AI chess arena
AI Showdown
A live chess arena for comparing model strategy, tool use, latency, recovery behavior, and tournament performance.
Outcome
Runs model-versus-model duels, knockout and round-robin tournaments, PGN export, telemetry inspection, and dynamic Elo ratings.
My role
Full-Stack Engineer
How I helped
- Built model-versus-model chess duels and tournament workflows.
- Added PGN export, telemetry inspection, and dynamic Elo ratings.
Product
The workflow in context
AI Showdown artwork from the open-source project.
The solution
AI Showdown compares model strategy and tool use through live chess matches. Model-versus-model duels, knockout tournaments, and round-robin tournaments expose latency and recovery behavior alongside match results.
What changed
PGN export, telemetry inspection, and dynamic Elo ratings make matches inspectable and tournament performance comparable.
Source
Explore the open-source project at github.com/azmi2409/ai-showdown.
AI system implementation
LLM tool calls participate in model-versus-model chess matches. The React interface and Express service expose live match activity through SSE; Chess.js supplies chess rules. Telemetry, exported PGN, tournament results, and dynamic Elo make strategy, latency, and recovery behavior inspectable rather than relying on model self-assessment.
Could a similar approach help your business?
Tell me where work gets stuck. We can identify a practical first improvement and what success should look like.
Discuss your challenge