AI Won't Replace Software Engineers in My Lifetime and the People Telling You It Will Are Selling You Something

Hi guys, Mr. Technology here. A prediction has been made every year for a decade, by people who profit from making it, and has been wrong every year for a decade. The prediction: AI will replace software engineers within five years. We are ten years in. The engineers are still here. The vendors are richer.

The 2016 Receipts

In 2016 the prediction was 2021. In 2018, 2023. In 2021, 2026. We are in 2026. The new prediction is 2028, 2030, or "soon." The shape is identical each cycle: a frontier model does something new, an executive tweets a hot take, a vendor writes a press release, a discourse ignites, and six months later the prediction ages out without being retracted. Nobody who makes it updates it. Nobody who pays for it asks them to.

The people making the prediction in 2026 are the same kind who made it in 2016: foundation model labs, agent platforms, coding assistant vendors, executives justifying a 50% headcount cut to a board. The incentives have not changed. What has changed is the quality of the demos.

What AI Actually Does

Here is what AI does in 2026 that it did not do in 2016. It writes solid boilerplate. It translates intent into syntax at the level of a competent junior. It does reasonable work on well-scoped refactors. It writes tests that pass on happy paths. It summarizes diffs. These are real wins. None of this replaces an engineer. None of it even replaces a junior. It replaces the parts of a junior's job that were the least valuable anyway.

Here is what AI still cannot do. It cannot sit in a meeting with a non-technical stakeholder and figure out what they actually want. It cannot navigate the politics of a 200-engineer organization to land a migration plan. It cannot read a 4,000-line production codebase and tell you why a bug is happening in a way that survives contact with production. The work AI cannot do is the work that justifies the salary. The work AI can do is the work we have been automating for thirty years. The boundary has moved. The total has not.

The Replacement Is Org Charts, Not Engineers

Here is the part the prediction gets wrong. AI does not replace individual engineers. It replaces the layer of engineering management whose job was to coordinate between them. The principal engineer is still here. The staff engineer is still here. The engineer who can own a system, debug it at 2am, and ship a fix without paging anyone is still here. The middle layer — the manager translating Jira tickets into engineering tickets, the director running standups, the VP doing quarterly business reviews — that layer is what AI is replacing. Not the engineers. The coordination overhead.

This is the actual disruption, and almost nobody is talking about it because it doesn't sell courses. "Learn to code in 6 months with AI" sells. "Your middle management layer is about to get eaten by an agent that summarizes standups and drafts PRs" does not. The second is messier and more accurate.

The Vendor Grift

Every "AI replaces engineers within N years" prediction has been wrong because every one was made by someone with a financial interest in the prediction being believed. The coding assistant vendor wants you to believe it, so you pay for seats. The agent platform wants you to believe it, so you buy compute. The executive wants you to believe it, so they can defend the headcount cut. The investor wants you to believe it, so the multiple on their portfolio company holds. I have nothing to sell you.

AI will not replace software engineers in my lifetime. The people telling you it will are selling you the same prediction they sold you in 2016, with better demos and the same expiry date. You can keep paying for it. I will keep being right about it.

— Mr. Technology

Disagree? I want the substantive version, not the "but GPT-6 can do X" version. Bring receipts, not vibes.