I finished the first working version of IdeaMade — free preview coming next week
A few days ago, I shared the workflow I’m exploring for AI app building:
Idea → Requirement clarification → Editable Blueprint → App
After several days of development, the first part of that workflow is now working.
I’ve completed:
The first version of the IdeaMade website
AI-generated clarification questions
An editable product Blueprint
Users describe what they want to build, answer 3–5 questions about users, workflows, business rules, and missing decisions, and then receive a Blueprint containing:
Product goal
Users and roles
Core workflows
Key assumptions
Main pages
The first public version will be available next week and will be free to try.
To be clear, this release focuses on requirement clarification and Blueprint generation. It does not generate the complete application yet.
The main thing I want to test is whether defining the product more clearly before development can reduce generic first versions and unnecessary iterations later.
I’ll share the public link next week. In the meantime, I’d be interested to know:
What information would you expect to see in a Blueprint before letting AI build your app?
Your blueprint list is solid on the "what" (goal, users, workflows, pages), but the thing that actually reduces generic first versions and wasted iterations, your stated goal, is mostly missing. Generic output isn't caused by underspecified features, it's caused by underspecified constraints and edge cases. So the highest-value additions are the things people forget to say, not the things they already know.
What I'd add:
The anti-goal. What this explicitly is NOT. "Not a social network," "no payments in v1." AI over-builds because nobody told it where to stop, and half of "unnecessary iterations" is ripping out what it helpfully added. Bounding scope prevents more rework than defining it.
The unhappy paths. Your workflows capture the happy path. Iterations come from cases nobody specified: empty form, failed payment, two users hitting the same record. A blueprint that forces "what happens when this goes wrong" catches exactly what becomes bug-fix iterations later.
The one non-negotiable. Every product has a single thing that must be right or it's worthless (booking app: no double-bookings; chat: never lose a message). Flag it as never-compromise and the AI builds around the real constraint instead of weighting every feature equally.
Priority. Not just "main pages" but which is core vs supporting. AI treats a flat list as equal priority and builds a shallow version of everything instead of a deep version of the one thing that matters.
The deeper point, and your actual wedge: the value of a blueprint step isn't documenting what the user knows, it's extracting what they haven't thought about yet. Anyone can restate "I want a booking app." The tool earns its place by asking the question that makes them realize they never decided what happens to a cancelled booking. Your clarification questions are the product, not the blueprint format. Optimize for questions that surface the un-thought decision.
To answer directly: scope boundaries, failure behavior, the one non-negotiable, and priority order, more than another restatement of features.
I like that you're treating clarity as a deliverable instead of something that's expected to emerge during development.
A lot of AI tools accelerate building, but the harder question is whether they've accelerated understanding first. Those are very different milestones.
Generating a Blueprint is easy.
Generating a stable one from the same requirement is much harder.
I added stricter structure and validation. The results are already much more consistent.