A Mathematician's Controversial Thesis: The World Doesn't Need This Many Mathematicians
Elite mathematician Jiang Zilin from Peking University argues that funneling the brightest minds into theorem-proving could paradoxically slow human progress.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Grading isn't scoring—it's negotiation and translation
IMO scores aren't handed down by judges; they're negotiated. Country leaders score independently, coordinators score independently, and when they disagree, they hash it out on the spot. Leaders have to ‘translate’ messy handwriting into the student's actual intent—when the same idea sprawls across different parts of a test, you have to assemble it like a montage. Even names get negotiated: the dragon problem was originally called ‘嫦娥’ but English speakers couldn't pronounce it, so it became ‘change’. The whole exam has to account for the linguistic habits of 120 countries.
— Jiang ZilinIMO gold selects for the longest chain of thought
About half of Fields medalists tried IMO young. An IMO gold medalist's odds of later winning the Fields Medal are 50 times higher than a math PhD from a top university. IMO doesn't test knowledge volume—it tests whether you can chase a complex chain of reasoning to the end under pressure, without giving up midway. In modern terms, it's filtering for chain-of-thought length. What IMO catches early is a kind of mental maturity that shows up ahead of schedule.
— Jiang ZilinThe 5% of competition winners who became entrepreneurs created the most unicorns
Over 25 years, researchers tracked about 18,000 IMO/IOI/IPHO medalists. Career split: 36% stayed in academia, 22% went into tech (Google hired the most), 12% into quant finance, and just 5% chose startups. But that 5% minted roughly 20 unicorns and 7 decacorns—OpenAI, Stripe, Databricks, Perplexity among them. This cohort's odds of becoming a billionaire are 1,500 times the baseline; their odds of founding a unicorn are 4,000 times baseline. The pattern traces back to clustering: the same schools (MIT, etc.) admitted them all, and proximity bred compounding advantage.
— Jiang ZilinDeeply studying an unrelated proof solved the 40-year-old conjecture
The ball-covering conjecture sat unsolved since 1973. His postdoc's first 8–9 months: nothing shipped. Then he and a colleague fell in love with an old 1950s proof on ‘flat tiling’—not because it was useful, just because it was beautiful. They mastered it so thoroughly they'd explain it to anyone who'd listen, no agenda. One day the colleague said: what if we tried that technique on the ball-covering problem? He immediately saw how to split the problem into two cases; one case yielded to the old proof exactly. He sketched the approach the same afternoon, confirmed the details at 11 p.m., and wrote up the whole proof overnight.
— Jiang ZilinFled to Israel but remained under his advisor's shadow
His doctoral advisor had just cracked something on equiangular lines. He deliberately sidestepped it—‘you've got first-mover advantage already’. His postdoc move to Israel was supposed to be an escape from the advisor's domain. Later he realized he'd never left at all; he was still operating within the shadow's reach. The ball-covering work was his true exit. He had to make that leap. Otherwise peers read your CV and conclude you have no independence; you're just an appendage to your advisor. And a postdoc is time-locked—you've got a fixed window to prove you're your own mathematician.
— Jiang ZilinWe don't need this many mathematicians
He pushes a claim he knows people will fight: if society creates too many mathematician positions, it actually slows human progress. The brightest minds get vacuumed up to prove theorems when they could solve harder problems with real social returns. Mathematician value lies more in the byproducts of their thinking than in the theorems themselves. The planet has 81 billion people. Maybe 100,000 math jobs would suffice. The rest of that genius should go elsewhere.
— Jiang ZilinAI independently solved a problem that stumped his mentor
Hangzhou, 2025: his doctoral advisor delivered a lecture and left an open problem unsolved. Early 2026, GPT-5.5 independently cracked it. He felt genuine shock—at least in extremal combinatorics, AI had solved a problem that had blocked a world-class mathematician without any human prompting. But he also notes that Deng Yu has said the chain-of-thought depth AI needs hasn't yet touched the threshold of what he needs.
— Jiang ZilinGrowth comes through multiplication, not addition
He quotes the US math coach Robison: the old way was ‘addition’—one person at the frontier, grinding out one more contribution, then another, then ten. A breakthrough, sure. But what if your role shifts? What if you have ten excellent students under you? You multiply their capacity by two, by three. That's multiplication, not addition—much higher ROI. Next: build a system that runs without him in the room, so multiplication—or exponential growth—happens every day automatically.
— Jiang ZilinIn their own words · checked verbatim
There's an IMF report saying that historically about half of Fields medalists participated in IMO in their youth, and the probability that IMO gold medalists later win the Fields Medal is 50 times that of top-10 math PhDs.
就是IMF的一份报告 说历史上约 一半的菲尔兹奖得主 都在年轻时参加过IMO 然后IMO金牌得主 后来拿下菲尔兹奖的概率 是top10名校数学博士的50倍
Jia Xiaojie37:17
The probability of this group becoming billionaires is 1,500 times that of ordinary people; the probability of becoming a unicorn founder is 4,000 times that of ordinary people.
这群人成为亿万富翁的概率 是普通人的1500倍 成为独角兽创始人的概率 是普通人的4000倍
Jia Xiaojie40:19
When you finally output the tokens, sure it took maybe half a day, but the entire thinking process must have been weeks—maybe two or three weeks worth of work.
就是最后输出token的时候 肯定是只花了半天的时间 但是整个的这个思考的过程 肯定是有一个长达 可能两三周的过程吧
Jiang Zilin51:26
When you think up the proof before sleep, whatever you do, don't write it down—that way you can actually get a good night's sleep.
在你睡前想出来证明的时候 千万别把它写下来 这样你才能够睡好一晚上的觉
Jiang Zilin1:02:29
That's true, but I had to break out of that cage. From a career development angle, it was absolutely necessary. If you just follow your doctoral path forward, when the time comes to find an academic position, most peers will think you lack your own independence.
大概有这样 但是我跳出这个五指山最好 确实从职业发展的角度 必须要做这个事情 如果你只是沿着你博士的轨迹 进一步向前的话 最后找工作的时候 大多数同行会认为 你没有自己的独立性
Jiang Zilin1:13:32
Imagine this: if we create too many mathematician positions, it would actually slow down all of human society, because you're putting the most brilliant people into proving theorems and solving those kinds of problems.
你可以想象一下 如果我们产生了很多数学家的岗位 反而会使得人类的整个的社会进一步放缓 因为你把最最天才的人都放到了证明定理解决这些问题上
Jiang Zilin1:32:52
If we purely pursue a result, the result itself might not mean much—but the process of pursuit and the sparks of thought generated along the way, I think those are worth far more than the answer itself.
我们如果单纯的去追求一个结果 可能结果本身并不会意味着什么 而是这个追求的过程 以及追求过程中产生的这些思想的火花 我觉得价值是远远大于答案本身的
Jiang Zilin2:56:24
What we did before was addition—I'm one person rushing to the academic frontier, adding one more to it, adding one more, adding one more, or adding ten. When you've made a big breakthrough. But you can also realize that your role can shift—you can become someone who multiplies.
我们以前可能在做的事情是做加法 加法意思是我一个人 我去冲在学术的前沿 我要去为这个学术再加一 再加一再加一或者加十 你说做了大突破的时候 但是你也可以意识到你的角色可以转变 你可以做一个做成法的人
Jiang Zilin3:17:39
Figures
| Probability of IMO gold medalists winning the Fields Medal | 50 times that of top-university math PhDs | 37:17 |
| Proportion of Fields medalists who participated in IMO | Approximately one-half | 37:17 |
| Sample size of IMO/IOI/IPHO medalists tracked over 25 years | Approximately 18,000 | 39:19 |
| Percentage of IMO medalists who chose entrepreneurship | 5% | 40:19 |
| Probability of IMO medalists becoming billionaires | 1,500 times that of the general population | 40:19 |
| Probability of IMO medalists founding unicorns | 4,000 times that of the general population | 40:19 |
| Year the ball-covering conjecture was proposed | 1973 | 58:27 |
| 2025 national math team: total members and women | 60 total, 0 women | 31:14 |
Glossary
- Chain of Thought
- In problem-solving, the sequence of intermediate reasoning steps from problem to solution.
- tenure
- A permanent faculty position at a university that provides long-term employment security.
- equiangular lines
- A set of lines in high-dimensional space where every pair makes the same angle.
- Condensed Mathematics
- A new mathematical language framework introduced by Peter Scholze.
- extremal combinatorics
- The branch of combinatorics studying extreme values of combinatorial structures under constraints.
How to listen
Entrepreneurs, parents, and STEM researchers interested in math competition culture, academic career paths, and AI's impact on mathematical research.
The opening pleasantries and Fields Medal gossip early on carry less insight and can be skipped.