Pundit ClaimsAI claims by Gary Marcus and Ed Zitron, with the evidence

Ed Zitron Where's Your Ed At Published

AI Is Really Weird

We recorded 42 claims from this post. 23 were checked against evidence. 19 can’t be checked. 28 more are held. Read the post at its source.

Claims from this post (42)

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.

  1. Assessment: UnresolvedSaid

    “As I discussed a few weeks ago, despite everybody talking about the hundreds of gigawatts of data centers being built “to power AI,” only 5GW are actually “under construction,” with “under construction” meaning anything from “we’ve got some scaffolding up” to “we’re about to hand over the keys to the customer.””

    What we tested: Only about 5GW of the hundreds of gigawatts of announced AI data centers are actually under construction.

  2. Assessment: UnresolvedSaid

    “And with that incredibly easy access, only 3% of households pay for AI.”

    What we tested: Only 3% of households pay for AI despite easy access.

  3. Assessment: Can’t be checkedSaid

    “Yet the “early days” argument is inherently deceptive.”

    What we tested: The argument that AI is in 'the early days' is inherently deceptive.

  4. Assessment: UnresolvedSaid

    “AI progress isn’t hampered by a lack of access, talent, resources, novel approaches, or industry buy-in, but by a single-minded focus on Large Language Models, a technology that has been so obviously-limited from the very beginning that Gary Marcus was able to call it in 2022.”

    What we tested: AI progress is not hampered by a lack of access, talent, resources, novel approaches, or industry buy-in, but by a single-minded focus on Large Language Models, a technology that has been obviously limited from the very beginning.

  5. Assessment: Can’t be checkedSaid

    “Saying it’s “the early days” also doesn’t really make sense when faced with the rotten and incredibly unprofitable economics of AI.”

    What we tested: Saying AI is in 'the early days' doesn't really make sense given the rotten and incredibly unprofitable economics of AI.

  6. Assessment: Can’t be checkedSaid

    “In any case, anytime that somebody tells you that we’re in “the early days of AI” has either been conned or is in the process of conning you, as they’re using it to deflect from issues of efficacy or underlying economic weakness.”

    What we tested: People claiming it is the early days of AI are conned or conning, deflecting from efficacy and economic weakness.

  7. Assessment: Can’t be checkedSaid

    “Probably the weirdest thing about this entire era is how nobody wants to talk about the fact that AI isn’t actually doing very much, and that AI agents are just chatbots plugged into an API.”

    What we tested: AI isn't actually doing very much, and AI agents are just chatbots plugged into an API.

  8. Assessment: Can’t be checkedSaid

    “Compliance reporting? Summarizing or pulling documents from places, much like any LLM can do, other than the fact that it’ll likely get shit wrong, which is bad for compliance.”

    What we tested: Anthropic/Infosys AI agents for compliance reporting will likely produce errors.

  9. Assessment: SupportedSaid

    “To be clear, LLMs can absolutely write code, and can absolutely create software, but neither of those mean that the code is good, stable or secure, or that the same can be said of the software they create.”

    What we tested: LLMs can write code and create software, but that doesn't mean the code is good, stable, or secure.

    What happened: Anthropic's generated compiler demonstrates substantive software-generation capability while retaining functional limitations. That directly fits the distinction in Zitron's sentence. (Anthropic, )

  10. Assessment: Can’t be checkedSaid

    “They do not have ideas, nor do they create unique concepts — everything they create is based on training data fed to it that was first scraped from Stack Overflow, Github and whatever code repositories Anthropic, OpenAI, and Google have been able to get their hands on.”

    What we tested: Coding LLMs have no ideas or unique concepts; all output derives from scraped training data.

  11. Assessment: UnresolvedSaid

    “It’s unclear what the actual economic or productivity effects are, other than an abundance of new code that’s making running companies harder.”

    What we tested: The actual economic or productivity effects of coding LLMs are unclear, other than an abundance of new code that is making running companies harder.

  12. Assessment: UnresolvedSaid

    “As I wrote a few weeks ago, LLMs are good at writing a lot of code, not good code, and the more people you allow to use them, the more code you’re going to generate, which means the more time you’re either going to need to review that code, or the more vulnerabilities you’re going to create as a result.”

    What we tested: LLMs write a lot of code, not good code; broader use means more review time or more vulnerabilities.

  13. Assessment: Can’t be checkedSaid

    “Worse still, hyperscalers like Meta and Amazon are allowing non-technical people to ship code themselves, which is creating a crisis throughout the tech industry.”

    What we tested: Hyperscalers letting non-technical people ship LLM code is creating a crisis in the tech industry.

  14. Assessment: UnresolvedSaid

    “Worse still, LLMs allow shitty software engineers that would otherwise be isolated by their incompetence to feign enough intelligence to get by, leading to them actively lowering the quality of code being shipped.”

    What we tested: LLMs let incompetent engineers get by, lowering the quality of shipped code.

  15. Assessment: Can’t be checkedSaid

    “Software engineers don’t seem happier, nor do they seem to be paid more, nor do they seem to be being replaced by AI, nor do we have any examples of truly vibe coded software companies shipping incredible, beloved products.”

    What we tested: Software engineers don't seem happier, better paid, or replaced by AI, and we don't have any examples of truly vibe-coded software companies shipping incredible, beloved products.

  16. Assessment: UnresolvedSaid

    “That “growth” is only possible because all of the companies in question heavily subsidize their products, spending $3 to $15 for every dollar of revenue.”

    What we tested: The growth of the AI companies in question is only possible because all of them heavily subsidize their products, spending $3 to $15 for every dollar of revenue.

  17. Assessment: UnresolvedSaid

    “Back in November 2025, I reported that OpenAI had made — based on its revenue share with Microsoft — $4.329 billion between January and September 2025, despite The Information reporting that it had made $4.3 billion in the first half of the year based on disclosures to shareholders.”

    What we tested: Based on its Microsoft revenue share, OpenAI made $4.329 billion in January–September 2025.

  18. Assessment: Can’t be checkedSaid

    “I was not reached out to by or otherwise acknowledged by any other outlets, and every outlet has continued to repeat that OpenAI “made $13 billion in 2025,” despite that being very unlikely given that it would have required it to have made $8 billion in a single quarter.”

    What we tested: It is very unlikely OpenAI made $13 billion in revenue in 2025, as that would require $8 billion in one quarter.

  19. Assessment: Can’t be checkedSaid

    “Though I cannot say for certain, both of these situations suggest that Anthropic and OpenAI are misleading their investors, the media and the general public.”

    What we tested: Revenue discrepancies suggest, though not certainly, that OpenAI and Anthropic are misleading investors, media, and the public.

  20. Assessment: SupportedSaid

    “The above quote has been used repeatedly to suggest that Anthropic has 50% gross margins and is “profitable,” which is extremely weird in and of itself as that’s not what Dario Amodei said at all.”

    What we tested: Amodei's remarks don't actually claim Anthropic has 50% gross margins or is profitable.

    What happened: Amodei explicitly separates the illustrative numbers from information about Anthropic. The surrounding discussion frames the calculation as hypothetical industry economics rather than a current company financial disclosure. (Dwarkesh Podcast, )

  21. Assessment: Can’t be checkedSaid

    “Yes, I know it’s a “stylized fact” or whatever, but that’s what he said, and I think that their IPO might have a rude surprise in the form of a non-EBITDA margin calculation that makes even the most-ardent booster see red.”

    What we tested: Zitron thinks Anthropic's IPO might bring a rude surprise in the form of a non-EBITDA margin calculation that makes even the most ardent booster see red.

  22. Assessment: Can’t be checkedSaid

    “Having two sets of earnings is extremely dodgy! Especially when the difference between them is billions of dollars.”

    What we tested: OpenAI and Anthropic showing investors two sets of earnings, with and without training costs, is extremely dodgy, especially when the difference between them is billions of dollars.

  23. Assessment: UnresolvedSaid

    “That “fast-growing” part is only possible because both Anthropic and OpenAI subsidize the compute of their subscribers, allowing them to burn $3 to $15 for every dollar of subscription revenue.”

    What we tested: OpenAI and Anthropic's fast growth is only possible because they subsidize subscribers' compute, burning $3-$15 per dollar of subscription revenue.

  24. Assessment: Can’t be checkedSaid

    “Nevertheless, something is up with Anthropic’s revenue story.”

    What we tested: Something is off with Anthropic's revenue story.

  25. Assessment: UnresolvedSaid

    “At this point, I estimate Anthropic’s revenue split to be more in the region of 75% API and 25% subscriptions, based on its supposed $2.5 billion in annualized revenue (out of $14 billion, so a little under 18%) in February coming from “Claude Code” (read: subscribers to Claude, there’s no “Claude Code” subscription).”

    What we tested: Anthropic's revenue is roughly 75% API and 25% subscriptions, by the author's estimate.

  26. Assessment: ContradictedSaid

    “If that’s the case, I truly have no idea how it could’ve possibly accelerated so aggressively, and as I’ve mentioned before, there is no way to reconcile having made $5 billion in lifetime revenue as of March 9, 2026, having $14 billion in annualized revenue on February 12 2026, and having $4.5 billion in revenue for the year 2025.”

    What we tested: There is no way to reconcile Anthropic having made $5 billion in lifetime revenue as of March 9, 2026, $14 billion in annualized revenue on February 12, 2026, and $4.5 billion in revenue for 2025.

    What happened: The sentence immediately preceding the quotation says cumulative revenue exceeded $5 billion; it supplies no upper bound. This permits additional revenue beyond $5 billion and defeats the asserted incompatibility of the reported figures. (U.S. District Court, Northern District of California / CourtListener RECAP, )

  27. Assessment: UnresolvedSaid

    “In simpler terms, Anthropic is cherry-picking four-week windows of API spend — ones that are pumped by big announcements and new model releases — and annualizing them.”

    What we tested: Anthropic cherry-picks four-week API spend windows boosted by launches and annualizes them.

  28. Assessment: UnresolvedSaid

    “This leads to context bloat that absolutely rinses your token budget.”

    What we tested: Anthropic's default 1 million token context window causes context bloat that greatly raises token spend.

  29. Assessment: Can’t be checkedSaid

    “if Anthropic actually made $2.5 billion in a month — we’ll find out when it files its S-1! — it likely came not from genuine growth or a surge of adoption, but in its existing products suddenly costing a shit ton more because of how they’re engineered.”

    What we tested: If Anthropic really made $2.5B in a month, it likely came from products costing more per use, not genuine growth or adoption.

  30. Assessment: Can’t be checkedSaid

    “There’s also one other option: that Silicon Valley is effectively subsidizing Anthropic through an industry-wide token-burning psychosis. And based on some recent news, there’s a chance that’s the case.”

    What we tested: There is a chance Anthropic's revenue is being propped up by industry-wide wasteful token burning.

  31. Assessment: Can’t be checkedSaid

    “If this is what’s actually happening, and this is what’s contributing to Anthropic’s growth, this is not a sustainable business model, which is par for the course for Anthropic, a company that has only lost billions of dollars.”

    What we tested: If Meta is really spending around $300 million or more a month on tokens and this is contributing to Anthropic's growth, that is not a sustainable business model, which is typical for Anthropic, a company that has only lost billions of dollars.

  32. Assessment: Can’t be checkedSaid

    “In any case, TokenMaxxing is not real demand, but an economic form of AI psychosis.”

    What we tested: TokenMaxxing is not real demand for AI.

  33. Assessment: UnresolvedSaid

    “I have confirmed with a source at that there is no actual metric or tracking of any return on investment involved in token burn at Meta, meaning that TokenMaxxing’s only purpose is to burn more tokens to go higher on a leaderboard, and is already creating bad habits across a company that already has decaying products and leadership.”

    What we tested: Meta does not track any ROI on token burn; tokenmaxxing there exists only to climb a leaderboard.

  34. Assessment: UnresolvedSaid

    “Furthermore, TokenMaxxing is exactly the kind of revenue that disappears first.”

    What we tested: TokenMaxxing spending is exactly the kind of revenue that disappears first.

  35. Assessment: UnresolvedSaid

    “Combined with mandates to “use as much AI as possible,” this naturally increases the cost of having software engineers, which — and I say this not wanting anyone to lose their jobs — does the literal opposite of replacing workers with AI.”

    What we tested: Model providers reportedly pushing enterprise clients onto API billing, combined with mandates to use as much AI as possible, naturally increases the cost of having software engineers, the opposite of replacing workers with AI.

  36. Assessment: UnresolvedSaid

    “Said data centers make up 90% of NVIDIA’s revenue, which means that 8% or so of the S&P 500’s value comes from a company that makes money selling hardware to people that immediately lose money on installing it.”

    What we tested: Data centers make up 90% of NVIDIA's revenue, so about 8% of the S&P 500's value comes from a company selling hardware to buyers who immediately lose money installing it.

  37. Assessment: UnresolvedSaid

    “It’s also weird that people discussing agents don’t seem to want to discuss that OpenAI’s Operator Agent does not work, that AI browsers are fundamentally broken, or that agentic AI does not do anything that people discuss.”

    What we tested: OpenAI's Operator does not work, AI browsers are fundamentally broken, and agentic AI doesn't do what people say.

  38. Assessment: UnresolvedSaid

    “Stargate was a lie! The whole time it was a lie! That time that Sam Altman and Masayoshi Son and Larry Ellison stood up at the white house and talked about a $500 billion infrastructure project was a lie! They never formed the entity!”

    What we tested: The $500B Stargate project was a lie; the entity was never formed.

  39. Assessment: ContradictedSaid

    “The revenues are terrible, Large Language Models are yet to provide any meaningful productivity improvements, and the only reason that they’ve been able to get as far as they have is a compliant media and a venture capital environment borne of a lack of anything else to invest in.”

    What we tested: AI revenues are terrible, LLMs have yet to provide any meaningful productivity improvements, and the only reason they have got as far as they have is a compliant media and a venture capital environment that comes from having nothing else to invest in.

    What happened: The study examines an LLM-based assistant in actual customer-support work, with rollout extending into winter 2021. The estimated gain is material under the ordinary workplace reading, although it is not an economy-wide estimate. (The Quarterly Journal of Economics; author-hosted copy, )

  40. Assessment: UnresolvedSaid

    “Coding LLMs are popular only because of their massive subsidies and corporate encouragement, and in the end will be seen as a useful-yet-incremental and way too expensive way to make the easy things easier and the harder things harder, all while filling codebases full of masses of unintentional, bloated code.”

    What we tested: Coding LLMs are popular only because of massive subsidies and corporate encouragement, and in the end will be seen as a useful-yet-incremental and far too expensive way to make easy things easier and hard things harder, while filling codebases with bloated, unintentional code.

  41. Assessment: Can’t be checkedSaid

    “If everybody was forced to pay their actual costs for LLM coding, I do not believe for a second that we’d have anywhere near the amount of mewling, submissive and desperate press around these models.”

    What we tested: Zitron firmly believes that if everybody had to pay the actual costs of LLM coding, there would be nowhere near as much mewling, submissive and desperate press about these models.

  42. Assessment: Can’t be checkedSaid

    “At the end of this era, I think we’re all going to have to have a conversation about the innate credulity of the business and tech media, and how often that was co-opted to help the rich get richer.”

    What we tested: Zitron thinks that at the end of this era everyone will have to have a conversation about the innate credulity of the business and tech media and how often it was co-opted to help the rich get richer.

Held from this post (28)

These records did not pass every check, so nothing about them is published. Held is not a finding.