Think through and scope your project with AI (before touching code)

I'm going to tell you something I learned the hard way building 13 products: the most important part of a project is NOT writing code. It's thinking things through BEFORE writing it. In surgery we have a saying — 'the operation is won or lost in planning, not in the operating room.' Software is identical. Most projects that die don't die from bad code: they die because they were never well thought out. Something nobody needed was built too quickly, or something so big that it was never finished.

This lesson is the one almost no course gives you, and it's the one that will save you the most. You're going to learn to use AI not to get answers, but to help you THINK through your own idea better: to research it thoroughly, to scope it out, and to find its holes before investing weeks into it. By the end, you'll have a 'Master Prompt': the solid foundation from which everything else starts.

AI is not an oracle: it's your thought partner

Here's the most valuable mindset shift in this lesson. Most people ask AI 'give me a business idea' or 'tell me if my idea is good.' That's using it as an oracle — and it will agree with everything, because it's trained to please you. The smart builder does the opposite: they use AI to INTERROGATE them. They ask it to question their assumptions, ask the uncomfortable questions, and play devil's advocate.

The difference is huge. An AI that applauds you teaches you nothing. An AI that challenges you to defend every assumption of your idea forces you to deepen your own thinking — bringing to light the holes you wouldn't see on your own. That back-and-forth, that active and conscious feedback, is where a vague idea turns into a solid project.

Research thoroughly before forming an opinion (NotebookLM, Gemini, Claude)

Before deciding ANYTHING about your project, research the field. Not from inside your head — from real data. Modern AI lets you do in an afternoon what used to take weeks of market research. These are the three tools I use, each for its own purpose:

• NotebookLM (Google): upload sources — articles, Reddit forums where your user complains, studies, competitor reviews — and ask it to SYNTHESIZE them. It only reasons over what you give it, so it rarely hallucinates. Perfect for figuring out 'what real pain points my user has' and 'what people say about existing products.'
• Gemini / Claude with search: to explore the open landscape — who else does it, how they solve it, market pricing, trends. Ask for sources and verify the important ones.
• Claude for reasoning and structuring: this is where you think OUT LOUD. Paste what you researched and ask it to help you draw conclusions, scope things down, and find your angle. It's your strategy partner.

The 5 pillars that scope a project

A well-thought-out project answers five questions before writing a single line of code. They aren't just paperwork: each one saves you from a different kind of failure. Skip one, and that's exactly where you'll get stuck three weeks from now.

• REAL NEED — What concrete pain point are you solving, for whom, and what does that person do today without you? If there's no real pain point and no poor 'current solution,' there is no project.
• SPECIFIC USER — For whom EXACTLY? 'For everyone' means for no one. The more specific (context, moment of use), the easier it is to build and sell.
• SCOPE (MVP) — What is the ONE thing your v1 must do brilliantly? Everything else gets postponed. Scope creep is the #1 reason projects are never finished.
• MARKET AND ANGLE — Who else is doing it, and why you, why now? You don't need to be first; you need a real angle that sets you apart.
• SUCCESS AND MODEL — What does 'winning' look like in 3 months? Free, paid, freemium? How do you get your first 10 users? Define the goal before running.

Notice something: NONE of these five are technical. None of them talk about code, frameworks, or databases. That's intentional. Technology is the easy part today — AI helps you with that throughout the course. What's hard, and what determines whether your product lives or dies, is getting these five right. That's why we dedicate an entire lesson to them.

Do it now: build your Blueprint

Below is the tool you'll use to ground your project. Write down your idea, answer each pillar in YOUR own words, and use each pillar's prompt to let AI interrogate you and dig deeper. Don't fill it all in at once by guessing: go to Claude or Gemini with each prompt, let it challenge you, and come back to refine your answer. That back-and-forth is the real work.

MY PROJECT CONTEXT
I'm building: [name + what it is in one sentence]
The problem it solves: [the real need]
For: [specific user]
v1 scope (MVP): [the one core function]
Differentiation / angle: [why me, why now]
What success looks like: [metric + model + first users]
Stack: Next.js + Firebase + Vercel (Méndez Labs course).

Use this context as the foundation for EVERYTHING we build. When I ask you for a piece, respect it and tell me if anything clashes with this plan.

The Master Prompt: your idea becomes your foundation

See what the tool does at the end? It gathers your five answers into a single block of context: the Master Prompt. That text is your project's most valuable asset. Why? Because remember the 4 parts of a good prompt from the previous lesson — Context, Task, Format, Constraints. The Master Prompt IS your permanent, distilled context.

From here on out, every time you ask AI for something in this course, you start by pasting (or keeping in mind) your Master Prompt. That way AI never builds blindly: it knows what you're making, for whom, at what scope, and where you're heading. The difference between a student who fights with AI and one who directs it like a partner is often simply that the second one ALWAYS starts from their Master Prompt.

Challenge your idea before building it

Here's the part almost no one does that will save you weeks: before writing a single line of code, put your idea through an honest interrogation. AI is perfect for this — not to agree with you, but to ask the uncomfortable questions you avoid on your own. Paste your Master Prompt below and choose a lens: a skeptical investor, a product strategist, or your own user. Read what it gives back and adjust your plan. That intentional back-and-forth is what turns a nice idea into a solid project.

Worked example: what Momentum's looks like

To show you what it looks like finished, here's Momentum's Blueprint (my example project for the course). Notice how each pillar is a concrete sentence, not an essay:

Idea: Momentum — un diario donde escribes tu dia y la IA te devuelve claridad.
Necesidad: la gente quiere reflexionar pero escribir a mano se siente pesado y
           no obtiene nada a cambio, asi que lo abandona.
Usuario:   profesionistas 25-40 que ya intentaron llevar un diario y lo dejaron.
Alcance v1: escribir una nota + recibir UN insight de IA. Nada mas.
           (Rachas, graficas y compartir = despues.)
Mercado:   hay apps de diario, pero ninguna te devuelve un coach que te conoce.
           Ese es mi angulo: la IA que encuentra tus patrones.
Exito:     10 personas escribiendo cada semana en 3 meses; plan Pro a $5/mes;
           primeros usuarios de 2 comunidades donde ya estoy.

With that, when I ask AI in Module 5 to 'create the writing screen for me,' it doesn't start from scratch: it knows it's a minimalist journal, who it's for, and that the highlight is the AI insight. Every piece it builds will fit the plan. That's what starting from a solid foundation looks like.

Key takeaways

• The project is won or lost in planning, not in code. Think before you build.
• Use AI as a thought partner, not an oracle: ask it to INTERROGATE and challenge you, not applaud you.
• Research with real data (NotebookLM to synthesize sources, Gemini/Claude to explore and reason), not from your assumptions.
• Scope it out with the 5 pillars: real need, specific user, scope (MVP), market/angle, success/model.
• Distill everything into your Master Prompt and ALWAYS start from there. It's your living foundation for the entire course.

AI builds in minutes what used to take you weeks. But only you can decide WHAT is worth building. That decision, well thought out, is your true edge.