Prototyping
Prototyping is building a quick, rough version of something to learn about it before committing to a production implementation. A prototype is a disposable artifact for exploring design questions that conversation and specification cannot resolve – how an interaction feels, whether an API shape is ergonomic in real calling code, whether a layout works at real data sizes.
Some questions cannot be answered in words. Past that point, discussion circles, so you build the thing, look at it, and come back with an answer. The workflow is iterative: react, ask for a change, react again. Each round resolves another decision against the real artifact at higher fidelity than conversation allows.
Prototypes trade polish for speed and learning. The value is in what they teach, not in the code itself – though you can build the pieces you are actually evaluating to production quality, so the reacted-to component transfers into the real codebase rather than being thrown away.
AI coding agents lower the cost of building, which makes prototyping practical to do routinely. A rough version that used to take a day to mock up now takes minutes. A prototype does not have to be all-scrappy: the cheap build cost means you can afford to make the parts that matter solid, and vibe coding a first cut in an afternoon. An agent handoff spec often references prototypes as the concrete artifact a future session should react to or replace.