Show HN: A Claude Code skill to analyze your chess games
This Show HN demonstrates a Claude Code skill that leverages Stockfish and user audio to generate personalized, narrated video post-mortems of chess games. It aims to make engine analysis more human-understandable and engaging than traditional methods, sparking discussion on the true capabilities and practical applications of LLMs in complex tasks. While innovative in its personalized approach, the project prompts debate about its cost-effectiveness and whether AI truly "understands" the game or just articulates engine output.
The Lowdown
This project, "chess-postmortem-skills," introduces a novel approach to chess game analysis using Anthropic's Claude Code augmented by the Stockfish engine. The system transforms a player's chess game and their "think-aloud" audio notes into a detailed, human-readable post-mortem, including a narrated video. Its core innovation lies in Claude's ability to interpret Stockfish's output in plain language and specifically address the player's in-game thoughts and decisions.
- Personalized Analysis: The system integrates player audio notes, transcribing them and aligning them with game moves, allowing Claude to comment on the player's actual thought process.
- Enhanced Explanations: Unlike raw Stockfish output which provides numbers and complex variations, Claude interrogates the engine to explain mistakes and strategic choices in an accessible manner.
- Comprehensive Output: The result includes an annotated PGN, a standalone HTML analysis board, and a narrated video, offering multiple ways to review the game.
- Technical Requirements: Users need Claude Code installed, along with Python, Stockfish, ffmpeg, and optional whisper.cpp for audio transcription. The setup is designed to be largely automated by Claude itself.
- Considerations: A full analysis can take about an hour and cost around $15 in API tokens. The author emphasizes its suitability for slower, thoughtful games rather than quick blitz sessions, and encourages providing detailed notes for the most valuable analysis.
- AI's Role: The author notes that while Claude provides the commentary, Stockfish acts as the authoritative source for move evaluations, mitigating potential AI "hallucinations."
In essence, this tool bridges the gap between raw, highly technical engine analysis and a player's desire for understandable, personalized insights into their game, making the learning process more engaging and tailored.
The Gossip
LLM Limitations vs. Tool-Augmented Power
Commenters fiercely debated the true chess-playing and analytical abilities of LLMs. Some argued that LLMs inherently lack understanding, hallucinate illegal moves, and are no better than a 1300 Elo player without external tools. Others countered that when LLMs are effectively integrated with powerful engines like Stockfish, they become invaluable tools for translating complex analysis into understandable human language, bridging the gap between raw engine data and player learning.
Costly Computations and Competitive Comparisons
The discussion often circled back to the project's practicality and value proposition. Many questioned the reported $15 cost per analysis and the hour-long processing time, especially given the existence of free and instant analysis tools on platforms like Lichess and Chess.com. Several users also highlighted their own similar, or earlier, projects, suggesting the concept isn't entirely novel, while the author acknowledged the project's "hacky" nature and lack of optimization.
The Ethics and Efficacy of AI-Generated Code and Explanations
A significant point of contention was the author's admission that the project's code, README, and even the "SKILL.md" prompt files were largely generated by Claude itself. This led to a sharp critique about "laziness" and "delusion" from some, questioning the intellectual value of such projects. Conversely, others found it to be a fascinating demonstration of models generating instructions for themselves, highlighting the evolving nature of AI development and the trade-offs between "deep understanding" and "efficient vagueness" in AI explanations.