LLM & AI for developers who ship, not read announcements. Model releases, API changes, pricing per token, benchmarks that survive contact with production. What it means for the person who has to run it.