First LLM runtime that preserves chain-of-thought under continuous user interruption. S(t+1)=f(S(t),ΔU)

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

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.