Building AI software,
documenting the work
DogXAI is an independent lab for building and documenting AI products, agents, engineering systems, and open technology.
I write about technology, product judgment, and the trade-offs that appear in real projects. The goal is not to chase every new term, but to understand which capabilities actually change software and how to turn them into products that are reliable, maintainable, and useful to people.
Content should be verifiable whenever possible: code should run, architecture should state its boundaries, and opinions should expose their assumptions. When something cannot be confirmed, the uncertainty should remain visible.
The projects page presents working results. Articles document their implementation, design choices, and lessons from building.
Editorial identity: DogXAI is both the author and publisher of this site. Important technical claims prioritize primary sources; DogXAI measurements, interpretations, and recommendations are separated where practical. Explore the results and build notes on the projects page.
Build something that runs before explaining what actually happened.
Prefer tests, data, and reproducible processes over vague AI narratives.
Share code, process, and failures when possible so other people can keep building.