arc-agi-3

ARC-AGI Prize Explainer (Includes ARC-AGI-3!)

Benchmark methodology, task sets, and evaluation results for RA²R

Minimalistic All-in-One Toolkit for ARC-AGI-3 AI reasoning benchmark

Prove you achieved AGI at home by testing your claude-code setup against arc-agi-3 benchmarks

Worm Arena

Don't step in lava! Navigate from the bottom-left to the top-right based on the heat of each tile. ARC-AGI-3 challenge game.

UB: an object-centric ARC-AGI-3 game reasoning agent with visual memory, adaptive planning, BFS recovery, and a live viewer

TWIN: Test Time World Model Inference

Fully-offline, deterministic symbolic agent for the ARC Prize 2026 ARC-AGI-3 track. Hybrid occam (MIT) + step-wise solver, ~9/25 public games, no LLM/GPU in the loop.

My lab for learning agent tooling by building ARC-AGI-3 game-playing agents

The ARC-AGI-3 track of ARC Prize 2026 — interactive, agentic reasoning.

Deterministic world-model engine, Grid-SLAM workbench, and visual discovery harness for ARC-AGI-3

🤖 ARC-AGI-3 Agent with 100% success rate. EfficientAgentV4 uses BFS + numpy + deque.

Adaptive AI Concept | Testing with ARC-AGI-3

A glass-box, safety-shielded agent for the ARC-AGI-3 interactive reasoning benchmark: object-centric perception, a human-readable learned world model, BFS planning, and an online-learned safety shield. CPU-only, no hosted LLM, honest live-API evaluation.