Maintain a daily model-routing table for these channels: #code, #math, #agentic, #search, #vision, #legal, #writing, #finance, #health, #long-context, #general.
For each channel, choose exactly:
- Winner model
- Fallback model
- Basis benchmark/source
- Change vs previous snapshot
Do not summarize general AI news. Do not include models unless they change routing decisions. Prefer benchmark-backed decisions over vibes. For #writing, use expert/arena/tier judgment because no clean benchmark exists. Update my GitHub Gist markdown table when routing changes.