CertiK and Columbia University Bring Math Proofs to the GPU Code Behind Every Large AI Model
This story was originally published on HackerNoon at: https://hackernoon.com/certik-and-columbia-university-bring-math-proofs-to-the-gpu-code-behind-every-large-ai-model. CertiK and Columbia's TensorLean uses machine-checked proofs to verify 100+ production AI kernels from DeepSeek and FlashAttention, finding 25 bugs. Check more stories related to undefined at: https://hackernoon.com/c/undefined. You can also check exclusive content about #certik, #columbia-university, #gpu, #software-engineering, #technology, #web3, #good-company, #cybersecurity, and more. This story was written by: @ishanpandey. Learn more about this writer by checking @ishanpandey's about page, and for more stories, please visit hackernoon.com. CertiK and Columbia University researchers have built TensorLean, a system that uses machine-checkable proofs written in Lean to confirm that the GPU code behind AI training and inference computes what it should. The team verified more than 100 production kernels across several GPU generations and found 25 bugs, 23 in FlashAttention and two in DeepSeek's kernels, while Big Tech plans about $725 billion of AI spending in 2026.
