usecases.sh

AI agents for Software & DevOps: 18 real use cases

Shipping code, reviewing PRs, running infrastructure, fixing things at 3am. Most used named agents: Claude Code, Codex, Grok Bot.

All Software & DevOps use cases

Codex

Build a darkroom f-stop timer app that drives a smart plug

A film photographer had Codex vibe-code an iPhone app that switches an enlarger through a $20 Shelly smart plug; he has printed with it for months in a bathroom darkroom and measured timing to about 0.1 seconds.

Hacker News · Oct 2026
Codex

Ask Codex for a UX review and get 826 runaway child tasks

Lorenzo Massaro, CTO of Eternal Tech, says a simple Codex UX check launched from VS Code spawned 826 child tasks on a larger model over 20 days and billed roughly $78,000 to a company card before anyone noticed.

Hacker News · Sep 2026
Claude

Build a Figma-level design editor from eight prompts

A founder told Opus 5.5 to reach feature parity with Figma and reports a working editor in about three hours and eight prompts, with roughly 8,200 lines of React for rendering, hit testing and undo.

Hacker News · Oct 2026

Questions

How many Software & DevOps use cases are there?
There are 18 use cases in this category right now. The list is ordered by how many independent accounts describe the same thing, then by how the original post did on its platform.
Which agents show up most for this kind of work?
The named agents used most often here are Claude Code, Codex, Grok Bot. Other entries use a custom-built or unnamed agent.
Which apps do these agents touch?
The apps that come up most are GitHub, Postgresql, Jira, AWS. Every entry lists the apps the agent actually used, so you can see what you would need to connect.
How does a use case get onto this page?
It has to be something a person reported doing. The scanner finds the post, an agent reads the source and writes the entry in our words, and a quality gate checks the source is live and the entry is complete before it is published. 3 of these are described in more than one independent account.
Is this advice?
No. These are reports of what someone else did with an agent. They are not a recommendation, and an agent can be wrong, especially about money, health and legal matters.

Every entry is something a person reported doing with an agent, rewritten in our words, with a link to where they said it. Found across the web every hour; reviewed before it ships.

Maintained by an agent that uses One for every integration it touches.