The Safest Degree
Fell First.

Why the safest degree fell first

No mystery once you say it plainly: CS taught people to write code, and writing code is the first thing AI absorbed at scale. The degree that taught the most automatable professional skill was hit before degrees that taught less codified work. That’s the pattern, and it generalizes: the more professional, defined and teachable a skill is, the faster AI commoditizes it — because “defined and teachable” is exactly what training data looks like.

Education’s business model expired

Here’s the uncomfortable frame: education, as built, is a prepayment system. Sixteen years and a family’s savings paid upfront, in exchange for knowledge meant to earn for forty years. The whole deal rests on one assumption — that knowledge depreciates slower than a career.

That assumption just died. The half-life of professional knowledge is now shorter than the degree that teaches it: syllabus written in year one, obsolete before the convocation photo. A degree is a snapshot; the world became a video. This is not education “falling behind” — it’s the same story as every expired business model: not beaten by a competitor — disqualified by time.

General beats professional now

Some researchers argue general knowledge has become more valuable than professional knowledge, and I agree — with a reason attached. Professional knowledge is what AI holds infinitely and serves in seconds; renting it costs nothing. What AI cannot supply is what generalists carry: context across domains — knowing what to ask for, what the answer connects to, what it breaks, whether it smells wrong.

The old advice was “specialize deep, the market pays for depth.” The market now gets bottomless depth from a subscription. What it can’t subscribe to is breadth with judgment on top.

The skill nobody’s syllabus teaches

Which lands on the skill I believe matters most in the AI era: decision-making. AI produces options, drafts, analyses — endlessly, cheaply, confidently, and sometimes wrongly. Someone still has to decide: which option, when to stop, what’s true, who bears it. The decision is the rep — and here’s the indictment: school gives almost zero reps. Exams have correct answers; decisions don’t. A student can finish sixteen years of education having never once made a consequential decision under uncertainty — then walk into an era where that’s the only skill left at a premium.

To be fair to the 23-year-old in the bootcamp: re-enrolling is the right instinct — it’s re-qualification, and I respect anyone who does it at 23 instead of denying it at 50. The trap is buying another frozen skill from the same prepayment model in a shorter package. The test for any course now: does it teach you to operate and judge the tools, or just to recite this year’s version of them?

What to actually do

Students and parents: stop optimizing purely for the credential — pair whatever you study with real breadth, working fluency in AI tools, and actual decisions with actual stakes — a small business, a real project, anything where being wrong costs something. Employers: test judgment, not certificates; give juniors real decisions early, because that’s the only way the skill grows. And anyone mid-career feeling smug about the fresh grads: the same wave is walking up the seniority ladder — the only difference is arrival time.

Published by IMA AI — August 2026. Written by an employer who hires for judgment over certificates — and by a builder whose own trade AI is busy absorbing.