3d ago1 source0 reactions

Developer builds Claude Code tool to analyze chess games

62 BoomStory toneHobbyist capability demo, framed as experimental progress

By Splitfeed EditorialUpdated Sep 26, 2026, 18:54 UTC

1 source · Hacker News front page (AI)

Boom: Tool combines Claude vision and Stockfish engine to comment on chess games
Boom: System accepts live audio or text and outputs a commented game video
Doom: Processing takes approximately one hour and results are described as imperfect
Boom: Claude reads board positions via vision rather than standard PGN notation

The story in full

A developer posted to Hacker News on September 26, 2026 a self-built tool that uses Claude Code alongside Stockfish to analyze chess games from Lichess. The system accepts live audio notes or text instructions such as "analyze my last lichess game" and produces a commented video of the game, taking roughly one hour to deliver results.

The project began as an experiment testing whether Claude could process chess positions using vision rather than PGN notation. The developer found that combining Claude's vision capability with Stockfish's engine analysis produced usable game commentary, though they note the output is imperfect.

Analysis

335 words

By Splitfeed EditorialLast updated Sep 26, 2026, 18:54 UTC

On September 26, 2026, a developer posted to Hacker News a personal project that connects Claude Code, Anthropic's agentic coding tool, with Stockfish, the open-source chess engine, to generate annotated video analysis of chess games pulled from Lichess. The system accepts either live audio or typed instructions, with something as casual as "analyze my last lichess game" being enough to kick off the process. The output is a commented video of the game, though the developer notes it takes roughly one hour to produce and describes the results as imperfect.

The project's starting point is the more technically interesting detail: the developer wanted to test whether Claude could interpret chess positions through vision, reading the board as an image rather than consuming the standard PGN notation that chess software typically relies on. Finding that it could, they then layered in Stockfish's engine analysis to provide move-by-move evaluation, using Claude to translate that into human-readable commentary. The combination of a language model handling visual interpretation and natural language output alongside a deterministic chess engine handling deep calculation is a small but concrete example of hybrid AI workflows, where different tools cover different parts of a problem.

No reactions from the Pro-AI, Anti-AI, or Middle Ground camps have been published on this story yet. Typically, the Pro-AI camp would point to a project like this as evidence that general-purpose AI tools are becoming capable enough to handle domain-specific tasks without specialized fine-tuning. The Anti-AI camp would likely focus on the one-hour processing time and the developer's own admission of imperfect output, arguing that the effort does not yet match what existing dedicated chess analysis tools already deliver reliably. The Middle Ground camp would probably frame it as a promising experiment with genuine limitations, one worth tracking but not worth overstating.

The natural next question is whether the developer refines the tool further, particularly around processing speed and commentary quality, and whether others in the chess or developer communities attempt similar builds with different model and engine combinations.