Leonids

hello eyb

[Leonids @leonids1115]

@YangFanYun GLM-5.2 is the best open-weights model I’ve used so far — not just on benchmarks, but in actual day-to-day use. This result matches the experience.

[Leonids @leonids1115]

I put ChatGPT Work + GPT-5.6 Sol to work building an AI wuxia sandbox: game engine, KP mechanics, cards, NPC cognition, UI, and eval rubrics. The AI doesn’t just write dialogue—it forms beliefs, changes strategy, and creates a different jianghu every run.