# The 97-to-Zero Problem: Why Frontier AI Fails When You Remove the Training Wheels - Date: 2026-04-07 - Category: Artificial Intelligence Opus 4.6 hit 97% on an ARC-AGI-3 environment with a hand-crafted harness. Drop that same model into an unfamiliar task and it scores zero. That is not an intelligence gap. That is memorization dressed as reasoning. ---