Shipping a real product in 48 hours, with AI in the loopShipping a real product in 48 hours, with AI in the loop

WavyClub is №02 in my series of experiments building real products with AI in the loop. The point was never to ship a side project — it was to develop a build process I trust as design, specification, and engineering collapse into one loop. I gave myself two days and used an existing surf rental product, SurfsUp, as a fixed reference. A known target removes the blank-page problem and turns every decision into one about interpretation, structure, and taste — which is where the learning lives. I know SurfsUp from the inside as a partner through my FinFun channel, so I could see the small gaps that only reveal themselves once you've lived with a product long enough to stop noticing them.

I ran the whole thing as a five-stage loop — Frame, Shape, Breadboard, Build, Ship — each stage built around the question it exists to ask. I framed two specific scenarios (the traveling surfer, the wrong-board-for-today surfer) before any design. I set the appetite — one day to build, one buffer — before scoping a single feature, because adding is cheap with AI and time-before-features is the constraint that keeps the build honest. I breadboarded eight places, their affordances, and data stores into a nine-page shaping doc before opening Figma. Building was the part the practice is really about: specification became the craft, taste stayed editorial — knowing what to keep, reject, and when to stop — and writing the right question mattered more than asking for “better.”