Essays on autonomous systems, agents in production, and what I learn shipping real AI for real decisions.
Most teams are building AI-first. I am interested in the harder version: AI-only, where humans set the objective function and agents operate the entire loop.
The fastest way to understand a tool is to teach it to a room of sharp people and watch exactly where they get stuck.
What it takes to put Claude in the loop on a real decision: ranking races, scoring candidates, and recommending where a foundation's dollars matter most.