Hong Letong
24-year-old Chinese founder of Axium, MIT/Oxford/Stanford alum building AI for Math.
“In December 2025, Axium Prover achieved a perfect score on the Putnam. That's the 6th perfect score in 98 years — the first by an AI.”
Source→“From one intern starting secretly in 2021 to AlphaProof in 2024 scoring 28 points at IMO — one point from gold. That was a cross-century moment.”
Source→“Amazon has one of the world's best automated reasoning teams. Without AI, they spent three to five years writing 260,000 lines of theorem-proving code to verify the memory isolation component of their hypervisor.”
Source→“Howard Morgan was just more optimistic about our business model than I was. He was telling me what my business model was... He's a mathematician, still teaches at NYU sometimes.”
Source→“Terence Tao has many podcasts and writings about formal proofs. My friend Kenny was one of the 5–7 students who built the foundations of Mathlib from scratch.”
Source→“Evan looked at one problem, drew a picture, and we all said 'oh, you've solved it.' The AI found no such elegant solution — it produced thousands of lines of Lean through brute-force case analysis.”
Source→“We're not aiming for AGI. We're aiming for ASI — Specialized Superintelligence. We're going straight for 'prove the Riemann Hypothesis' on the edge of the plate.”
Source→“Morgan Prize is like a thing in AI. Levon Alpert at Anthropic, Greg Young who is an xAI co-founder — the Morgan Prize winners have been flowing into AI.”
Source→“I respect my friends doing AI for Science. But if you want fast iteration with computation and logic, math and code are the better sandbox.”
Source→“Axium built 12-13 custom tools to make Lean (the formal language) run properly, including a proof verifier that is 100x faster than the community standard.”
Source→AI-extracted from podcast / newsletter / paper summaries. May contain errors.