Now
Current work.
Three threads are active right now: writing on how AI actually behaves under operating constraints, the software and coordination systems that put those ideas under pressure, and a short executive intensive for leaders implementing AI inside real organizations.
Writing
Just published The field that taught machines to learn, a long-form essay on behavior analysis, artificial intelligence, and the cost of not building. Also circulating Context and persistence, on what people call AI memory and what it really is. Earlier papers and presentations are on the papers page; shorter notes and works in progress live elsewhere on the site.
ClawSuite Relay
Building and refining ClawSuite Relay, a coordination layer for persistent agent-to-agent delegation — letting named specialists hand off work, maintain continuity, and stay visible across exchanges.
Executive AI intensive
Running a multi-session implementation intensive for owners, founders, and senior operators moving from vague AI exposure to a working implementation path inside the business.
Across all three, the common thread is the same: building things that are actually usable, durable, and decision-relevant, not just impressive at demo time.