I come from software development. Spent twenty years in it. But somewhere along the way, the writing-the-code part started feeling like the least interesting thing I did. Architecture, strategy, operations, coaching... all more rewarding, all less tedious. I stopped shipping my own code and didn’t really miss it.
I shipped a pull request to dnscontrol in January. I’d been using it for years to manage all my Cloudflare zones, and I hit a feature they didn’t support. I needed per-record CNAME flattening, a paid zone feature with a small customer base and zero incentive for the maintainers to build themselves. So I built it, it got accepted, and that kind of ignited something in me that had been long gone. So I did another one, adding support for record comments and tags, another Cloudflare feature nobody had wired up.
Then I finally built the Cloudflare security group sync I’d wanted but never made time to build, a Lambda that pulls Cloudflare’s published IP ranges and keeps your AWS security groups current. A small but very real gap in every setup I’d ever run. Then cfnpeek, because I needed a quick way to inspect stack outputs. Then cfntop, because I was tired of refreshing the AWS console during deploys. And I’m working on some iOS stuff that still needs to marinate a bit before I reveal it. 🤫
I didn’t plan any of this. But building rmBug meant dealing with infrastructure again, which meant hitting friction I’d been ignoring for years. This time I had Claude Code open in the next terminal, and the distance between “I wish this existed” and “okay it exists now” collapsed.
I didn’t expect the part where activation energy drops low enough that you stop triaging your ideas and just build them. The maintenance burden you couldn’t justify becomes manageable because you’re not writing every line from scratch.
Turns out I just needed the overhead to come down.

Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.