Where Is the AI Jobs Apocalypse? - by Peter McCrory - The Free Press

TEXT START: Artificial intelligence is changing everything—and fast.

The Dissection

This is not a refutation of the AI jobs apocalypse. It is a timing argument dressed as a debunking. The text takes stable aggregate employment as the decisive measure, then treats AI adoption and capability growth as evidence that displacement should already be visible.

The excerpt quietly shifts the question from “Is AI automating economically necessary human work?” to “Have mass layoffs appeared yet?” That is a narrower and lagging test. The author’s own description confirms the underlying mechanism: AI is already performing statistical modeling, equation solving, and visualization in an economist’s job. The human role has been reduced to directing increasingly automated production. The missing unemployment spike does not erase that transition; it shows that task automation has not yet fully translated into headcount elimination.

The article also turns Anthropic’s institutional position into authority. Its proximity to frontier AI gives it useful observational access, but the framing still serves the builder class: acknowledge transformation, emphasize present employment stability, and keep systemic alarm classified as premature.

The Core Fallacy

The central error is treating current unemployment as a real-time falsification test for structural displacement.

AI can suppress hiring, reduce hours, compress wages, eliminate junior pathways, increase output expectations, and replace departing workers without producing an immediate unemployment surge. Firms can absorb automation through attrition, labor hoarding, reassignment, and delayed restructuring. Aggregate job counts are therefore compatible with an economy whose productive participation is already being hollowed out.

The text also confuses three different variables:

• AI adoption is not worker replacement.
• Improved AI capability is not realized labor substitution.
• Stable employment is not stable human economic power.

Under the Discontinuity Thesis, this is precisely what lag defenses look like. Physical workflows, contracts, managerial habits, institutional inertia, and the need to redesign organizations delay the visible break. They do not reverse the mechanism.

Hidden Assumptions

• If unemployment has not risen sharply, displacement is not yet economically significant.
• Productivity gains will create enough new human work to absorb automated labor.
• Human direction of AI output will remain indispensable rather than becoming another temporary interface layer.
• The undefined measure of “quality-adjusted AI output” is a reliable proxy for economy-wide labor substitution.
• One-in-five business adoption indicates meaningful integration, even though the statistic says nothing about how much labor each deployment removes.
• Aggregate American job numbers reveal the condition of workers who are losing bargaining power, entry-level opportunities, or career ladders.
• The current employment pattern will persist as AI capability and organizational integration compound.
• The Anthropic economist’s institutional position has no influence on the interpretation.

These assumptions are doing the real argumentative labor. The employment data are merely the camouflage.

Social Function

Primarily ideological anesthetic and elite self-exoneration, with a partial truth embedded inside it.

The partial truth is narrow: the predicted mass unemployment event is not visible in the supplied figures. The anesthetic is broader: readers are encouraged to interpret delayed collapse as evidence against collapse. That gives AI developers, investors, and managers permission to continue accelerating while claiming that the social consequences remain speculative.

The article functions as transition management. It normalizes the early stage of automation by presenting the displaced worker’s problem as an unresolved economic mystery rather than as the opening phase of productive participation collapse.

The Verdict

The text records a lag, not a reprieve. It mistakes the absence of an unemployment crater for the absence of structural damage and treats a delayed macroeconomic symptom as proof against an accelerating technological mechanism.

On the supplied evidence, it does not disprove the Discontinuity Thesis. It documents its incubation phase: AI is already taking over cognitive tasks, while institutions still preserve the appearance of normal employment. The corpse has not entered the statistics yet. That is not evidence that the system is healthy; it is evidence that the killing mechanism is still clearing its lag defenses.