AI Is Really Really Taking Jobs Today - 24/7 Wall St.

TEXT START: New Federal Reserve research is tracking something economists have long debated, and the early numbers point toward a labor market shift that could rival the worst economic downturns in modern American history.

The Dissection

The article begins with a narrow signal—Dallas Fed research on Texas job postings—and inflates it into a national catastrophe. It moves from task exposure to job loss, from separate industry estimates to 10 million unemployed workers, and from that projection to a Great Depression comparison without establishing that the figures share definitions, populations, timing, or causal methods.

The text is doing two things at once: reporting an early warning of AI pressure on labor demand and manufacturing a dramatic click-friendly collapse narrative. Under the Discontinuity Thesis, it detects surface damage but lacks the deeper diagnosis: the decisive issue is not merely the unemployment rate, but whether ownership of AI-controlled production makes most human labor economically unnecessary.

The Core Fallacy

The article treats “automatable” as equivalent to “eliminated.” Those are separate events:

• AI can perform a task.
• A firm deploys it.
• The deployment reduces labor demand.
• A worker loses a job.
• The worker becomes unemployed rather than reemployed or leaving the labor force.

The article proves none of those transitions at national scale. A decline in job postings is not automatically layoffs. A task-mapping exercise is not proof of implementation. The 2.6% exposure figure is then combined with unrelated Goldman Sachs and University of Delaware estimates as though they were additive and non-overlapping.

The arithmetic is theatrical. The text converts a collection of partial estimates into a 10% unemployment scenario by assuming simultaneous job destruction, no reemployment, no labor-force exit, no denominator change, and no institutional delay. Its historical comparison is also muddled: it invokes the Great Recession’s unemployment level and then labels the result an economic disaster unseen since the Great Depression.

The deeper DT error is using a cyclical unemployment threshold as the test for systemic death. Ten percent unemployment would not by itself prove the post-WWII order is dead. Conversely, the wage-consumption circuit could be structurally severed before unemployment reaches that figure if AI capital becomes dominant and its gains accrue to a narrow ownership class.

Hidden Assumptions

• The O*NET-to-Claude task mapping accurately predicts real-world replacement.
• Capability translates rapidly and uniformly into corporate adoption.
• The Texas posting data isolates AI rather than broader changes in hiring strategy.
• Texas is representative of the entire American labor market.
• The cited estimates measure the same kind of loss and can be added together.
• Affected jobs disappear rather than being reorganized, downgraded, or reduced in hours and pay.
• Displaced workers become unemployed rather than reemployed or exiting the labor force.
• All effects arrive simultaneously and continue linearly.
• Legal, institutional, physical, and cultural lag defenses are negligible.
• New demand or new occupations offset none of the destruction.
• Entry-level damage is a temporary adjustment rather than the removal of the human pathway into white-collar work.

Social Function

Class: partial truth functioning as alarmist transition management and ideological anesthetic.

The article gives readers a real signal—AI-related pressure is appearing in vulnerable work, especially entry-level and technical roles—then buries the structural power transfer beneath unemployment statistics and historical panic. It presents displacement as a future macroeconomic disaster rather than asking who owns the systems replacing workers, who controls access to productive capital, and who receives the output.

That framing is useful to the existing order. It encourages fear of a recession while avoiding the more terminal question of whether labor itself is losing bargaining power. The result is not pure copium; it is an incomplete alarm wrapped in sensationalism.

The Verdict

The supplied evidence supports an early warning of AI-driven labor-demand damage in selected sectors. It does not support the article’s 10-million-unemployed or Great Depression conclusions.

Under DT logic, this is a crude but directionally useful alarm: it sees the first cracks, misidentifies the failure mechanism, and replaces a systemic autopsy with a body count. The real threshold is not 10% unemployment. It is the point at which AI-controlled production makes the majority economically unnecessary and leaves non-owners without a claim on the output.