AI CUDA Engineer: Agentic CUDA Kernel Discovery, Optimization and Composition

This was debunked - the agent was actually fooling the verification harness <a href="https:&#x2F;&#x2F;x.com&#x2F;SakanaAILabs&#x2F;status&#x2F;1892992938013270019" rel="nofollow">https:&#x2F;&#x2F;x.com&#x2F;SakanaAILabs&#x2F;status&#x2F;1892992938013270019</a>. One particular test that showed a 150x speedup is actually 3x <i>slower</i>.

Nvidia is doing work like this internally: <a href="https:&#x2F;&#x2F;developer.nvidia.com&#x2F;blog&#x2F;automating-gpu-kernel-generation-with-deepseek-r1-and-inference-time-scaling&#x2F;" rel="nofollow">https:&#x2F;&#x2F;developer.nvidia.com&#x2F;blog&#x2F;automating-gpu-kernel-gene...</a>