SGLang's New v1.0 Inference Engine Is Now Written by AI Agents for 95% of Merged Code

SGLang's new v1.0 Inference Engine is now written by AI Agents for 95% of Merged Code. Automated agents refactored 200,000 lines of code and managed 328 pull requests, resulting in a 2.1x speedup for its megakernels. SGLang is forming a consortium with Nvidia, AMD, Intel, Google, Meta, and Cloudflare.