AntiChess
An early platform for playing chess against different AI models and comparing their behavior.
Read the build from several angles.
Authentic screens, terminal output, and architecture plates are kept together so the page shows the shape of the work without inflating the claim.
This record uses a procedural architecture view because no authentic public screenshot has been approved.
AI
What exists.
AntiChess lets you play chess against multiple AI models and compare how they reason about the same positions differently.
Where it resisted.
Different AI models have different latency, reasoning depth, and move selection patterns. The interface needs to accommodate all of them without bias.
What I changed.
Built a monorepo with a shared chess engine package and a Next.js frontend using shadcn/ui with a custom chess theme.
The obvious implementation was not the product.
Separate the game logic from the AI provider layer so models can be swapped without touching the board.
Current proof
03- Monorepo architecture with shared engine package.
- Next.js 14+ with App Router and TypeScript.
- Custom chess theme with Tailwind design tokens.
What this build changed.
Game AI comparison is more about exposing reasoning patterns than about raw strength.