Alexandr Wang admits Muse Wonder wasn't trained on 3D game demos, insists it excels at real coding
alexandr_wang · x · 2026-10-13
Meta's chief AI officer Alexandr Wang confirmed the team did not explicitly train Muse Wonder on 3D demo games, admitting in retrospect they probably should have done a bit of that. Quoting theo's take that flashy 3D game demos don't represent real code work, he urged developers to judge the model on actual coding tasks instead.
- Muse Wonder's 3D demos drew comparisons with Astra and Opus, which were RL'd on 3D tasks
- Wang: no regrets, try Wonder on your real coding work
Related event: Meta AI Chief Admits Muse Wonder Wasn't Trained on 3D Game Demos(2 posts)→
More from coding & agent
- MecAgent 2.0 launches as its maker claims 'CAD is about to get solved' — Scobleizer · 2026-10-13
- Dev finds ~25% of AI-generated code introduces regressions in .NET 8+ — lucasmeijer · 2026-10-13
- Investor says coding agents left most software firms badly overstaffed, cut now or be disrupted — doodlestein · 2026-10-13
- The Turing Post to Host Free Online Event on Practical AI Agent Memory, Oct 13 — TheTuringPost · 2026-10-13
- New llama.cpp release adds multimodal model support and Metal performance gains — vboykis · 2026-10-13
- Using Claude in Cursor to Generate a History-of-Music Video — prasenx · 2026-10-13