Make it real on paper before it is real in code: the solution, the screens, the data, the plan.
a sign you are ready: demoable, scope crisp, technically sound
what happens inside, and how to do it
There is always more than one way to solve it. Spread out, then narrow down.
howSketch three to five rough approaches. Weigh each on how well it solves the problem against how complex it is, then pick one and write down why. Build a throwaway prototype only when words are not enough.
helps: generates options and rough prototypes in minutes.
watch: a quick prototype tempts you to ship it, and AI defaults make everything look the same.
common trap: grabbing the first obvious idea.
How it works, before how it looks. Most quality failures are experience failures, not visual ones.
howMap every screen and every state, not just the happy path: empty, loading, error, success, and the edges (huge lists, slow connection, mobile, the back button). The empty state is a teaching moment, not an afterthought.
helps: lists the states and edge cases you would forget.
watch: its list is generic. Your users have specific cases AI does not know.
common trap: designing only the happy path.
Data structures outlive code. Changing them after you have users is the most expensive kind of debt.
howList your entities, justify every field, map how they relate. Keep it one simple app on a proven stack until there is a real reason not to. Secrets never live in code.
helps: drafts a schema and an architecture from a plain description.
watch: do not accept a schema you cannot explain. Ask it 'why this choice' until you can.
common trap: over-engineering for scale you do not have yet.
Turn the blueprint into small, shippable chunks you can actually finish.
howSlice vertically: one whole thin feature, end to end, at a time. Order the slices by what depends on what, and do the risky, uncertain ones first.
helps: drafts the task list and a sensible order.
watch: plans look better than they are; they miss the real frictions in your code.
common trap: 'build the feature' as one giant ticket nobody can finish.
This is one way to know Shape is done: demoable, scope crisp, technically sound. But it is not a locked gate. You decide what this phase means for you and when you are ready to move. Read around it, think through the what and the why, then write your own definition. the cheapest time to change your mind. A sketch costs minutes, a built feature costs weeks.