enhanced human
The PulseThe Shortcut
Sign InSign Up
Lesson 1 of 5 · 15 min

Lesson 1: Picking the Right Idea

Course: Ship Your First AI App

Estimated time: 25 minutes

Difficulty: Beginner Builder

You have an idea. It's been sitting in your notes app for weeks, or maybe it just hit you in the shower this morning. Before you open your editor, before you think about the tech stack, before you do anything else - you need to know if it's worth building.

Most ideas aren't. That's not cynicism, that's just how ideation works. The filter is fast validation, not long analysis.

Why Most Ideas Die in Production, Not Conception

Here's the pattern: builder has an idea, spends 3 months building it, launches to silence. The problem wasn't the code. It wasn't the design. It was that nobody actually wanted the thing.

The graveyard is full of technically solid products that solved problems nobody had, or solved real problems but for audiences too small to matter, or solved real problems for real audiences that already had good enough solutions.

You can skip most of this graveyard with one framework applied early.

The Validation Framework: 5 Questions in Order

Run every idea through these five questions before you touch code. The order matters - each question filters out bad ideas before you spend time on the next one.

1. Problem: Is it real and recurring?

A real problem is one people are already paying to solve, already complaining about, or already building workarounds for. A recurring problem is one they hit every week, not once a year.

Ask yourself: do people Google this problem? Are there Reddit threads about it? Do people pay for imperfect solutions to it right now?

If you can't find evidence that people actively suffer from this problem, the idea is probably a solution looking for a problem. Those rarely work.

Good signals:

  • Existing paid tools in this space (validates the problem, not the competition)
  • Reddit/Twitter complaints about the current solutions
  • You personally have this problem and pay for a bad solution

Bad signals:

  • "I would use this" is the only validation you have
  • The problem only happens once or twice a year
  • Nobody is currently paying anything to solve it

2. Audience: Can you reach them?

A real problem with no reachable audience is still a dead end. "Everyone who uses a computer" is not an audience. "Freelance designers who do client work on Figma" is.

The question isn't whether the audience exists. It's whether you can find them, talk to them, and get your product in front of them without a $50k marketing budget.

Reachable audiences have:

  • Specific online communities (subreddits, Discord servers, Slack groups, forums)
  • Specific influencers or publications they follow
  • Specific job titles or tool stacks you can target

If you can't name three places where your target audience hangs out online, the audience definition is too broad.

3. Existing Solutions: What are people using now?

This is where most builders get discouraged and shouldn't. If there are no existing solutions, that's often a bad sign - it might mean there's no real market. Existing competition is validation.

What you're actually looking for here is the gap. What do existing solutions do badly? What do users complain about in their reviews? Where does the "good enough" solution fall short for a specific segment?

Go to G2, ProductHunt, and the App Store. Read the 3-star reviews of the market leader. That's a list of problems waiting to be solved.

Common gaps you can build on:

  • The existing tool is too expensive for smaller users
  • The existing tool requires technical setup that alienates non-technical buyers
  • The existing tool is great for one use case but terrible at the adjacent one
  • Nobody has built the "Notion for [specific industry]" version yet

4. Your Angle: Why would someone switch or choose you?

You don't need to be 10x better across the board. You need to be meaningfully better for a specific person in a specific situation.

"Same as Competitor X but cheaper" is a race to the bottom. But "same as Competitor X, built specifically for freelance designers, with client approval workflows they actually need" is a position.

Good angles:

  • Focused on one use case the general tool handles awkwardly
  • Built for a specific workflow that generalist tools ignore
  • Faster or simpler for users who are overwhelmed by the power tools
  • Cheaper and simpler for the segment that doesn't need the enterprise features

Your angle should fit in one sentence. If it takes a paragraph, it's not sharp enough yet.

5. The 1-Day Test: Can you validate it before building it?

This is the most important question in the whole framework. Before you build anything, can you find out if people would actually use it?

The fastest validations:

  • Fake door landing page. Put up a landing page describing the product. Add a "Join the waitlist" button. See if people sign up. Takes 2 hours to build with an AI-assisted Carrd or Framer page.
  • Manual version first. Do the thing manually for 3-5 people. If it's a tool that generates weekly reports, generate the reports by hand. If people find it valuable, build the automation.
  • Sell before building. Tell 10 people in your target audience about the idea and ask if they'd pay $X/month for it. Offer to add them to a founding member list. Actual payment commitments are the only real validation.

If you can't design a 1-day test for your idea, the idea is too vague. Make it specific enough to test.

Real Examples: Ideas That Worked vs. Flopped

Flopped: A developer built a "social network for book lovers" with AI-generated reading summaries. The problem: Goodreads already existed. The audience: too broad. The angle: nothing meaningfully different. The 1-day test: skipped entirely. Result: 6 months of work, 12 signups, abandoned.

Worked: A solo developer noticed that his Notion workspace kept getting messy because he and his clients edited the same documents. He built a simple tool to snapshot Notion pages as read-only shareable PDFs with one click. Specific problem. Specific audience (consultants and agencies). Clear gap (Notion has no clean client-sharing solution). Built a landing page and got 40 signups in 48 hours before writing a line of code.

Flopped: An AI writing tool for "everyone." Saturated market. No angle. No reachable audience. No validation. Launched to silence.

Worked: An AI tool specifically for writing cold outreach emails for SDRs at B2B SaaS companies. Specific audience on LinkedIn. Clear problem (writing 50 personalized emails per day is brutal). Validated by posting in 2 SDR Slack communities and getting 60 signups in a week. Built a focused MVP, got to $3k MRR in 3 months.

The pattern in the ones that worked: they're specific. Ruthlessly specific. The audience is named. The problem is named. The gap is named.

Using AI to Run This Framework Faster

Here's the part where AI actually helps. Instead of spending a week on this analysis, you can run through it in an hour.

Use this prompt with Claude or GPT-4:


I have an app idea: [describe your idea in 2-3 sentences].

Help me run through a quick validation framework:
1. Is the problem specific and recurring? What evidence would confirm it?
2. Who is the most specific audience that would have this problem?
3. What existing solutions exist? What do users complain about in their reviews?
4. What angle would make this meaningfully different for a specific segment?
5. What's the fastest possible 1-day test I could run?

Be honest. Tell me if this looks like a weak idea and why.

Then take that output and actually do the research it points to. Search the subreddits. Read the G2 reviews. Check if there's a Product Hunt page for competitors.

AI is good at suggesting where to look. You still have to look.

Exercise: Validate One Idea in 60 Minutes

Pick your best current idea - or generate one if you don't have one yet. Then:

Minutes 0-10: Write the one-liner

Write one sentence that completes: "This is a [tool type] for [specific audience] that [solves specific problem] by [your approach]."

If you can't write that sentence, keep refining until you can.

Minutes 10-25: Find the evidence

  • Search Reddit for the problem. Find 3 threads where people complain about it.
  • Find at least 2 existing paid tools that solve something similar.
  • Read the negative reviews on those tools. Screenshot 3 complaints that your idea could address.

Minutes 25-40: Find the audience

  • Name 2 specific subreddits, Discord servers, or online communities where your audience hangs out.
  • Find 1 person you could actually talk to about this problem in the next 24 hours.

Minutes 40-55: Design the test

  • Write the landing page headline you'd use to test this idea. One sentence that describes the value.
  • Choose one: fake door landing page, manual version first, or sell before building. Describe exactly what you'd do.

Minutes 55-60: Verdict

Score the idea on a simple 1-5 scale for each: Problem (real and recurring?), Audience (reachable?), Gap (does one exist?), Angle (do you have one?), Test (can you run one?).

If you score 4 or 5 on all five: move to the next lesson.

If you score 3 or below on any: either sharpen the idea or pick a different one.

Key Takeaway

The fastest path to a successful launch starts with 60 minutes of validation, not 60 days of building.

What's Next

Now you have an idea worth building. Lesson 2 is about the stack that lets you ship it as fast as possible - and why AI coding tools completely change which decisions matter.

Next