Musk at TomorrowX Summit: Grok 4.8 pretraining done, now in reinforcement learning

FuSheng_0306 · x · 2026-11-19

A thread relaying the first 3 of Musk's 24 takeaways from TomorrowX Summit (self-described as fact-checked from an investment notes account):

Later posts in the same thread cover his anti-benchmark stance, Grok's FSD-style training loop, the C/C++ infra rewrite, and an Optimus demo preview.

Related event: Musk Summit Highlights: Grok 4.8 Pretraining Done, Starship Reusability Vision(2 posts)→

Original post →

More from Companies & People

Companies & People channel →