Lab

Accessible design is agent-ready design. I've spent years describing the problem in meetings and essays. Description only goes so far. This is where I show it.

Each experiment builds two versions of the same interface, gives an AI agent the same task on both, and logs what happened. Each write-up includes the markup, every step the agent took, and the limits of the test.

• Experiment 001: Labels vs. Div Soup

  Two forms that look identical. An AI agent finished one in four actions. On the other, it never found the submit button.

  Read →