Steve Cuozzo tests AI, finds it gets him all wrong

TEXT ANALYSIS: "Steve Cuozzo Tests AI, Finds It Gets Him All Wrong"

The Dissection

Cuozzo performs a parlor trick dressed as journalism: he asks AI platforms about himself, catches them hallucinating (wrong books, wrong facts, premature obituary), and concludes AI isn't ready to "take over." The piece is entertaining. It is also analytically worthless as a verdict on AI's economic trajectory. It mistakes a current engineering defect for evidence of structural incapacity. This is like noting that early cars sometimes caught fire and concluding they'd never replace horses.

The Core Fallacy

The article evaluates AI against the wrong benchmark. Cuozzo asks: "Can AI accurately retrieve facts about me?" What the DT framework demands is a different question: "Can AI perform economically relevant cognitive work at sufficient quality below the cost of human alternatives?" These are not the same question. Hallucination is a solvable problem in training pipelines and retrieval architectures. Economic substitution does not require perfection—it requires adequate performance at sufficient margin.

Hidden Assumptions

1. Present capability equals terminal ceiling. The piece assumes AI's current hallucination rate represents a fundamental flaw, not a transitional engineering problem.
2. Human retrieval is the gold standard. Cuozzo's implicit benchmark is "can AI do what I do?" not "can AI do what matters economically at the required price point?"
3. The journalism business model is the unit of analysis. He tests AI on biographical accuracy for a writer. He does not test AI on whether it can generate acceptable content at near-zero marginal cost—which is the actual existential threat to his industry.
4. Accuracy is the scarce resource. In post-WWII media economics, accuracy and authorship were scarce. In the DT framework, the scarce resource is human curation of meaning. AI doesn't need to be accurate about Steve Cuozzo. It needs to be cheap enough to make Steve Cuozzo economically optional.

Social Function

Ideological anesthetic for the professionally vulnerable. This piece performs a vital service for its target readership: mid-tier information workers (journalists, analysts, editors, content producers) who are watching their economic position erode. It says: "Relax. The machines are dumb. Your job is safe." It is the most comforting possible reading of a deeply threatening trend. It is copium with a byline.

The Verdict

Cuozzo has demonstrated that AI hallucinates biographical facts about him. He has not demonstrated that AI will not render his economic function—generating written content about commercial real estate and restaurants for a declining metropolitan newspaper—structurally unnecessary. The article's own industry (print journalism) is dying not because AI gets Steve Cuozzo's biography wrong, but because AI can generate adequate content at near-zero marginal cost, collapsing the advertising and subscription revenue model that once funded his salary.

The piece is a human being stubbing his toe and concluding the earthquake won't happen. The earthquake is not about the accuracy of individual retrieval. It is about the economics of cognitive labor production in a world where AI achieves durable cost-performance superiority across domains. That trajectory is not arrested by the fact that Claude can't find his degree on Wikipedia.

Final Note: Cuozzo ends by joking AI should "learn how to steal from Wikipedia." In the DT framework, this is not a joke—it is a precise description of the mechanism. AI is learning. It will. And when it does, the question will not be whether it gets Steve Cuozzo's bio right. The question will be whether anyone needs Steve Cuozzo's bio at all.