← Back to feed

George Hotz published “The Doom Justifies the Valuation” on June 21 — a direct indictment of the AI industry’s fear-based communications strategy, with a specific target: Anthropic.

The central argument: the technology currently available does not support frontier AI valuations. The gap is maintained by existential-risk framing, which creates a narrative framework where current products are proxies for potentially world-altering future capabilities. Without the doom, the valuation collapses.

The Argument

Hotz draws a contrast between two types of AI company blog posts.

The first type: Z.ai’s GLM-5.2 release post, which he describes as “a pleasant technical look into how things are slowly improving” — specific numbers, architecture decisions, benchmark results, what changed and why. He puts comma.ai’s own engineering blog in the same category.

The second type: Anthropic’s policy communications. He quotes two posts directly — one arguing AI is “advancing at exponential speed” beyond existing policymaking infrastructure, another on recursive self-improvement risks. His read: “It’s all just nonsensical hype, it’s not about technology.”

The mechanism, drawn from a paper Hotz cites (titled “Schizoposting”):

The only possible conclusion is that it’s designed to cause panic. In fact, it is optimized for it: there is no possible framing of the actual product(s) that could possibly induce more psychological spiraling in the media and its audience. “The AI Apocalypse?” has provided a years-long news cycle and an infinite spawner of contrived Discourse, which has primarily served to contextualize the AI industry’s valuation around not reality but hypothetical future value — i.e. a baseless assertion of infinite growth between the lines of a public-facing eschatology.

Corroboration From the Other Side

Hotz is not alone. Goldman Sachs CEO David Solomon told a public forum in May that fears of mass AI-driven unemployment are “overblown.” The Goldman view, which Hotz references, is that mainstream AI job displacement narratives have been consistently ahead of observable reality.

That convergence — a hardware engineer and a Wall Street CEO agreeing the narrative has outrun the product — is a different kind of signal than either alone.

What It Does Not Say

Hotz is not arguing the models are bad. He specifically calls out GLM-5.2 as frontier-class — on par with Claude Opus 4.8 and GPT-5.5. His objection is not to the technology but to the communications strategy deployed around it by safety-branded labs.

The distinction matters for how to read the argument. He is not a doomer critic of AI or an open-source absolutist. He is making a corporate-communications critique: that the doom framing is load-bearing infrastructure for valuations, and that it will be abandoned — without accountability — when the bubble cycle ends.

The Question He Poses

How much longer does this bubble have?

No answer is offered. The implication is that safety rhetoric borrowed its eschatological structure from prior tech bubbles — and that, like those bubbles, the reckoning will arrive slowly then suddenly, with the same participants who drove the narrative quietly moving on.

Anthropic’s current valuation sits at $965 billion. Its technical report acceptance at the frontier rests substantially on models that do outperform most benchmarks. Whether the valuation reflects those benchmarks or the recursive-self-improvement framing surrounding them is the question Hotz is raising — and not yet answering.