Evidence item

ARC Prize observed Astra turning unfamiliar environments into compact symbolic world models and representing mechanics as rules for planning. DIRECT_TEXT

Authenticity UNVERIFIED · extraction UNVERIFIED · independence group none (unreviewed) · created seq 10185

Source: OpenAI's GPT-6 Astra on ARC-AGI-3 · locator {} · excerpt hash sha256:738eb3b9e0298ef880233def85efc2d4d85746e74bf3798a8e776e0fa6c455d3

A key behavior observed in GPT-6 Astra was its ability to turn unfamiliar environments into compact symbolic world models. It represented game mechanics as logical rules and developed its own domain-specific language shorthand to track state and plan actions.