On Nov. 3, 2026, the Google Developers Blog said per-token API prices keep falling but engineering bills for autonomous coding agents keep climbing, and that cost per successful task—total spend across every loop iteration, tool call, retry and internal thinking pass divided by the number of tasks that pass verification—is now the dominating metric. It suggested adding a lightweight routing layer to intercept routine decisions before they reach an expensive reasoning model, citing TypeSafe's Jev classifier in front of Google Antigravity's model calls; Jev evaluates structured classification primitives in a single forward pass: a Choice from a list, a numeric Score or a Noul, which returns a 0 to 1 probability for a binary statement. Google disclosed it has no partnership with TypeSafe and the article is not an endorsement. The blog said a 50-question, auto-graded, coding-agent benchmark against live Gemini and Jev APIs tested whether a near-free classifier can route each task to the cheapest model that still gets it right. It reported that regex and token-length harness routers collapse on unseen agent tasks: on 16 coding-agent tasks, a single broad Choice question routed nearly everything to simple, even tasks requiring an LRU cache trace or modular exponentiation, and replacing it with one atomic Noul question fixed the separation immediately.