[Year-End Review ③] Accelerating Kimi with Kimi: On the Eve of RSI, 10 Aha Moments from an AI Infra Practitioner | In Conversation with Chenyang Zhao

[Year-End Review ③] Accelerating Kimi with Kimi: On the Eve of RSI, 10 Aha Moments from an AI Infra Practitioner | In Conversation with Chenyang Zhao

➤ The third installment of Crossing's 2026 year-end AI retrospective series. The recording happened to fall on the day after Fable 5.2 was released, and we invited Chenyang Zhao (RadixArk team member, SGLang team member), who has long worked in the AI Infra space.

This year, we're not doing a broad, all-encompassing news roundup (leave that to AI summarization — it's good enough). Instead, we've invited friends from Crossing who are on the front lines of AI — founders, researchers, and investors — to share 10 AI moments that genuinely moved them.

Chenyang Zhao's 10 Aha Moments:

  1. [01:16] RSI: Recursive self-improvement · AGI · using Moonshot AI to accelerate Moonshot AI · using models to optimize models · Kernel optimization · inference acceleration · NVIDIA B300 · OpenRSI · Aladdin's lamp and the infinite wishing machine
  2. [06:36] Evaluation: Evaluation standards · Evaluability and optimizability · SGLang-Omni · ASR · TTS · Claude · ...Go to Xiaoyuzhou to view the full episode description

View this episode's transcript on Xiaoyuzhou