Outside of my day job, I've been building a few small SaaS tools. The biggest change AI has brought for me isn't that it does the work for me—it has dramatically reduced the implementation cost. The hard part is still figuring out what to build.

Most of these projects come directly from problems I run into.

One example is cloud cost management. There are plenty of dashboards that show where your cloud bill goes, but I haven't found one that actively points out the small, easy-to-miss services that quietly charge you month after month. Each charge is tiny, so nobody notices, but over a long period they add up.

Since I work in the data infrastructure space, I also built a tool that helps people install and configure databases. Setting up a database still involves a surprising amount of manual work, and I wanted something that made it much more straightforward.

Another project comes from my appreciation for GitHub Actions. I like the workflow, but much of our code isn't hosted on GitHub, and our deployment environments are all different. So I started building an offline CI/CD system. It can package code, copy artifacts, deploy them, run validation, and execute AI-generated deployment scripts. As long as your code lives in a Git repository and the target machine is reachable over the network, it works regardless of where the Git server is hosted.

All of these tools run on top of a backend platform I call *MotherBoard*, which provides the common pieces: authentication and SSO, metering and billing, products, subscription plans, and other shared services.

I've noticed that almost everything I build starts with a problem I personally encounter. In the past, I'd think, "That would be nice to have," and stop there because building it would take too much time. Now the implementation cost is low enough that I can actually build it. And if it turns out other people have the same problem, maybe it'll make a little money too.