AI's Employment Damage Is Landing on the Youngest Workers, New Research Finds

Tenth iteration of the Zero Cost Thesis reports a 16% relative employment decline for workers aged 22–25 in AI-exposed occupations — and finds the technology getting cheaper faster than institutions can absorb it

THOUSAND OAKS, Calif., July 31, 2026 — Artificial intelligence is displacing work at the entry level faster than anywhere else in the labor market, according to the tenth iteration of a long-running research program published today and available through the synED research library.

The report, The Zero Cost Thesis: A First Principles Analysis, is produced by the State of the Singularity research program at ListenToSee, Inc. It draws on payroll microdata analysis from Stanford's Digital Economy Lab showing roughly a 16% relative decline in employment for workers aged 22 to 25 in occupations most exposed to AI — the most rigorous displacement evidence the program has recorded in six months of tracking.

The finding matters because of where it lands. Displacement is concentrating at the point of entry, among workers who have not yet built the judgment that currently protects more senior practitioners. The same data shows demand for experienced people holding up or rising. The effect is an economy that still wants skilled workers while removing the rungs by which people become skilled.

Findings

The report tracks ten premises across ten iterations since January. Four findings define the current window:

Abundance is arriving as a price collapse, not a concentrated breakthrough. Anthropic's Claude Opus 5 shipped in July at half the price of its own frontier tier. A Chinese laboratory worth roughly two percent of its largest rivals reached the same performance frontier and published the model weights.

For the first time in ten iterations, the broad, diffuse scenario is now more likely than a concentrated AI takeoff — 36 to 46 percent against 35 to 45 percent. Capability is spreading faster than it is consolidating.

Claims that AI has begun improving itself without human direction did not survive review. Three separate such claims were examined; none held as stated, and two were contradicted by the technical participants of the panels making them.

The binding constraints are physical, not computational. Electrical transformers now carry multi-year lead times, and memory scarcity is forecast to persist past 2030 — beyond the window in which the transition is supposed to occur.

Why it matters for the workforce

synED works at the junction where education and employers meet, including through its Cyber-Guild® program for cybersecurity talent development. The report's entry-level finding speaks directly to that work: if the first rung of the career ladder is the rung being removed, the pipeline problem is no longer about supply of candidates but about where practitioners acquire judgment at all.

The report also finds that the transition is running well ahead of any mechanism designed to absorb it, placing the probability of a disorderly transition at 90 to 96 percent — the highest the program has recorded.

On method

The program's stated purpose is not to refute forecasts of AI abundance but to test them. Every source in the published report was located through a live search result and then independently fetched: 99 sources, of which 84 resolved to the expected content, 15 were blocked to automated retrieval and are marked as such, and none were dead.

That verification pass corrected the report in roughly twenty places, and the corrections are published rather than quietly absorbed. The entry-level figure itself was revised upward from 13% to 16% after the earlier number was traced to a superseded draft.

Quotes

“We did not set out to argue that the optimists are wrong. We set out to ask them to prove they are right, and to describe the transition they are asking everyone else to live through,” said Scott Young, who leads the State of the Singularity research program at ListenToSee. “This window is the first where a real part of that proof arrived. It arrived as a price collapse, which is a better world than the concentrated one — and a much harder one to govern.”

“The number that should concern every workforce program in the country is the one about twenty-four-year-olds,” Young added. “Cheap intelligence was always going to be the easy part. Everything unfinished is the part its advocates never costed.”

Availability

The Zero Cost Thesis: A First Principles Analysis, tenth iteration, is available through the synED research library at syned.org. The report runs 35 pages and includes a fully cited source appendix, eleven data figures, and a published record of its own corrections.

About synED

synED connects education and industry to strengthen the technology and cybersecurity workforce, including through synED's Cyber-Guild® program. Accomplish More. Learn more at syned.org.

About ListenToSee, Inc.

ListenToSee, Inc. is a governance, risk and compliance consultancy. Its State of the Singularity research program maintains a continuous, adversarially reviewed record of AI capability, cost and workforce impact, now in its tenth iteration.

Media Contact

synED

P: 866.420.4573 ext. 801

E: [email protected]

W: syned.org