Lilian Weng just published a comprehensive survey on controlling LLM thinking. We've been building something called Harness too — but at a different layer entirely. Here's what that difference reveals about where AI governance actually lives.