tl;dr: GLM 5.2 from Z.ai is the first open-weights model the author finds genuinely competitive with Opus/GPT-5.5 for coding, at ~15-20% of the price, and it works as a drop-in replacement via OpenAI/Anthropic-compatible endpoints in tools like Claude Code. Current weaknesses are slower inference, no vision support, and poor web search, but switching costs are trivially low—threatening the high inference margins (~90% on compute) that fund frontier labs. Part two will cover the industry implications of collapsing inference margins.