Notes on AI adoption, process design,
and the gap between the two.

I ran local models so you don't have to

Local LLMs promised privacy, speed, and zero API costs. I tested them seriously — including building tools around them. Here's what actually works, what doesn't, and why Ollama's move to cloud/subscription validated the constraints I hit.

The question every AI conversation skips

Everyone wants to know how to implement AI. Almost nobody asks whether they should — not yet, not for this problem, not without solving the upstream process issue first. That question is what Wolflow is built around.

Automate the process, not the workaround

The most common automation failure: digitizing a broken workflow instead of fixing it. How to tell the difference, and why the redesign conversation is always worth having before the build conversation.