High-Performance Chess Engine Built on MIND

Copyright (c) 2025 STARGA, Inc. All rights reserved.

NikolaChess is a state-of-the-art chess engine built on the MIND programming language - a next-generation systems language designed for AI/ML workloads with native tensor operations and CUDA support.

MIND is a modern systems programming language developed by STARGA, Inc. that enables:

• Native Tensor Operations: First-class tensor types (tensor<f16, (16, 8, 8)>)
• GPU Acceleration: Seamless CUDA integration (on(gpu0) { })
• SIMD Intrinsics: Hardware-optimized vector operations
• Rust-Level Safety: Memory safety without garbage collection
• Python-Level Ergonomics: Clean, readable syntax

// Example: GPU-accelerated batch evaluation
on(gpu0) {
    parallel for i in 0..65536 {
        let board = boards[i];
        let eval = network.forward(board.to_tensor());
        results[i] = eval;
    }
}

The MIND compiler and runtime are proprietary technologies of STARGA, Inc.

• NNUE Evaluation: Stockfish-compatible neural network evaluation
• CUDA Backend: GPU-accelerated search via MIND Runtime
• Syzygy Tablebases: Perfect endgame play (7-piece support)
• Cloud Eval Integration: Lichess cloud evaluation for openings
• Predictive Analysis: Cloud eval exploitation for opponent prediction
• Pondering: Continuous analysis during opponent's time
• Advanced Time Management: Aggressive allocation strategies

• +600 Elo improvement with NNUE evaluation
• Optimized for competitive online play
• Sub-millisecond move generation

Download the latest release from the Releases section.

Available builds:

• nikola-cuda.exe - Windows CUDA build (recommended)
• nikola-cpu.exe - Windows CPU-only build
• nikola-cuda - Linux CUDA build
• nikola-cpu - Linux CPU-only build

Source code access is restricted to authorized personnel.

NikolaChess is built on proprietary MIND language technology. For licensing and partnership inquiries: info@star.ga

NikolaChess implements the Universal Chess Interface (UCI) protocol.

> nikola-cuda.exe
NikolaChess v3.28.0-nnue by STARGA, Inc.

> uci
id name NikolaChess
id author STARGA, Inc.
uciok

> position startpos
> go depth 20
info depth 1 score cp 20 nodes 21 pv e2e4
info depth 2 score cp 15 nodes 84 pv e2e4 e7e5
...
bestmove e2e4

Run the Lichess bot wrapper:

python lichess_bot.py [depth]

Requires LICHESS_API_TOKEN environment variable.

NikolaChess
├── MIND Core (Proprietary)
│   ├── Board representation
│   ├── Move generation
│   ├── Search algorithms
│   └── Evaluation functions
│
├── MIND Runtime
│   ├── CPU Backend
│   └── CUDA Backend (GPU)
│
├── Mind NNUE Module (native/mind_nnue/)
│   ├── Teacher model (full precision training)
│   ├── Student model (distillation target)
│   ├── SCReLU activation + WDL head
│   └── NKNN v2 quantized export
│
├── Training Pipeline
│   ├── generate-lichess-shards (PGN → labeled shards)
│   └── train-mlp (GPU training via Mind Runtime)
│
└── Integration Layer
    ├── UCI Protocol
    ├── Syzygy Tablebases (6-man, 149GB)
    └── Lichess API

PROPRIETARY AND CONFIDENTIAL

This software is the exclusive property of STARGA, Inc. All rights reserved.

• Unauthorized copying, modification, or distribution is prohibited
• Reverse engineering is prohibited
• Published binaries are for evaluation purposes only
• Commercial use requires explicit licensing agreement

For licensing inquiries: info@star.ga

STARGA, Inc. develops next-generation programming languages and AI systems. The MIND programming language represents our vision for the future of systems programming.

NikolaChess v3.18 - Built with MIND. Engineered to win.