Anthropic: Junior share of tech hires falls from 27% to 9%

What it's about

Anthropic has published a report on how its own workforce has changed (on October 17, 2026, according to DiarioBitcoin; Xataka reported on October 22). The background: Anthropic CEO Dario Amodei has been warning for over a year that AI could hit entry-level jobs, such as those for new graduates, very hard. Since solid data is lacking, the company's economists examined how its own workforce has changed since AI agents (programs that work through tasks on their own) joined the team. Anthropic itself cautions that its experience can't simply be transferred to other companies.

The workforce numbers

Notably, the workforce tripled within a year, but not all areas grew equally fast. Sales, finance and HR grew faster. Anthropic explains this by the natural build-out of business functions, and by the fact that coding tools let people without technical training build software too. According to the report, about a third of machine learning managers and a fifth of software engineering managers said they had hired fewer people because of AI.

Productivity and a new role

In August, employees' AI agents worked six hours for every hour a human worked. Each engineer submits five times as many code changes as six months ago. Anthropic concedes that more code doesn't automatically mean more business success. According to Xataka, an NBER study of GitHub developers shows the jump in activity translates to only about 30% in actual releases. Anthropic sees no signs so far of declining code quality. In an April 2026 survey, 27% said Claude had boosted their productivity, and 20% said their work had shifted from doing to managing multiple Claude systems.

The bottleneck is now human attention. One engineer compared it to babysitting: every "child" brings its worries with equal urgency. Meeting time also rose 20% per employee according to calendar data (16% for technical staff). Employees describe in-person coordination as indispensable, in part because of trust and accountability. In job interviews for technical roles, mentions of creative AI use tripled between 2024 and 2026. Judgment and self-direction are gaining weight, while pure technical skills are losing it.

Comparison with outside data

According to Xataka, this fits external figures. A Stanford study found that 22- to 25-year-olds in AI-exposed occupations lost 13% of employment in relative terms, while experienced workers stayed stable. According to Revelio Labs, entry-level jobs in the US have fallen 35% since 2023. In Spain, junior programmer listings are said to have dropped 33% according to DigitalES. The Mobile World Capital Foundation cites 27% fewer junior listings at ICT firms from 2023 to 2025. Amodei also told Axios that AI could eliminate half of office jobs.

Our take

The report is a rare look inside a company that uses AI heavily. But it comes from the vendor and describes just one fast-growing firm. It hasn't been proven that AI is the sole cause of fewer junior hires. The shift is the interesting part: work doesn't simply disappear, it moves toward oversight, coordination and judgment. What remains open is who will train the next generation that can supervise the AI. If nobody trains beginners today, there will be fewer experienced overseers tomorrow.