I use AI agents in my open source work, and I'd like to be open about how, so you can judge the result on its merits.

Where I come from. I'm a computer scientist by training, and I've spent almost 25 years building software. I'm not someone who jumps on every new model and sees what falls out. I know what good code looks like, and I put my name on what I release.

AI is my collaborator, not my autopilot. I design the architecture, interfaces and data models, and the agents help me implement, refactor, document and review them. We work through the code together, and I steer and correct when they drift. The agents are listed as contributors in each project, because that's what they are.

Everything gets tested. The code runs through automated tests, and I use the software myself, on my own real data. If it doesn't hold up in daily use, it doesn't ship.

The code that matters most, I review myself. I read all security-relevant and low-level code line by line. That means anything that touches kernel APIs, drivers, filesystem attributes or permissions. SEN, for example, was handcrafted until recently, and its foundations stay under my direct review. AI assistance is there to make it more stable and better tested, not to replace my judgment where it counts.

Licensing is checked proactively. Every build runs a license linter, so licensing issues are caught before they reach a release.

It is maintained, not thrown over the fence. These are not throwaway prototypes. I fix bugs, answer issues and refactor when the structure needs it. Together with the agents, I keep the code modular, commented and documented, so that other people can read it, review it and take it over.

What I don't claim. I don't proofread every line of generated code, and I won't pretend I do. My assurance rests on architecture, modularity, tests, real-world use and focused reviews. If you need more than that, please look at the code yourself. I welcome it, and I'm happy to fix what you find.

Why I do it this way. I can't build all of this by hand in my spare time. With AI, ideas that would have stayed on the drawing board become software that anyone can use, study and improve. I understand the skepticism, and I share the dislike of careless, unreviewed slop. I hope you'll judge my work by its care and quality, not by the tools behind it.