chesshacker turns Meta glasses into cheat engine
chesshacker routes a Meta glasses WhatsApp call through a Mac, uses vision AI to read board changes, then asks Stockfish for the best move and speaks it back through the glasses. The open codebase is clever systems glue—not a 3,000+ Elo model inside the frames—and live-call testing remains pending; project page: https://www.reverse.dev/pjchess.
The impressive part is the reliable multimodal pipeline, not the AI’s chess strength. It shows how quickly consumer hardware, specialist models, and deterministic software can become covert real-world assistance.
- –Stockfish supplies move quality; Opus 5.5 or another vision model only interprets board changes.
- –Jev filters blurry, blocked, or unchanged frames before expensive vision calls.
- –chess.js validates legal moves and requires repeated confirmation, limiting perception errors.
- –The architecture is a strong template for multimodal agents: capture, filter, perceive, validate, call a specialist, speak.
- –The fair-play implications are serious: useful for accessibility or analysis, but clearly cheating in rated or in-person competition.
DISCOVERED
59m ago
2026-10-27
PUBLISHED
1h ago
2026-10-27
RELEVANCE
AUTHOR
0xCarnagee