I build practical AI and developer tools, experiment with coding agents and model infrastructure, and then document what happens when those ideas meet a real project.
The recurring theme is control: local-first software when it makes sense, transparent routing instead of magic, and tools that explain enough of themselves to be trusted.
The projects are the résumé.
I would rather show a working repository, a benchmark, a failed experiment and a build log than a long list of percentage bars claiming how good I am at a technology.
TunnelLab is part of the same loop.
The channel is where I turn tests and builds into useful explanations. That means the code influences the videos, and the questions raised by viewers often feed the next experiment.
Build → test → break → fix → document.
That is the short version.