Watch each software release for problems with agents
An engineer built a prototype where agents plan monitoring for each release and investigate alerts; he says it caught regressions in systems he didn't understand.
Done with Other
What they did
Ben Congdon prototyped agent-driven monitoring around software releases and settings changes. An agent reads the service code and writes a plan for what to watch. A regular program runs that plan. A second agent digs into anything that trips, and hands off to a person when needed.
How it went
He says even the early version caught problems in systems he knew nothing about, faster and more accurately than he could before.
Worth knowing
He found a pure agent loop worked worse. Having the agent write a monitoring plan that a plain program runs gave cleaner signals and used far fewer tokens.
Try it yourself
Read the code for [service name] and the change in [release or config change]. Write a monitoring plan for the next [hours] after it goes live: which metrics and log lines to watch, normal ranges, and what counts as a problem. Save it as a file I can run on a schedule. When a check trips, investigate using [your logs and metrics tool], summarize the likely cause and evidence, and message me before taking any action like rolling back.
Swap the highlighted [parts] for your own details.
Source: Hacker News · bcongdon · Oct 19, 2026
Written with AI assistance from the original post and checked against our editorial guidelines. Results are as reported by the source, not independently verified. Spotted a mistake? Tell us.
