machines such as LLMs, it is not going to end well.
No horizon
Only if: putting society in the hands of giant, superficial correlation machines
Source: “New Ways to Corrupt LLMs”
Assessment: Can’t be checkedSaid
Observation
“And again their is an avenue that could be easily exploited.”
Our reading: Weird generalization in LLMs is an avenue that bad actors could easily exploit.
No horizon
Source: “New Ways to Corrupt LLMs”
Assessment: Can’t be checkedSaid
Observation
“As Evans noted, this is no joke. A bad actor could easily use this techniques to do nasty things:”
Our reading: Bad actors could easily use subliminal learning techniques to corrupt LLMs for harmful purposes.
No horizon
Source: “New Ways to Corrupt LLMs”
Assessment: Can’t be checkedSaid
Observation
“The problem with generative AI has always been that large language models associate patterns together without really understanding those patterns; it’s statistics without comprehension.”
Our reading: Large language models associate patterns without really understanding them; they are statistics without comprehension.
No horizon
Source: “New Ways to Corrupt LLMs”
Assessment: Can’t be checkedSaid
Causal claim
“Especially because of the speed at which chips like GPUs (a key component of that infrastructure) depreciate, it may be that the real winner is whichever country doesn’t overextend itself to the point of financial ruin, in a foolish effort to win a race that can’t be won.”
Our reading: Especially because chips like GPUs depreciate quickly, it may be that the real winner is whichever country doesn't overextend itself to the point of financial ruin in a foolish effort to win a race that can't be won.
No horizon
Qualifier: especially because of the speed at which chips like GPUs depreciate
Source: The core misconception that is driving American AI policy
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Building our AI policy around a fantasy that we are somehow going to crush China in LLM war (or vice versa!) is misguided.”
Our reading: Building US AI policy around a fantasy that the US is somehow going to crush China in an LLM war, or that China will crush the US, is misguided.
No horizon
Source: The core misconception that is driving American AI policy
Assessment: Can’t be checkedSaid
Conditional forecast
“All the more so if LLMs turn out to be a dud, or if LLMs are replaced by smaller, more efficient systems that don’t demand such immense amounts of infrastructure.”
Our reading: If LLMs turn out to be a dud, or are replaced by smaller, more efficient systems that don't demand such immense infrastructure, it may be all the more the case that the real winner is whichever country doesn't overextend itself to the point of financial ruin.
No horizon
Only if: if LLMs turn out to be a dud
Source: The core misconception that is driving American AI policy
Assessment: Can’t be checkedSaid
Causal claim
“he has come to believe (probably based on the self-interested whisperings from Silicon Valley investors who often lean on FOMO to China as a way to manipulate the government) that generative AI is a winner-take-all-race”
Our reading: Trump has come to believe that generative AI is a winner-take-all race, probably because of self-interested whisperings from Silicon Valley investors who often lean on FOMO about China to manipulate the government.
No horizon
Qualifier: probably
Source: The core misconception that is driving American AI policy
Assessment: Can’t be checkedSaid
Forecast
“My own hunch is that all this is going to blow up on Trump (or at least on his heir apparent J.D. Vance) badly.”
Our reading: Marcus's hunch is that the AI preemption executive order will badly backfire politically on Trump, or at least on J.D. Vance.
No horizon
Qualifier: or at least on J.D. Vance
Source: A sad day for America
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“If you live in the US and don’t like what’s happening — and don’t want to leave American citizens almost completely vulnerable to the downsides of AI — reach out to your governor and state attorney general, and let them know how you feel.”
Our reading: People who live in the US, don't like what's happening, and don't want to leave American citizens almost completely vulnerable to the downsides of AI should contact their governor and state attorney general and let them know how they feel.
No horizon
Scope: if you live in the US
Source: A sad day for America
Assessment: Can’t be checkedSaid
Rhetoric
“Instead President Trump has trampled over states’ rights and left our citizens in the lurch.”
Our reading: President Trump has trampled over states' rights and left US citizens in the lurch.
No horizon
Source: A sad day for America
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“What we obviously need is something in between, fostering innovation but also protecting our citizens.”
Our reading: The US should adopt AI regulation between over- and under-regulation that fosters innovation while protecting citizens.
No horizon
Source: A sad day for America
Assessment: Can’t be checkedSaid
Causal claim
“But maybe, maybe, just maybe, China’s reticence to stock up on H200s is also a sign that China has realized that GPU’s aren’t the royal road to AGI.”
Our reading: China's reluctance to stock up on H200s may also be a sign that China has realized GPUs aren't the royal road to AGI.
No horizon
Qualifier: maybe, maybe, just maybe
Source: Has the Chinese government figured out the fact that GPUs ≠ AGI?
Assessment: Can’t be checkedSaid
Forecast
“And realized that loading up on AI infrastructure that is likely to rapidly depreciate might be premature — and not the massive competitive or economic advantage many once people thought it was.”
Our reading: AI infrastructure such as GPUs is likely to depreciate rapidly, so loading up on it may be premature and not a massive competitive or economic advantage.
No horizon
Source: Has the Chinese government figured out the fact that GPUs ≠ AGI?
Assessment: Can’t be checkedSaid
Observation
“When the host asked “What is writing with ChatGPT doing to college education?”, I explained that writing with ChatGPT was outsourcing education:”
Our reading: Writing with ChatGPT is outsourcing education, in the context of college education.
No horizon
Scope: in college education, in answer to the host's question
Source: Has the Chinese government figured out the fact that GPUs ≠ AGI?
Assessment: Can’t be checkedSaid
Forecast
“(Like me, he also worries about overall coherence of the current policies, and about the US’s large investments in AI infrastructure that may not pay off.)”
Our reading: The US's large investments in AI infrastructure may not pay off.
No horizon
Source: Has the Chinese government figured out the fact that GPUs ≠ AGI?
Assessment: Can’t be checkedSaid
Observation
“At best, the motivations for the sale remain unclear.”
Our reading: The motivations for allowing H200 sales to China are, at best, unclear.
No horizon
Source: Has the Chinese government figured out the fact that GPUs ≠ AGI?
Assessment: Can’t be checkedSaid
Causal claim
“Trump is risking fracturing MAGA, and alienating the public, at a moment when his approval ratings are falling.”
Our reading: With his AI policies, Trump risks fracturing MAGA and alienating the public.
No horizon
Assessment: Can’t be checkedSaid
Observation
“People expect an Executive Order to follow, quite possibly illegal, banning states from regulating AI.”
Our reading: The expected Executive Order banning states from regulating AI may well be illegal.
No horizon