I've spent the better part of a year building demo environments for a living. Not prototypes, not MVPs, not production apps. Demos: software whose entire purpose is to be watched for eleven minutes by someone deciding whether to spend money.
It is a strange discipline. Most of what I learned as a frontend engineer over the previous decade applies, but the priorities invert in ways that took me months to internalize. Here is what actually transferred, what didn't, and what I got wrong.
The demo is a narrative, and the code serves it
The first demo I built was technically excellent and completely useless.
I had built a content modeling walkthrough with proper error boundaries, optimistic updates, a loading skeleton for every async boundary, and a genuinely nice type-safe API layer. During the run-through, the account executive stopped me forty seconds in and asked why we were looking at a spinner.
