Skip to content
X · @teortaxesTex · X / Twitter

Each and every open lab is working on this. Moonshot derisked Muon, DeepSeek did the most architecture R&D, ZAI seems to be at the frontier of SWE RL,…

Each and every open lab is working on this. Moonshot derisked Muon, DeepSeek did the most architecture R&D, ZAI seems to be at the frontier of SWE RL, Qwen on vision, Minimax on videogen, maybe Dots Lab on ARC-style reasoning… Thus the virtuous cycle goes faster and faster.Zain: im gonna need kimi pre-training + deepse