---
title: "Comment by Yunus Emre Vurgun on \"Before AI Moves Money: The Eval Gate I Built for AutoKon\" | Hashnode"
description: "Putting the judge in the payment path and failing closed is the right call, and narrowing the prompt to per-claim grounding rather than overall quality is the k"
url: "https://hashnode.com/posts/before-ai-moves-money-the-eval-gate-i-built-for-autokon/6ad4c7ca5d5bf6bbe9352fbf/comment/6ad4d6c8b32f28f6765938af"
---

[Discussion](https://hashnode.com/posts/before-ai-moves-money-the-eval-gate-i-built-for-autokon/6ad4c7ca5d5bf6bbe9352fbf)

Comment [Before AI Moves Money: The Eval Gate I Built for AutoKon](https://hashnode.com/posts/before-ai-moves-money-the-eval-gate-i-built-for-autokon/6ad4c7ca5d5bf6bbe9352fbf)

[Oct 18](https://hashnode.com/posts/before-ai-moves-money-the-eval-gate-i-built-for-autokon/6ad4c7ca5d5bf6bbe9352fbf/comment/6ad4d6c8b32f28f6765938af)

Putting the judge in the payment path and failing closed is the right call, and narrowing the prompt to per-claim grounding rather than overall quality is the key detail. One thing I'd be curious about: do you ever run the judge on a deliberately mismatched photo/description pair (e.g. a valid description attached to a photo of a different unit) as a canary? Since both models can share the same blind spots, a small rotating set of known-bad pairs is a cheap way to catch the judge drifting toward agreeing with the generator.
