Simon Willison's May 2026 PyCon US lightning talk summarized six months of LLM progress since the November 2025 inflection point, documenting five leadership changes among providers and the emergence of functional coding agents (Simon Willison, 19 May 2026). Model superiority transitioned from Claude Sonnet 4.5 (29 September 2025) to GPT-5.1, Gemini 3, GPT-5.1 Codex Max, then Claude Opus 4.5, evaluated via SVG pelican-riding-bicycle tests (Simon Willison, 19 May 2026). Reinforcement Learning from Verifiable Rewards applied by OpenAI and Anthropic produced coding agents crossing from often-work to mostly-work thresholds by November, paired with Codex and Claude Code harnesses (Simon Willison, 19 May 2026). The Warelay repository, initiated November 2025 by Pete, evolved into OpenClaw by February 2026, spawning the generic Claw category of personal AI assistants and driving Mac Mini demand in Silicon Valley (Simon Willison, 19 May 2026).