This is the most important thing you do before the course starts, and it takes less time than people fear. You are going to pick the one product you will build across all four weeks. Everything you make each week is a piece of this single thing, so choosing it now is the difference between finishing with a real portfolio piece and finishing with four disconnected demos.

A good capstone has three properties. It is small enough to finish in four weeks, which means one clear job done well, not five jobs half-built. It is real to you, drawn from your work, a side project, or a problem you personally have, because you will push harder on something you actually care about. And it has all four layers: a user signs in, does something, and their work is still there when they return. If it has those, it is a product, not a page.

Some things to avoid, because they quietly sink capstones. Anything that needs a huge dataset or a custom-trained model is out of scope. Anything so sprawling it cannot be shown in a three-minute demo is too big. And a vague theme with no clear user is not a capstone at all: "something with AI for marketing" gives you nothing to build, whereas "a tool that turns a sales call transcript into a follow-up email draft" tells you exactly what to make on day one.

To make this concrete, here are five that fit the course perfectly: a client intake tool that takes a form and drafts a proposal; an internal dashboard that pulls in data and answers questions about it; a booking flow that sends reminders automatically; a tool that turns long documents into short briefs for your team; a support assistant that answers from your own help docs. Notice what they share: one clear user, one core action, and an obvious first thing to build.

Your task is a one-page seed. Answer four questions. What does it do? One sentence; if it takes more than one, narrow it. Who is it for? Name the actual person who would use it. What is the single most important action it needs to do? This is what you will build first, in Week 1. What would make you proud to demo it? That is your Demo Day target.

Post your one-pager in Slack before Week 1. You will get feedback, and reading everyone else's seeds is half the value, it is the fastest way to see what "well-scoped" looks like. Do not overthink it. You can adjust as you go. What matters is walking into Week 1 with a real thing to build, not a blank screen.