shadcn reports weird eval results: Sonnet and Gemini pass the most tasks
shadcn · x · 2026-11-07
shadcn (creator of shadcn/ui) posted that "weird things" are happening in his evals, attaching a leaderboard screenshot showing Claude Sonnet and Gemini pass the most tasks, with other model names and figures redacted. Given his influence as an independent evaluator, the ranking shift is notable — but the post itself is thin; full details require the original link.
More from coding & agent
- The AI agent to-do list meme: fix a bug, solve the Riemann hypothesis, cure cancer — fofrAI · 2026-11-07
- A 4B router model beats fixed Opus 5.5 on Terminal-Bench, 77.78 vs 69.44 — YangYou1991 · 2026-11-07
- Testing and verification infra is quietly becoming a strategic advantage in the AI era — arpit_bhayani · 2026-11-07
- Indie dev shares AI-generated action RPG concept with top-down and first-person modes — Dimillian · 2026-11-07
- OpenAI built its own Git backend to keep up with agent-scale parallel coding — Hesamation · 2026-11-07
- Detecting browser-using AI agents remains an unsolved problem, say practitioners — teodorio · 2026-11-07