I've spent the last month at the front edge of a genuine technological shift. I want to share what I've learned.

I've read both camps. Ed Zitron's financial doomsday scenarios to the singularity absolutists. Neither extreme quite captures it. What I can tell you from the inside: the current state of AI is nearly impossible to overstate. And this was not a surprise. Two years ago it became clear that agentic AI would matter more than the underlying language models. That bet is paying off.

Yet the toolkit most of us in leadership are working with today does not scale to this moment. The shift is large enough that we probably all need to go back to school, with the advantage of business insight but the humility of a student. Consulting firms are learning alongside us. Nobody has the playbook yet.

The Work

I built a rigorous, high-throughput research program organized around four areas: representational learning to understand and calibrate what AI actually learns; edge-friendly inference frameworks that are low-cost and privacy-preserving; secure software development automation; and novel architectures I hope will open a third axis of AI scaling, beyond models and agents. Papers and details on GitHub.

I also started writing for people newer to AI. The Incomplete Guide to AI is my own publication. I contributed a guest piece to Zane Hall's newsletter. I continued mentoring across industries and advising companies beginning to adopt modern AI.

IRL: Month One Checkpoint

P.S. If you're heading to NVIDIA GTC, message me. I'd like to connect.