About

The thing that makes AI expertise meaningful right now isn't knowing how to use AI — it's knowing what it can actually do. Not what the benchmarks say or what the pitch deck claims. What it actually delivers in production, under real constraints, with real accountability for the output. That calibration is rare, and it only comes from stress-testing these systems honestly over time.

I've spent years doing that — building systems that had to work, for organizations where the stakes were real. What I learned is that the gap between AI as a concept and AI as a production capability is almost always organizational, not technical. The model isn't the problem. The surrounding system is: the software logic, the data infrastructure, the evaluation mechanisms, the human judgment applied at the right points. That's the framework I wrote about in Collaborative AI, and it's what I bring to every engagement.

Compound Agent OS is the running proof of concept. It's the AI operating system I built and run for my own serious knowledge work — research, writing, analysis, and strategy. Building and running it every day is how I stay calibrated on what's actually working at the frontier versus what's still more promise than production.

I work with a small number of companies and leaders each year on the decisions that determine whether AI projects actually deliver. If that's relevant to you, the Work With Me page has the details.