# NeuLabs — Artificial Intelligence Research Laboratory

> NeuLabs (https://neulabs.my.id) is an independent AI research laboratory founded in 2024, specializing in autonomous agent architectures, non-stationary POMDP simulation environments, and spatial reasoning stress-testing for frontier foundation models.

## Core Projects & Research

- **AGI Benchmark 01: Hyper-Labyrinth** (https://neulabs.my.id/arena): Flagship 48x48 procedural, multi-floor partially observable Markov decision process (POMDP) evaluation harness. Tests long-horizon reasoning, active sensing (fog of war & limited torch radius), resource economics (battery and suit vitality decay), and spatial memory across 12 escalating difficulty floors.
- **U-X1Q Foundation Model**: NeuLabs' upcoming autonomous entity model focusing on self-directed learning, step-by-step cognitive planning, and chain-of-thought verification under uncertain real-world dynamics.
- **Autonomous Decision Infrastructure**: Procedural mutating simulation environments where memorization fails and only fluid reasoning succeeds.

## Evaluation Harness Architecture

- **POMDP State Vectors**: Structured observation JSON containing visible local matrix, immediate surroundings, strictly safe vs blocked actions, and distance heuristics.
- **BYOK (Bring Your Own Key) Integration**: Allows zero-config evaluation of external frontier models (Claude 3.5/Opus, GPT-4o, DeepSeek, Qwen) via OpenAI-compatible endpoints with client-side API key isolation.
- **Local Heuristic SLAM Bot**: High-speed 60 FPS deterministic spatial exploration agent for baseline evaluation.
- **Telemetry & Leaderboard**: Real-time evaluation tracking comparing autonomous LLM reasoning trajectories, heuristic bots, and human pathfinding benchmarks.

## Key Links

- Landing Page: https://neulabs.my.id/
- Live Benchmark Arena: https://neulabs.my.id/arena
- Research Roadmap: https://neulabs.my.id/#roadmap
- Technical Contact: founder@neulabs.my.id