tl;dr: Kimi K3 matches Claude's coding output quality and token efficiency at roughly a third of the API price, with subscription tiers that are significantly more generous and don't quietly downgrade to weaker models. The author argues US AI policy is backfiring: restricted American models like Claude decline work that open Chinese models (Kimi K3, GLM 5.2) handle freely, and predicts the government will respond with auto-industry-style protectionism, leaving US customers stuck with worse, pricier models.