AI, layoffs, and the developer who stays valuable

Mid-2020s tech has two headlines stuck together: copilots that write code, and layoff rounds sold as "efficiency." For developers, the real question is simpler — is this still a craft you grow into, or a cost line that shrinks every time a demo hits the board deck?

The anxiety is rational — the blame often isn't

AI changes unit economics: junior-sprint work can become a prompt-plus-review loop. That sits on a cold hiring market after overhire and higher rates. But not every pink slip is "replaced by ChatGPT." Plenty of cuts are runway, margins, or dead products — with AI as convenient PR. The sharper shift is at the margin: slower hiring for automatable roles, higher bars for juniors, same team expected to ship more.

The broken ladder

The ugliest cost is the entry rung. Juniors used to learn taste via reviews, on-call, and ugly code. If a senior plus a model covers that grunt work, the industry eats its own training pipeline. Learn AI or be left behind — while the same AI may erase the first job that taught you.

What keeps a developer valuable

Don't abstain, and don't become a prompt tourist who can't debug the model's output. Use copilots for boilerplate; spend the saved time on what models still botch — requirements, architecture, security, incidents, and saying no. Review AI like a junior PR. Prefer work that leaves decisions (design docs, SLOs, migrations) over code someone else can regenerate. Fundamentals, domain depth, and taste beat shallow tool-chasing — the people who stay valuable are trusted with outcomes when the tool is wrong.

The point

Use AI tools where they make you faster, keep building the skills they can't replace, and help newer developers learn the job. Coding assistants will keep improving, but they will still be confidently wrong at times, and developers who can catch and fix those mistakes will stay in demand.