Skip to content

Note

November 17, 2026

2 min read

Build it for now

By Cristiano Pierry

How this site is made

Software can be worth building for one investigation, then retired once it has served its purpose.

ai-codingdeveloper-toolscodex
A cream-colored replicator holds a translucent blue app tile with one control and a pixelated edge. A paper request slip in front shows the same outline, and a tiny toucan perches on the machine.

A piece of software can be worth building even if I only use it for one investigation. With Codex, I can make a tool for the problem in front of me, use it while I need it, and retire it once it has served its purpose.

Over the past 72 hours, both Mac Minis and several smart devices in our home have been losing connectivity while other devices in the house keep working. The affected devices stay disconnected until we reboot the main Orbi router.

I used ChatGPT Codex to check the AT&T fiber gateway and our Orbi configuration. The results looked normal, so we built a network monitor to record what happens during an outage.

The tool is now running on both Minis as we wait for the next outage to find out whether it collects the evidence we need to diagnose the failure.

My usual approach would have been to research monitoring software, download a few candidates, and work out which one fit. I might have bought a license or, more likely, added another subscription. This time, I used Codex to build software for the exact network configuration of my house.

Frontier coding tools make it practical for more people to describe a problem to an AI assistant and ask it to write a tool for that situation. We may only need our monitor until we understand the outages.

Once the investigation is over, we can shut the monitor down and remove it. It doesn’t have to survive the next OS upgrade or keep receiving fixes for newly discovered security vulnerabilities after we’ve finished using it.

For users, the benefit is being able to make progress without first finding an existing product that fits their particular setup. An issue that might otherwise frustrate them for months can become something they can investigate now.


This writing reflects my personal perspectives on product management, AI, and content discovery. It does not represent the official position of my employer or any affiliated organization.