How long have you been "planning to understand" how modern LLM inference works?

We just gave you a readable version of SGLang you can finish over the weekend.

Introducing mini-SGLang ⚡

We distilled SGLang from 300K into 5,000 lines. Kept the core design…