In 1999, friends helped him land a humble, non-tenure lecturer gig at the University of New Hampshire.
Finally having a quiet office, he set his sights on a monster: The Twin Prime Conjecture.
Prime numbers (like 3 and 5, or 11 and 13) sometimes appear in pairs just 2 units apart. Mathematicians since Euclid believed there are infinitely many such pairs—but for millennia, no one could prove prime gaps didn't just grow to infinity.
At 58, he was working at a Subway sandwich shop to pay rent.
He had no permanent academic job, lived in a cheap apartment, and had published only one paper in his entire life.
Weeks later, he solved a 2,000-year-old math mystery and shocked the world.
The greatest underdog story in modern science: Yitang Zhang 🧵👇
Most people would have given up on mathematics forever. Zhang didn’t.
Even while making sandwiches, his mind was living in a different universe: analytic number theory.
He didn't need a lab. He didn't even like taking notes. He paced around alone for hours, running complex proofs entirely inside his head.
After a severe fallout with his PhD advisor in 1991, Zhang graduated with zero recommendation letters. In academia, that is a death sentence.
For the next 7 years, one of the sharpest minds on Earth survived by: • Delivering Chinese food • Working at a motel in Kentucky • Keeping books for a Subway franchise • Sleeping in friends' basements and his car
In the 1980s, Zhang was a rising star at Peking University—China’s top institution.
He came to the US in 1985 for his PhD at Purdue. Brilliant and ambitious, everyone expected him to conquer the math world.
Instead, he vanished into obscurity.
最近看大家每天都在焦虑:AI 一天一个新模型,不学怕被淘汰,学了发现硬件在涨价、工资在缩水、人比以前更累了。
从数学优化的角度看,全人类正在陷入一场巨大的“梯度下降陷阱”:
每个人都用 AI 把自己的局部斜率卷到了极致,但因为所有人都顺着同一个负梯度方向狂奔,整个行业反而以最快的速度掉进了一个更拥挤、更内卷的局部极小值(Local Minimum)。
在大家都用 AI 追求“均值回归”的时代,真正值钱的反而不再是平滑的算力,而是你身上那些 AI 拟合不出来的“随机噪点”和“偏离度”。