There's at Least One Job That AI Isn't Killing - Gizmodo

TEXT START: One of the things artificial intelligence seems particularly adept at is cybersecurity—at least if you buy the marketing pitch of models like Anthropic's Mythos...

THE DISSECTION

This is anxiety management journalism dressed as contrarianism. The piece performs a valuable social function: it tells readers who want to believe the AI employment apocalypse is overblown exactly what they need to hear. Surface-level metrics—job listings up 11%, unemployment lower for "AI-exposed" roles—are paraded as evidence that the structural thesis is wrong. The self-aware bits about "vibe coding" and "AI washing" provide just enough honesty to make the reassuring conclusion credible. It's copium with a sophisticated wrapper.

THE CORE FALLACY

Counting warm bodies instead of diagnosing structural position.

The Discontinuity Thesis doesn't predict simultaneous job annihilation. It predicts the severing of the mass employment -> wage -> consumption circuit through AI's eventual superiority in cognitive work. The article commits the cardinal error of measuring quantity of jobs while ignoring quality and structural necessity.

A hiring spree in cybersecurity for the purpose of cleaning up AI-generated code messes is not evidence that humans remain economically necessary. It's evidence of defensive spending on a losing front. Every "Masturbation Consultant" and panic-hired philosopher is a data point in the "humans pushed to increasingly absurd niches" category, not the "AI has limits" category.

The BLS unemployment data captures lag indicators. The Yale Budget Lab research measures current flows, not terminal trajectories.

HIDDEN ASSUMPTIONS

1. Job listings = structural demand. They might be panic hiring, compliance theater, or desperate attempts to manage AI-created chaos that will itself be automated away.
2. Current employment stability disproves structural thesis. It doesn't. It confirms the lag phase exists. That's what lag means.
3. Humans hired to manage AI outputs = human relevance maintained. This is the strongest sleight of hand. Humans as AI janitors is not a thriving employment sector—it's a transitional hospice.
4. The cybersecurity hiring surge is permanent. If Anthropic's Mythos can spot vulnerabilities humans missed for 20 years, what happens when that capability matures? Who audits the auditors?

SOCIAL FUNCTION

Lullaby with a wink. This article wants to be read as the smart skeptic taking on AI hype, but its actual function is to soothe anxiety and provide intellectual cover for "we have time" thinking. The philosophical consultants and absurd job listings aren't evidence of human irreplaceability—they're museum exhibits of the species circling the drain while producing ever more specialized performative labor.

THE VERDICT

The article offers surface reassurance about a structural problem. It mistakes defensive lag for architectural resilience. The cybersecurity boom is not evidence that AI has a job-killing limit—it's evidence that AI is currently creating more security surface area than it eliminates, and that this gap is being papered over with human labor. That labor will itself be automated. The question is timing, not permanence.

DT verdict: The article diagnoses the fog. It misses the machinery underneath.