❓ Is this necessary?
There’s a type of idea that starts on the builder’s side: "I discovered you can do X with AI, so I’ll build an app for X." The missing question comes from the user’s side: Is anyone struggling because they don’t have X?
Anyone who talks with a lot of entrepreneurs sees an uncomfortable pattern: the vast majority of ideas presented, something like nine out of ten, aren’t pain points. They’re neat things, well executed, sometimes beautiful. But no one wakes up in the morning needing them.
✓ Signs the market is asking for it
- ✓People already pay for workarounds (spreadsheets, freelancers, consultants) to solve this.
- ✓You hear the same complaint from people who don’t know one another.
- ✓The problem measurably costs money, time, or sleep.
- ✓Someone asks "when will it be ready?" before you offer it.
✗ Signs it’s just your own excitement
- ✗The explanation starts with the technology, not the problem.
- ✗You have to convince people they have the problem.
- ✗The compliments are "that’s cool," not "how much does it cost?".
- ✗No user came back after the first test.
💡 The "how much does it cost?" test
Show the idea to five people in the right audience. If no one asks about price, timing, or "can I try it now?", you heard politeness, not demand.
🔥 Solve problems rooted in pain
If you’re not solving something rooted in people’s struggles, you’re solving it just for the sake of solving it. Then no one is in enough pain to justify using your product. This sounds obvious, but it’s the thing people forget most when building becomes easy.
⛽ Why pain fuels adoption
Every tool switch has a cost: learning the interface, migrating data, convincing your cofounder, entering a credit card. Users pay that cost only when the pain of staying as they are is greater than the pain of changing.
That’s why "neat" products get lots of sign-ups and little use. Signing up costs one click; daily use requires a recurring pain point.
| Pain dimension | Question to measure it | Strong answer |
|---|---|---|
| Frequency | How often does this happen? | Every week or every day |
| Intensity | What does it cost when it happens? | Money, a lost customer, a sleepless night |
| Urgency | Does it need to be solved now? | Yes, there’s a deadline, penalty, or risk |
| Awareness | Does the person know they have the problem? | They know and have already looked for a solution |
⚠️ Trap: a pain point only you feel
A problem can be real for the way you work and rare for the rest of the world. Before building, confirm that people who aren’t you or your close friends experience the pain point.
⏳ Will it deserve to exist in 3, 6, and 12 months?
A good discipline is to try living three months in the future. When a new product catches your attention, the question isn’t "is this cool?" It’s: how easy will it be to add this same thing anywhere else three months from now?
The three-horizon exercise
Three months from now
Could a coding agent already do this in an afternoon? If so, your window is short.
Six months from now
Will one of the major assistants offer this as a built-in feature? If so, your audience will get for free what you charge for.
Twelve months from now
Will the problem still exist, even with much better models? If it changes along with the technology (like security), that’s a good sign.
💡 Why ignore most launches
This filter explains why it’s worth ignoring most of the new products that appear every week. If their difference is a feature anyone can add in a month, they’re not worth your attention, as a user or a competitor.
🛡️ Case study: security for AI-built apps
Imagine someone choosing a niche for their next SaaS. Instead of something flashy, they choose security for AI-built apps. There are three reasons, and none is "because it looks good."
No one likes it
Most people avoid security as a topic. That means less competition from people who only want to do things that get likes.
It’s not sexy
It won’t make a viral "look what I built in 10 minutes" video. People who enter do so out of necessity, and they stay.
An endless game of cat and mouse
Every advance in defense leads to an advance in attacks. The problem is never "solved once and for all."
🔁 Why it’s a game of cat and mouse
Frontier models are getting better at defending. At the same time, open models running locally without safety guardrails, often trained by copying the behavior of closed models, are getting better at breaking in.
And there’s one detail: frontier models have barriers that prevent them from helping ordinary people with offensive security tasks. That leaves room for specialized products that harden the infrastructure of companies that aren’t large enterprises.
✓ What’s good about this niche
- ✓Demand grows as AI advances.
- ✓Customers feel afraid, and fear is a strong pain point.
- ✓Solving it requires ongoing updates, which supports a subscription.
- ✓Data about discovered vulnerabilities becomes an asset and content.
✗ What it requires from you
- ✗Credibility: no one trusts security to someone who seems like an amateur.
- ✗Constant updates, with no "finishing" the product.
- ✗Explain a boring topic simply.
- ✗Real accountability when something fails.
🕳️ The information gap is a pain point
Imagine someone in the middle of a client contract who suddenly realizes that the delivered system may have security flaws. Their first reaction is to think they need to hire an expensive specialist. The discomfort isn't technical; it's emotional: they don't know what they don't know.
Even people who could solve something with help from an AI agent often prefer not to spend time or tokens on it. It's the kind of problem they want to hand off to someone they trust and forget about. That "I want to outsource the worry" is a great product signal.
💡 How to find information gaps
Look for phrases like "I don't know if I'm doing this right," "I'm afraid I'm exposed," and "Do I need a specialist?" They show up in forums, groups, and comments, and point to pain that no one has organized yet.
💄 Unappealing problems beat appealing ones
People love chasing flashy problems because they look good on screen: you can record an impressive video in ten seconds. The result is predictable: dozens of nearly identical products competing for the same attention.
A classic example of a bad niche today is content creation automation. There are a thousand ways to do it, and most people already use ChatGPT, Claude, or both for it. Plus, creating content is a personal process: a generic tool never works exactly the way someone likes.
Copy and run
Ask an agent (Claude Code, Codex, or chat) to honestly evaluate your idea using the questions in this module.
Act as a skeptical investor evaluating a product idea.
Idea: <describe your product in 2 sentences>
Audience: <who uses it and in what situation>
How the person solves this today: <tools, spreadsheets, freelancer...>
Answer with a score from 0 to 10 and one sentence explaining each score:
1. Is this necessary, or was it built because it was possible?
2. How painful is it in terms of frequency, intensity, urgency, and awareness?
3. Will this still deserve to exist in 3, 6, and 12 months? Could it become a single prompt?
4. Is this a "flashy" (crowded) or "boring" (low-competition) problem?
5. Is there an information gap (fear of what the client doesn't know)?
End with: "Is it worth continuing?" (yes / no / yes, if...) and the 3 questions I should ask real clients this week.💡 Be wary of high scores
Language models tend to be agreeable. Run the same prompt explicitly asking the model to act as a critic, and compare the two answers.
🧪 Quick module quiz
Three questions. Click an option to see the answer.
1. What's the strongest sign that an idea meets real demand?
2. What does it mean to ask whether the idea "will still deserve to exist in 3, 6, and 12 months"?
3. Why is security for AI-built apps a good example of a niche?
📋 Module summary
Next module:
2.2 - Giants, Data, and Difficulty