Qwen3-235B Advances, GPT-5 Teasers, and LLM Reasoning Progress. Hardware, Inference, and Training Bottlenecks. CLI Agents, Code Tools, and Ecosystem Wars. LLM Benchmarks: Markets, Collusion, and UI Generation. Structured Decoding, LLM Output Control, and WGrammar