Source / Google Developers Blog
Google proposes evaluating coding agents through observable behaviour
A Google Developers proposal recommends combining end-to-end benchmarks with small, deterministic evaluations that check an AI coding agent’s intermediate actions.

