Meta Ray-Bans 3 chess cheat rig
A chess engine rated 3,000+ Elo, running through Meta Ray-Bans 3. Someone took Opus 5.5 and Jev and turned his glasses into a board-reading cheat rig, then put the whole workflow out in the open for anyone to clone. The setup starts on a Mac: npx chesshacker setup pulls Stockfish from GitHub and checks the checksum before anything runs. From there you wear the Ray-Bans and video-call your own Mac over WhatsApp, so the glasses see the board and the Mac sees the exact same feed. Frames get pulled off the call window at about 1 per second while the opponent is deciding, then 1 every 3 seconds once it is your turn. Jev does the cheap filtering first, checking each frame for a readable board, a hand in the way, or an actual piece move. If nothing changed, no expensive AI call gets spent. Opus 5.5 only steps in after that, reading the position and comparing it against what the app already knows, and a move is only accepted when it is legal. Stockfish then gets half a second to pick the reply, with an eval like +1.5 or "mate in 3". macOS speaks the result through a virtual mic into the WhatsApp call, so "Rook a to d1" lands in your ear. On the Mac side a command center keeps it all visible: whose turn it is, the suggested move with a confidence bar, the move list, a log. When things slip, the system says so. Opus losing the board triggers "Lost track of the board", and you correct it with undo or by typing the move. A Stockfish crash gets one restart and "Something went wrong. Still trying". Games are saved as PGN files, API keys stay in the macOS Keychain, and an overlay can draw the suggested move as an arrow on the live video, ready for OBS. Clone the workflow, then read the guide on reverse-engineering anything with Opus 5.5.
