Evidence as of November 1, 2026. Every figure below comes from the broadcast record of the Comuvia News video published November 2, 2026, and carries the label that video gave it. Trackers that count differently are cited by name and date rather than reconciled into one number. This companion piece adds no evidence of its own. Not investment advice, and not career advice.
The headline is a crisis. On the record, the picture is narrower and stranger than the headline.
In September 2026 the unemployment rate for computer and mathematical occupations was 3.3%, below the 4.2% national rate, by the Bureau of Labor Statistics. Over roughly the same stretch, hiring of new graduates at the largest technology firms fell to about half its pre-pandemic level, by SignalFire. Both of those are true at once, and any account that uses only one of them is not describing the market.
The strain is real, and it is specific
This is not, in aggregate, a collapse. The Bureau still projects employment of software developers to grow about 10% over the decade to 2035, faster than the average job. What has frozen is the entrance.
New-graduate hiring at the largest firms is down by roughly half against the pre-pandemic years. Stanford's Digital Economy Lab finds workers aged 22 to 25 in the jobs most exposed to artificial intelligence sitting about 19% below where their employment would otherwise be — an effect that runs through slower hiring, not layoffs. The Federal Reserve describes a low-hire, low-fire market: few people let go, few taken on.
Above that frozen entrance, senior, security, cloud and AI skills are scarce while generalists and juniors are not. ISC2's workforce study reports 59% of security teams seeing a critical or significant skills shortage, up from 44% the year before.
So the market is not one market. It is a scarce top and a crowded bottom, and a single national number hides both.
A posting is not a hire
The clearest single picture of this labour market is the gap between vacancies and actual hires. In the Bureau's JOLTS survey for August 2026 there were about 7.1 million openings across the economy, but the hires rate was 3.3%, near the low end of the past decade. In the information sector — much of tech — about 123,000 openings produced just 44,000 hires, a 1.6% hires rate, half the national figure.
The listings are there; the hiring is not keeping pace. That gap feeds the ghost-job problem and the brutal funnel: roughly one employer in three admits to posting roles with no current intent to fill, and about 180 applications stand behind each hire.
What October added
Two findings landed after the story was first cut, and both complicate the simplest reading.
The displaced are being hired — by other industries, at higher pay. In the Randstad Tech Talent Report, built on 20 million job postings and a survey of 1,100 decision-makers across 11 countries, 55% of non-technology employers said they were recruiting workers released by the technology sector. Healthcare and life sciences offer up to 17% more than technology companies for a software engineer; banking offers 13% more. The demand did not vanish. It moved industry.
Some of the cuts are being undone. Orgvue found 32% of organisations that cut staff because of AI have since had to rehire. In a survey of more than a thousand American workers run for SHL, nearly one in five said they had been laid off or replaced by AI and then asked to return. Ford and Amazon both reversed AI-related cuts. Those asked back are not returning cheaply: almost half want a significant pay rise, 43% want written job-security guarantees, 40% want full transparency about the AI's role, and 30% want an apology.
Against all of that, the firm that counts the cuts reports hiring up about 25% on last year, and puts it plainly: while AI is shifting the labour market, it is not dismantling it.
How to read the numbers
They do not all mean the same thing. The Bureau's computer-and-mathematical grouping and CompTIA's tech-occupations grouping are different baskets, and both bounce month to month. The layoff totals genuinely disagree — Layoffs.fyi, Crunchbase and TechCrunch use different scopes and dates — so the video cites the tracker and the date rather than a single truth. Stanford's entry-level figure is a moving target, revised from 13% to 16% to 19%, and its authors call it early evidence, not proof.
The record fragments
Five recordings carry the argument, each from the official channel of the person or outlet named on screen, with original audio: Kevin Warsh, Chair of the Federal Reserve, at the FOMC press conference of October 28, 2026, on job gains keeping pace with the workforce; Dario Amodei on the alarm; Jensen Huang on why a task is not a job; a candidate's own account of the interview grind; and Erik Brynjolfsson on the oldest distinction in the field — machines that replace against machines that complement.
Comuvia's complexity judgement is labelled as a judgement, not as a finding. This is research and analysis, not investment or career advice.