Stockfish is an open-source chess engine. Recent coverage describes it being used in experiments: a researcher distilled Stockfish's value function into a ResNet/ViT model using 1 billion positions from the Gigafish dataset and released a 3.9 billion-position dataset built from 37 months of Lichess games; a developer combined Claude with Stockfish to explain games at roughly $15 per analyzed game; and an open-source Jev-like model's chess checkpoint scored 4 wins, 46 draws and 0 losses against a random mover.
Open 3.9B chess dataset and Stockfish distillationA researcher distilled Stockfish's value function into a ResNet/ViT model using 1 billion positions from the Gigafish dataset, and released a 3.9 billion-position dataset built from 37 months of Lichess games. The CNN learned faster early on than the ViT, but combining both gave the best results.
Claude Code Skill Analyzes Your Chess GamesA developer shared an experiment showing Claude can play chess using vision instead of PGN notation, and combined with Stockfish can explain games. The system takes live audio or text notes plus vague instructions and, after about an hour, returns a commented game video; one analyzed game costs roughly $15 at API prices.
Reverse-engineered Jev-like model releasedAn open-source repo releases a Jev-like model that scores multiple text options in one pass, with Doom and chess examples. Reported results: Doom checkpoint averaged 0.60 kills and -97.50 reward over 10 episodes; chess checkpoint scored 4 wins, 46 draws, 0 losses vs a random mover but 0 wins, 2 draws, 48 losses vs Stockfish level 0.
The A.I. Daily