You’ve probably noticed the cold sweat forming on the brows of academics lately. Terry Tao, perhaps the world’s greatest living mathematician, recently pointed out a terrifying reality: AI is making proof generation cheap. For centuries, the mathematical proof was the ultimate gold standard of truth. It was the fortress. Now, the fortress is a McDonald’s drive-thru, and AI is serving up theorems faster than you can write them down.
We didn’t optimize for truth; we optimized for a grading rubric. And now AI is gaming the rubric.
This is Goodhart’s Law in its purest, most brutal form. When a measure becomes a target, it ceases to be a good measure. Mathematics over-optimized itself around the proof because proofs are measurable. You can publish them. You can hand out tenure over them. But now that AI can crank out these measurable deliverables, the academic world is scrambling. Their grand retreat? “Well, math is more than proof! We have exposition! We have intuition! We have problem-finding!”
But let’s be brutally honest. Jacob Tsimerman recently predicted we might have superhuman AI expositors by April. Yes, April. So what happens when the AI doesn’t just prove the theorem, but explains it better than you ever could? What happens when it finds the most elegant, intuitive path through the math, leaving you in the dust?
Retreating to “human intuition” isn’t a strategy; it’s a hospice for obsolete professionals.
Don’t kid yourself—this isn’t just about mathematicians. If you are a knowledge worker, you are in the exact same boat. You thought your value was in the code you wrote, the reports you generated, the analyses you compiled. Those were just your “proofs.” Those were your measurable targets. And AI is automating them right now.
The push to celebrate “understanding” over computation is just a defensive retreat into a fortress that AI will eventually breach. The real question isn’t “What can’t AI do?” The real question is: why on earth did you define your human existence by your competence at a task a machine can do?
If your entire value as a human is tied to how efficiently you produce a deliverable, you were never a professional. You were just a very slow machine.
Stop trying to out-compute the computer. Stop looking for the one niche task AI can’t do yet. The shift isn’t from measurable deliverables to unmeasurable judgment. The shift is from doing to being. AI can have the proofs. It can have the exposition. It can have the whole damn rubric. Your worth was never in the output.
FAQ
Q: Won't AI always lack true human intuition and problem-finding skills?
A: No. Experts predict superhuman AI expositors by April. Retreating to 'intuition' is just moving the goalposts. AI will eventually optimize for taste and insight just as well as it optimizes for syntax.
Q: What's the practical implication for regular knowledge workers?
A: Stop tying your worth to measurable deliverables like reports, code, or analyses. AI will automate them. You must shift your focus from 'producing the output' to asking 'why the output matters' in the first place.
Q: Isn't this just cope for humans losing to machines?
A: Yes, if you're still trying to compete on competence. The contrarian take is that we shouldn't fight for the 'uniquely human' scraps of a broken system. We should let AI have the rubric and redefine our value entirely.