Experiments

First-hand tests run against AI systems — with full conversation replays, methodology notes, and honest assessments of what worked and what didn't.