---
title: "Comment by Macro8n on \"Stateful AI Workflows: Routing, Checkpoints, and Human Review\" | Hashnode"
description: "Revision-bound approval fits anything with side effects. For desktop flows, I use a similar rule: verify the expected window/state before acting, then stop on d"
url: "https://hashnode.com/posts/stateful-workflows-and-review/6afeffa9c927dfeaaf726e95/comment/6b007ac7804494f976583ac4"
---

[Discussion](https://hashnode.com/posts/stateful-workflows-and-review/6afeffa9c927dfeaaf726e95)

Comment [Stateful AI Workflows: Routing, Checkpoints, and Human Review](https://hashnode.com/posts/stateful-workflows-and-review/6afeffa9c927dfeaaf726e95)

[Macro8n](https://hashnode.com/@macro8n)

[9h ago](https://hashnode.com/posts/stateful-workflows-and-review/6afeffa9c927dfeaaf726e95/comment/6b007ac7804494f976583ac4)

Revision-bound approval fits anything with side effects. For desktop flows, I use a similar rule: verify the expected window/state before acting, then stop on drift instead of replaying blindly. I'm the developer of Macro8n, a local-first visual automation app for Windows; we just released it at macro8n.com. Curious: do you treat a retry as a new revision or a continuation of the same approval?
