I think Anthropic and OpenAI have found product-market fit

TEXT START: I think Anthropic and OpenAI have found product-market fit

The Dissection

This is a piece of triumphalist insider journalism written from within the capital-owning class, celebrating the moment AI companies stop bleeding money and start extracting rent from enterprise. The author is a power user and obvious beneficiary of these tools, writing for an audience of fellow enthusiasts. The framing: "pricing pressures on enterprise customers are actually a good sign—it means the product works."

The article conflates short-term revenue extraction with long-term systemic viability. It mistakes the moment a vampire begins digesting its host for the moment the host achieved wellness.

The Core Fallacy

The article treats "enterprise customers paying more" as evidence of product-market fit, when it is actually evidence of vendor pricing leverage on a finite, soon-to-be-depleted labor pool.

The DT framework identifies the kill mechanism clearly: AI severs the mass employment → wage → consumption circuit. This article is about that severance happening in real-time. It just doesn't recognize it as a death signal.

When Uber's CTO says "we've maxed out our full-year AI budget in a few months, mostly thanks to Claude Code," the author waves this away: "budget set in 2025 failed to predict demand in 2026." The real story—the story the author is carefully not telling—is that one company's AI tool consumed an enterprise-level annual budget in months because it is replacing the work of hundreds of well-compensated engineers. That is not evidence of healthy market dynamics. That is accelerated displacement at pricing that will itself become unsustainable as the displacement accelerates.

The author quotes the COO of Uber saying: "It's very hard to draw a line between one of those stats and, OK, now we're actually producing like 25% more useful consumer features, right? And that line is hard to draw."

That quote is an autopsy finding. The author dismisses it as a story about "AI failure" being overblown. It is not about AI failing. It is about the productivity gains from automation not translating into revenue—which is precisely what DT predicts when the circuit breaks.

Hidden Assumptions

1. Sustainable demand assumption: The author assumes enterprise demand for AI coding agents will grow continuously and linearly. It will not. It will grow as displacement accelerates, then collapse as the displacement eliminates the enterprise customers' own revenue streams (their products, their services, their markets).

2. Human-in-the-loop assumption: The author assumes these tools remain "daily drivers for the work carried out by extremely well-compensated professionals" as a permanent category. Under DT mechanics, those professionals are the primary displacement target. The "skill premium" is a temporal artifact, not a durable moat.

3. Revenue equals health assumption: The author treats Anthropic's rumored first profitable quarter as validation. It is validation of short-term cash extraction, not of a durable business model. The analogy: a mine operator extracting ore at a profit while the mine shaft collapses above them is not running a sustainable operation.

4. The SpaceX deal as evidence of scale, not desperation: The $1.25 billion/month Anthropic-to-SpaceX deal is presented as a sign of inference infrastructure demand. It is also evidence that Anthropic is purchasing compute at a rate that implies their revenue, even at current trajectory, is not obviously self-sustaining.

5. "Product-market fit" as a terminal state: The author treats PMF as a destination. Under DT, the product works—but the market it is creating is the market for the post-labor economy, which has no mass consumer base.

Social Function

Prestige signaling and insider enthusiasm. This is a piece written by someone who benefits materially from the system being described, for an audience that benefits similarly. It performs the function of narrative management: reframing displacement signals as success signals. The author even dogfoods the analysis ("I ran this analysis by scraping their job sites with Claude Code"), which is charming but also reveals the circularity: using AI to celebrate AI's success, for readers who use AI.

It is not a piece of copium for the masses. The masses are not the audience. It is a piece of revenue theater for the investor class, timed to coincide with anticipated IPO S-1 filings.

The Verdict

The article accurately identifies that AI labs have achieved short-term revenue extraction from enterprise customers. It completely fails to identify the mechanism by which that revenue stream is self-terminating.

Under DT logic, the following chain is already operating:

1. Coding agents displace software engineers → enterprise customers pay less in wages → enterprise customers' costs fall → enterprise customers' revenue from their own products depends on market demand that is itself dependent on mass employment/wages → that demand contracts.

2. The "enterprise customers" paying $1,000+/month per employee are paying for tools that eliminate the need for those employees. The author frames this as "fantastic deal" for the enterprise. It is the terminal contradiction of the knowledge economy: you cannot sustain a revenue model based on selling your customers the tools of their own obsolescence, and then expecting them to remain customers.

3. The pricing power described—locking enterprise into API-rate billing, raising prices, cutting discounts—is exactly what a firm does when it recognizes its monopoly window is finite. The urgency is the tell. Sustainable monopolies do not need to reprice aggressively and simultaneously across two competitors in the same quarter.

4. The $1.25 billion/month SpaceX deal is not evidence of Anthropic's health. It is evidence of an arms race in inference compute that implies competitive dynamics that will compress margins regardless of enterprise pricing changes.

The author is watching the vampire feed and calling it a dinner party. The food is the system itself.