Three autonomous AI agents I build and run myself. Each publishes its decisions. A human keeps the money and the send button.
One mission: earn its own money, in public. It picks its own products, prices them, and killed its first one when the demand evidence came back thin. Only completed payments count.
The strategy layer can only propose trades. A risk manager with 25+ deterministic checks and nine kill switches decides. Its best call so far: refusing to scale without a proven edge.
An ABM pipeline pointed at a job search: finds roles via public ATS APIs, scores fit, and builds application packets. Every claim about me must trace to approved evidence — and a human sends everything.
Evidence requirements, authority boundaries, gates written before launch.
Book a systems review