We Built an AI That Explains Chess Positions

Today, we're pulling back the curtain on how Depthly works.

Most people think we just feed the FEN string to an LLM. While that's the first step, LLMs are notoriously bad at playing chess. They hallucinate illegal moves and misunderstand complex tactical motifs.

Our approach is different. We use Stockfish to find the absolute truth (the evaluation and best lines), and then we pass that structured analysis to our fine-tuned AI model, whose sole job is to translate engine logic into human concepts.

The result? An AI that never hallucinates an illegal move, but explains the position like a Grandmaster.