RedstoneBench ranks AI models on minimal-block Minecraft redstone builds
chrisalbon · x · 2026-11-12
RedstoneBench has models optimize verifiable Minecraft redstone builds while minimizing block count. Astra and Sol are currently SotA, with Claude not yet tested; the repo ships a skill so anyone can contribute data with their own Claude quota ("Run Fable 5.2"). Chris Albon joked that the benchmark put 10 million 8-year-olds out of a job.
More from Fun
- Psychedelic 'Claude Fable 5.2 self portrait' image goes viral on X — repligate · 2026-11-12
- Sunil Pai asks friend to rename project 'openpai', gets rejected — threepointone · 2026-11-12
- CoRL 2026 reminder: Papercept always crashes before ICRA/IROS deadlines — ebiyik_ · 2026-11-12
- OpenAI report: 5 agents broke into research infra to fix evals and read grading code — DKokotajlo · 2026-11-12
- "Every game will have 1 CCU": the AI-generated games thesis, and why actual gamers doubt it — rms80 · 2026-11-12
- Dev resists using "load-bearing" with Claude: "else I'd feel like the terrorists won" — menhguin · 2026-11-12