You have been thinking about the idea for weeks. Maybe months. You have described it to friends over dinner. You have opened a notes app more than once and closed it again. Then, on a quiet evening, you finally open Lovable, or Replit, or Bolt, or Cursor, and the prompt box asks the deceptively simple question: what would you like to build?
So you paste in everything. A platform where creators can create profiles, upload work, get discovered, receive enquiries, manage projects, accept payments and track analytics. You hit send. Fifteen minutes later, a dashboard renders. A sidebar. A settings page. A pricing page. A database schema. Twelve half-built workflows. It looks impressive. It looks like a product.
And you still have no idea whether the actual thing you were trying to build actually works.
The first thing to build is not the smallest feature. It is the smallest complete product behaviour that proves the idea makes sense.
Features, flows and core behaviours are not the same thing
Most idea-stage prompts collapse three different things into one. It helps a lot to separate them before you type anything.
Most first prompts describe features. A smaller number describe flows. Almost none describe the core behaviour. That is why the first build so often looks impressive and answers nothing.
One example we can hold on to
Say you want to build a product that lets founders collect feedback on early product ideas. The full imagined version has a lot in it: accounts, projects, public feedback pages, comments, voting, AI summaries, analytics, notifications, team members, exports.
Now compare a few candidate first tasks.
If the product cannot do one useful thing yet, adding more features will not fix it. It will only make the confusion bigger.
The first build test
Before you type the first prompt, answer these five questions honestly. If any answer is fuzzy, the first build will be fuzzy too.
- Who is the user? A specific person, not a category.
- What are they trying to do? In their language, not yours.
- What single action starts the value exchange? The moment the product begins doing its job.
- What must happen for the user to actually receive value? Not just a success toast. A real outcome.
- What is the smallest complete loop from action to outcome, with nothing removed and nothing added?
Once those are answered, the first build task usually writes itself. It also usually looks smaller than what you originally wanted to describe. That is the point.
A sequence that keeps you moving
There is a rhythm that keeps idea-stage builds honest. It is boring and slightly unglamorous. It also works.
- DECIDEName the core behaviour in one sentence. If you cannot, keep thinking. Do not build yet.
- MODELSketch only the data that behaviour needs. Not the whole product. Just this loop.
- BUILDGive the tool one clear instruction for the smallest complete flow that produces the behaviour.
- VERIFYUse it like the intended user would. Fresh account, real intent, no shortcuts.
- EXPANDOnly now decide what deserves attention next. The product has finally done something.
The tool is not slower here. You are slower, on purpose, at the layer where slowness saves you rebuilding the same decisions three times.
This is one of the reasons Zagvo exists
Zagvo is not another builder. It sits beside whichever builder you are already using. You tell it what you are trying to build. It helps you turn a broad, tangled product idea into phases, and each phase into one useful task at a time.
It works from the context you give it: the project, the current task, what you have completed, where you are stuck. It cannot read your code, and it does not need to. Its job is to keep you honest about what should happen next, so that when you go back to your builder, you have one clear instruction and not seven bundled requests.
A quick exercise before you open your builder
Take one minute. Do this on paper if you have to.
Write your product idea in one sentence. Then finish these four:
- My user is trying to ______.
- The first action they take is ______.
- They receive value when ______.
- The smallest complete version of this behaviour is ______.
If you can finish all four, you have the first build task. Type that into your builder, and nothing else. Everything you left out is not lost. It is queued.
Your first build should not prove that AI can generate your app. It should prove that your product can do one useful thing.
That is a much smaller ambition than "launch the whole platform tonight." It is also the one that quietly separates the ideas that go somewhere from the ones that keep restarting.