A prototype ends the meeting
Two people can argue about a flow for an hour. A clickable prototype settles it in five minutes. I prototype early to turn abstract debate into a concrete thing people can react to. Showing beats telling, every time.
Fidelity matches the question
I match the prototype to what I am testing: a paper sketch for structure, a Figma click through for flow, a coded prototype for feel and technical feasibility. Over polishing a throwaway is as wasteful as under building a real test.
Prototype to learn, not to impress
The goal is to be wrong cheaply. I build the smallest thing that answers the riskiest open question, put it in front of a real user, and expect to throw it away. A prototype that only confirms what I hoped taught me nothing.
Code prototypes with AI change the economics
With modern AI tooling I can stand up a working prototype myself, a real flow against a real API, without booking an engineering sprint. That lets me validate feasibility and feel before committing the team, which changes what is even worth prototyping.
Real example
I regularly prototype to align stakeholders, turning a contested spec into something clickable so the debate is about the real experience, not everyone's mental image of it. Decisions get made faster because people are reacting to the same concrete thing.