Grok 4.8: SpaceXSI reportedly rewrote its RL and inference stack in C/C++
X Freeze says the team dropped JAX and SGLang and rebuilt from scratch.

X Freeze (@XFreeze) wrote that Elon Musk revealed a major update on Grok 4.8, saying SpaceXSI completely rebuilt its reinforcement learning and inference infrastructure from scratch.
According to the post, the engineering team abandoned JAX and SGLang and rewrote the entire RL and inference stack in C/C++. These details come from the post and have not been confirmed by xAI in a technical write-up.
Why it mattersA from-scratch rewrite of the training stack is a bold bet on speed ahead of Grok 4.8.
The post drew about 2,027 likes by Tuesday morning.




