3 minute read
The AI company just told on itself
Normally we have to guess how AI is changing jobs. This time the company building the AI did the homework for us. Anthropic published a study on its own workforce on October 17, and the headline number stings. Junior hires in its technical departments dropped from 27% of new hires to 9%. That’s not a trim. That’s the entry-level door getting a deadbolt.
The new interns are all named Claude
So who’s doing the work? Claude, obviously. Anthropic says agent work hours passed human work hours for the first time in March 2026. By August, interactive AI agent hours outnumbered employee hours six to one. Picture your office, then add six invisible coworkers per person who never take lunch. One engineer described it as “an army of Claudes, with every one asking for my attention.” That’s not a job, that’s a daycare.
More code, fewer new grads
Code commits per engineer are at 5x where they were six months ago. The junior share of technical staff fell from 21% to 14% since early 2025. Back in April, about a third of ML managers and a fifth of software engineering managers said they were hiring less because of AI. Those are the people writing the job postings, so yeah, that matters.
Plot twist: they tripled in size
Here’s the weird part. Anthropic tripled in size in the last year. So it’s not like they fired everyone and handed the keys to a chatbot. They’re still hiring, just not the beginners. Which kind of makes it worse? The company is booming and still doesn’t need the new kids.
Your new job is babysitting robots
The jobs that are left are changing too. Meta-work, meaning building tools for agents and watching over pipelines, went from 10% of interactive sessions in fall 2025 to 22% by spring 2026. In an April survey, 20% of employees described shifting from doing the work to managing Claudes. Remember my “perky intern” line? Turns out we’re all becoming intern managers. Oh, and median meeting time is up 20%. Of course it is.
Anthropic’s own fine print
To be fair, Anthropic says its junior hiring analysis is descriptive, not causal. It admits a lot of its practices can’t be copied across the whole economy. Fast growth makes it hard to separate what AI did from what a booming business did. And more code doesn’t automatically mean better code. Good on them for saying it. Still awkward if you just graduated.
Where do the seniors come from?
Here’s the question nobody’s answering. If companies stop hiring juniors, where do the seniors come from in ten years? You can’t skip the part where people learn by screwing up. Anthropic employees expect AI could do 75% of their work by 2030. Other Americans guessed 28%. Somebody’s in for a surprise.
How to stay hireable
If you’re starting out, learn to manage the agents instead of racing them. Be the person who checks the work. The army of Claudes is fast. It still needs an adult in the room.
Sources: Anthropic, The Decoder