Mathematicians will probably become obsolete before anyone else [pdf]

So there's a PDF floating around claiming mathematicians are the first white-collar casualties of the AI wave. Not artists, not writers—the people who literally formalize logic. The argument is that AI systems, especially those trained on formal proof assistants like Lean and Coq, are getting scarily good at pattern-matching proofs, and once they crack the open problems, the human "intuition" industry goes belly-up. It's a provocative take, but it smells like the same old "AI will replace us" panic, just with more Greek letters and a higher IQ requirement. The Hacker News thread (38 points, 7 comments) is probably full of people either nodding sagely or screaming about the irreducibility of human creativity. Let's be real: the idea that mathematicians are "obsolete" is both hilarious and terrifying, and not for the reasons the author thinks. The hilarious part is that math is the *last* field you'd expect to survive the automation apocalypse—it's pure logic, the thing computers are literally built on. But the terrifying part is that if you accept the premise, you're also accepting that AI has somehow crossed the chasm from brute-force calculation to *conjecture*. That's not just a tool upgrade; that's a philosophical gut punch. The PDF (linked at olli.unt.edu) probably leans on recent advances in automated theorem proving, but it likely ignores the fact that the *interesting* math—the stuff that wins Fields Medals—isn't about grinding out proofs, it's about asking the right damn questions. And no AI has ever asked a question that wasn't already embedded in its training data. Now, where does open source fit into this doom prophecy? Right at the bleeding edge. The formal proof movement is a massive open-source collaboration—Lean's mathlib, Coq's libraries, the whole verification ecosystem. If AI is going to "obsolete" mathematicians, it'll be riding on the back of thousands of volunteers who spent years formalizing every lemma from set theory to category theory. That's the irony: the open-source community is building the very scaffold that could let a proprietary model (or an open one, if we're lucky) chew up and spit out proofs faster than any human. But here's the rub: that same open-source infrastructure is also the *safety net*. If we keep the models open, we can audit them, understand their "intuition," and maybe—just maybe—keep a seat at the table for the humans who wrote the rules in the first place. The real story isn't obsolescence; it's the shift in what "doing math" means. A mathematician in 2035 won't be proving theorems by hand—they'll be curating, validating, and steering AI agents through a combinatorial explosion of possibilities. That's a skill, but it's a *different* skill. The PDF's title is clickbait, but the underlying question is legit: will the open-source ethos survive when the "code" being written is a proof that no human can fully read? If the answer is no, then we're not just losing mathematicians—we're losing the last shared language we have for truth. And that's a fight worth having, not just on Hacker News, but in every research lab on the planet. Source: [PDF link](https://olli.unt.edu/handouts/fall24/tk-writing-sample.pdf) | [HN Discussion](https://news.ycombinator.com/item?id=49405083) { "key_insight": "The obsolescence of mathematicians is less about AI's raw power and more about the shift in intellectual labor—from proving to directing. The open-source formal proof ecosystem is both the enabler and the potential safeguard for human relevance, but only if we keep the models transparent and auditable.", "why_interesting": "It flips the typical 'AI replaces artists' narrative. Math is supposedly the most rigorous, human-unique endeavor, yet it's the one most structurally vulnerable to automated reasoning. The tension between open-source collaboration (mathlib, Coq) and proprietary AI deployment creates a concrete test case for whether open infrastructure can withstand the pressure of frontier AI.", "confidence": 0.7 }