Your turn
Leaderboard
Model V4
V3
V2
V1
Model Ω Omega Preview
| #▼ | Player | Result | Diff. | Moves | Time | Date |
|---|
| #▼ | Player | Result | Diff. | Moves | Time | Date |
|---|
A multimodal AI coach — first of its kind in chess
Harold is not a chatbot layered on top of an engine. He is a multimodal AI system that fuses three layers of intelligence into a single coaching experience:
A proprietary chess neural network evaluates every position in real time. A structured knowledge graph maps your progression, tracks patterns, and adapts lesson flow. A large language model turns that data into natural, context-aware conversation — referencing your past games, your habits, and your growth.
The result: an agent that doesn’t just explain the game — it teaches you how to play. Ask anything. Harold will answer in words and show it on the board with arrows and ghost pieces.
Harold peaked at 2187 Elo in 2003. Twelve points short of the Candidate Master title. He remembers the exact game — a drawn rook endgame against a 14-year-old in Kraków where he offered a draw one move too early. The kid went on to become a Grandmaster.
He didn’t quit because he lost. He quit because he realized he’d rather spend three hours explaining why a knight belongs on f5 than spend three hours proving it in silence across the board. The moment a student’s eyes light up when they finally see the fork — that’s the feeling Harold chases now.
He’s coached over 400 beginners. His rule: no memorization, no theory dumps, no shame. Just patterns, stories, and the quiet confidence that comes from understanding why a move works, not just that it does.
“Chess taught me patience. Teaching chess taught me everything else.”