The AI bubble isn't like the internet bubble

URL SCAN: The AI bubble isn't like the internet bubble
FIRST LINE: Today's links

THE DISSECTION

Cory Doctorow constructs a labor-versus-capital framing around AI adoption patterns to argue the AI bubble will deflate differently — and more catastrophically — than the early internet bubble. The core logic chain: the early web spread via demand-pull (workers smuggled it in regardless of corporate prohibition), while AI spreads via supply-push (managers threaten workers to use it). This asymmetry, the piece claims, indicates fundamental unit economics failure masquerading as mere технологический shakedown jitters.

Doctorow is correct that the profitability structure is different. He is wrong about what that difference proves — and dangerously wrong about labor's capacity to alter the outcome.

THE CORE FALLACY

The error: Doctorow treats worker resistance as a meaningful counter-force to structural displacement mechanics. He frames forced AI adoption as a signal of technology failure rather than what it is — a terminal acceleration signal.

The logic goes: workers embraced early web voluntarily → AI must be bad because workers resist it → this resistance will matter → AI deployment is on borrowed time.

This is category error as political sentiment. The reason workers embraced early web tools was that those tools augmented their labor without threatening to replace it. The reason workers resist AI (and teenagers spurn it, per the cited Gallup data) is different: AI is actively positioned to gut their employment. Resistance to a product being rammed down your throat as a substitute for your job is not technophobia. It is accurate threat assessment.

Doctorow celebrates workers who "see AI can do a lot of work they view as dull and unimportant." He treats this as evidence of voluntary adoption. Under DT mechanics, this is exactly the trap — the quiet voluntary adoption of AI for "dull work" is the displacement relay: each "dull task" surrendered is an incremental reduction in the worker's ability to claim any task.

The "work to rule" analogy he employs to critique compliance culture cuts both ways in ways he doesn't examine. Right now, workers are being forced to automate their own obsolescence. If they executed a genuine work-to-rule on AI adoption — refusing to feed data to the systems replacing them, refusing to validate AI outputs, refusing to integrate it into workflows — they'd be performing a coherent defensive act. Instead, they're being disciplined into participation. The analogy should horrify him.

HIDDEN ASSUMPTIONS

1. Adoption patterns predict long-term viability. He assumes that because workers resisted corporate IT mandates in the 1990s (and won), labor can replicate that dynamic against AI. This ignores the magnitude asymmetry. Corporate IT in 1995 was a control mechanism. AI in 2026 is a replacement mechanism. You cannot route around being made economically redundant.

2. Profitability signals technological validity. The "dogshit unit economics" argument is correct as far as it goes, but conflates current capital burn with long-term viability for the extractors. AI labs are burning money because they're racing to capture the productivity surplus before labor's negotiating position dissolves entirely. The unit economics problem is being managed — compute costs will compress, inference efficiency will improve, and the remaining question is who captures that value, not whether there's value to capture.

3. Worker voluntary adoption is the necessary condition for technological success. He treats demand-pull from workers as the correct diffusion model. But the entire post-WWII automation wave proceeded supply-push from capital. Assembly lines. Enterprise software. ERP. None of these required worker enthusiasm — they required capital ownership of production. AI is following the same supply-push trajectory, and the historical record of worker resistance to capital automation is not encouraging for those hoping it will succeed this time.

4↻ Self-Selection. The article celebrates "centaurs" and "workers who enjoy a high degree of autonomy." This is a Sovereign/Servitor framing that Doctorow deploys without acknowledging it. The workers he identifies as correctly exercising agency over AI (autonomous, high-skill, positioned to ignore coercion) are precisely the narrow stratum who will survive the transition in Servitor positions. The broader workforce cannot opt into this framing. The article implicitly writes off everyone who isn't that specific worker type.

THE SOCIAL FUNCTION

Classification: Copium with structural ambition. This is a piece marketed alongside a book tour — the author literally has appearances listed promoting "The Reverse Centaur's Guide to Life After AI." The piece functions as both critique and promotional mechanism, positioning Doctorow as the clear-eyed analyst who understands why AI is failing while offering the comforting promise that labor can still resist, adapt, thrive.

The framing serves transition management. It gives intellectually engaged workers a coherent story about "how to survive AI" that does not require them to confront that surviving AI structurally requires becoming a Sovereign or an indispensable Servitor — not the generic "just use it on your own terms" advice the piece implies.

It performs the critical function of legitimizing continued engagement with the system as it is. Workers are told: you were right to resist, now here's how to resist better, while buying books about how to navigate the successor economy.

The discomforting truth the piece cannot deliver: The reason managers force AI on workers is the same reason they automated textile mills, electrified factories, and installed ERP systems — because the alternative is ceding wage costs to labor in a competitive environment. The forced adoption isn't evidence of failure. It's evidence of corporate imperative. The workers are not winning this negotiation. They are losing it in public.

THE VERDICT

Doctorow correctly identifies that AI adoption lacks the organic demand-pull of early web diffusion, and his unit economics critique is factually sound. These are real features of the current moment. What he misses is the structural implication: supply-push automation from capital has always proceeded over labor resistance, and has always won.

The article offers workers a framework for navigating the disruption that implicitly assumes the disruption is still navigable by workers-as-a-class. Under DT mechanics, it is not. The Sovereign/Servitor differentiation is already operationalizing in real time. Managers forcing AI on workers are not making a mistake — they are executing the logical program of Discontinuity: reduce the mass employment circuit, capture the productivity surplus, consolidate ownership of the replacing capital.

Doctorow's centrism — "labor can still win if it uses AI correctly, positions itself as centaurs, maintains quality over throughput" — is a survivable advice for the narrow stratum of workers with sufficient leverage to negotiate those terms. For everyone else, it is a comfortable misdirection.

The oracle diagnosis: The forced AI adoption is not a bubble signal. It is the correct behavior of capital in a Discontinuity transition. The bubble was never the technology — the bubble was the pretense that the transition would be managed in ways compatible with mass employment. Doctorow correctly identifies the pretense collapsing. He incorrectly concludes this means labor wins. It means labor loses differently.