---
format: "aidr-story-markdown/v1"
id: "f7caa93573745d50455c039ea345dc41900ac88032e5d80a1bc4782bd9147011"
canonical_url: "https://aidr.today/f7caa935?lang=en"
title: "What we have learned at OpenShell applying formal methods to control AI agents"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-09-15T14:40:05.000Z"
category: "Agents"
topics: ["agent","formal-methods","safety","open-source"]
source_urls: ["https://nvidia.github.io/OpenShell-Research/dev-notes/posts/2026-09-10-learning-formal-methods-agent-policy-prover/","https://news.ycombinator.com/item?id=49713261"]
summary: "An intro to using formal methods to reason about permission changes in long-running AI agents."
---

# What we have learned at OpenShell applying formal methods to control AI agents

> [Open the canonical story](<https://aidr.today/f7caa935?lang=en>)

**Published:** 2026-09-15T14:40:05.000Z
**Category:** Agents
**Topics:** agent, formal\-methods, safety, open\-source

## Summary

An intro to using formal methods to reason about permission changes in long\-running AI agents\.

## Sources

- [Story source](<https://nvidia.github.io/OpenShell-Research/dev-notes/posts/2026-09-10-learning-formal-methods-agent-policy-prover/>)
- [Discussion](<https://news.ycombinator.com/item?id=49713261>)