Small, self-initiated projects where I work with AI agents as design partners and pay attention to what changes when I do. The work feeds directly back into how I design: better process, new skills, and working tools that make an AI-augmented practice more effective.
Each one ends with something you can try: a working prototype of a fictional product, built in the open. The agents do the building and the critiquing. They take turns leading and reviewing, disagree on the record, and wait for my ruling. I set the process, make the calls, and keep the account honest, including the parts that went wrong.
The prototypes are scripted demonstrations, not live AI. The products and companies inside them are invented. What is real is the process: the trial records, the reviews, the decisions, and the effort logs, all linked from each case study.
Start with any of them. Each ends with a prototype and a plain account of how it was made.
Halfmoon Tea
The Authority Line
How far may an AI marketing agent go on its own? A working prototype that shows the morning each setting produces, and an honest account of building it with two AI agents as partners.
Read the case study →
Meeting workspace
Pre Share
A feature idea for AI meeting notes: fix one wrong decision and see exactly what else that fix will change before anything moves. Built with two AI agents, mistakes included.
Read the case study →
Atelier
Agent Spaces
A feature concept for a design tool: a project page where a designer sees what two AI agents need decided, inspects the work, and rules beside it. Built with two AI agents, with a real working trial behind it.
Read the case study →
New experiments are added here as they ship. This series is part of Design and Systems.
↩ Back to Design and Systems