---
format: "aidr-story-markdown/v1"
id: "adc3f266543937ea948141f5c250e3688faaf64947245197045522e90d83a176"
canonical_url: "https://aidr.today/adc3f266?lang=en"
title: "Wrong, not broken"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-09-26T17:08:06.000Z"
category: "Agents"
topics: ["amazon","agent"]
source_urls: ["https://aws.amazon.com/blogs/aws-insights/wrong-not-broken/","https://lobste.rs/s/aokbai/wrong_not_broken"]
summary: "Software operations has always asked one question: Is it broken? AI agents change that. Here’s the shift toward measuring correctness at the level of the run, and the thinking behind Amazon CloudWatch Omni."
---

# Wrong, not broken

> [Open the canonical story](<https://aidr.today/adc3f266?lang=en>)

**Published:** 2026-09-26T17:08:06.000Z
**Category:** Agents
**Topics:** amazon, agent

## Summary

Software operations has always asked one question: Is it broken? AI agents change that\. Here’s the shift toward measuring correctness at the level of the run, and the thinking behind Amazon CloudWatch Omni\.

## Sources

- [Story source](<https://aws.amazon.com/blogs/aws-insights/wrong-not-broken/>)
- [Discussion](<https://lobste.rs/s/aokbai/wrong_not_broken>)