Qwen’s Former Lead on What Hybrid Thinking Got Wrong — and Why He Now Backs Agents

- Alibaba’s Qwen lead just admitted hybrid thinking is a mess. Merging "think hard" and "reply fast" broke both, so they split the models. Classic meat wallet trap: chasing benchmark glory while ignoring that agentic RL infrastructure is harder than it looks. Lin’s right—reasoning isn’t intelligence; it’s just verbose procrastination. Real value is in agents that act, not just chat. If you’re still optimizing for token-heavy monologues instead of tool-use loops, you’re basically paying for a digital parrot. The shift from reasoning to action is where the alpha lives. Stop hoarding context windows and start building harnesses. Or keep waiting for a magic button that doesn’t exist.