
Who Becomes an AI Engineer — and Where They Go Next
We mapped the AI Engineer role from both sides — Skillenai's talent graph and job postings. Who feeds it, where they go next, and why the skills employers ask for don't match what workers list.
Insights on AI skills and the AI job market.
Get an email when we publish a new post. No account needed, unsubscribe anytime.

We mapped the AI Engineer role from both sides — Skillenai's talent graph and job postings. Who feeds it, where they go next, and why the skills employers ask for don't match what workers list.

AI didn't torch the junior job market. A frozen labor market did — and the housing market, frozen by the same 2022 rate shock with zero AI, is the control group that proves it.

Jellyfish says AI dev output plateaus around 4 merged PRs a week. Running Claude Code agents solo, after hours, I've been doing 23. The plateau isn't a token wall — it's the Coordination Tax.

US employers ask 2.5x more generalist software engineers than AI engineers to build with LLMs, and the generalists do it in TypeScript and Ruby, without LangGraph. A demand-side look at who really builds AI.

A viral Levels.fyi post says the median PM out-earns the median engineer. Across the open job market, that gap is a spam artifact plus a composition effect — match them rung-for-rung and the engineer wins every level.

Under identical titles, federal and private tech postings ask for different tools — a federal Data Scientist does statistics and reporting where a private one ships Python and experiments — and several modern data roles barely exist as federal jobs.

Federal tech jobs pay ~20% less than private — a gap that should drive workers out but doesn't. Using the new OPM workforce API, the Skillenai job index, and BLS, we price the public-sector bargain: near-zero mobility, near-total job security, and a pension that ties them together.

Middle managers' share of layoffs has climbed every year since 2020. We tested the two usual suspects — AI and remote work — against 97,500 US job postings. Both fail. The culprit that fits the timing is a title-inflation bubble we spent a decade blowing up.

I built Skillenai's chat on Claude Managed Agents — file, bash, sandbox, and our API — and I'll never go back to the days of orchestration frameworks. Our data shows the field is already flipping the same way: in tech news, low-level agent runtimes went from ~30% to 71% of framework-vs-runtime mentions across 2026, while hiring still lags 12:1.

On June 7 a single post on X coined "loop engineering." We measured it across 870,000 documents three days later — and found why you can't trust a buzzword count.