The 10-Year Idea Test: The Validation Question Nobody Asks (And the Only One That Matters)

It was 2023 and I had validated my idea perfectly.

Surveys. Interviews. A landing page with a waitlist. First payments before writing a single line of code. Everything the books say you should do — I did it.

Two years later, the project was dead.

Not from lack of initial demand. It died because I validated what most people validate: whether someone would pay in month one. Nobody told me that’s only half the question.

The other half is whether someone will still be paying in year ten.

The problem with how we validate today

Only 34.9% of businesses survive ten years. Most validation frameworks we consume are designed to get you through the first twelve months, not the first decade.

And here’s the interesting part: they’re completely different questions.

This week I was reading the case of a researcher with no coding background who launched a SaaS using Claude and Cursor, reaching paying customers within 25 days [2]. Flawless launch validation. But ask the 10-year question — what happens when five other people do exactly the same thing next month with the same AI? — and the answer gets complicated.

Contrast that with this other case: YouTube videos with barely a hundred views generating a solid base of paying subscribers month after month [1]. From the outside it looks like a distribution failure. From the inside, it’s a compounding search engine that gets stronger every week.

One is validated for the sprint. The other for the marathon.

The two validations you actually need

Launch validation: Is there enough demand to get started?
Longevity validation: Will that demand still exist, and will I be the one capturing it?

We do the first one reasonably well. Almost nobody does the second.

Here are three concrete questions for the 10-year test. Apply them to your next idea before writing a single line of code.

The 3 questions of the 10-year test

Question 1: Will this problem still exist in 2036?

Not the market. The problem.

Some problems disappear with technology (converting file formats, basic document translation). Some problems grow larger over time (managing regulatory complexity, communicating clearly, making decisions with incomplete information).

The question isn’t “are people searching for this now?” but “will people be searching for this in ten years, or will AI have solved it for free?”

Exercise: Write down the problem you solve. Then ask: what would have to be true in 2036 for this problem to no longer exist? If that list of conditions is short and plausible, you have a fragile problem.

Question 2: Does your business get stronger or weaker over time?

This is the question I most enjoy asking and that fewest people can answer.

Some businesses accumulate advantages over time: proprietary data, organic search reputation, network effects, community. Every month that passes, they’re harder to compete with.

Other businesses erode: competition copies you, margins fall, differentiation disappears. They need to run faster just to maintain position.

Back to the YouTube example with few views but significant revenue [1]: that’s a business that gets stronger. Old content keeps generating searches. Accumulated authority is hard to replicate overnight. The filtered audience (people with a specific problem actively searching for it) is more valuable than a viral audience.

Exercise: Draw your business at year 1, year 3, and year 7. What does year 7 have that year 1 doesn’t? If the answer is “more customers,” that’s not enough. You need structural advantages that can’t be bought or easily copied.

Question 3: How many people can replicate this in 90 days with AI?

In 2026 this is no longer a theoretical question. A researcher with no React knowledge built a functional SaaS with paying users in less than a month [2]. The technical barrier to entry has collapsed.

What hasn’t collapsed: distribution, accumulated trust, proprietary data, and deep domain knowledge.

If someone can replicate your product in three months with Claude and Cursor — and sell it to your same audience — your launch validation guarantees nothing about year ten.

The right question isn’t “can I build this?” but “why will customers choose me in five years when there are ten equally functional alternatives?”

Exercise: Describe your idea to a hypothetical competitor with access to the same AI tools you have. How long would it take them to have a competitive MVP? If the answer is less than ninety days, you need to identify which element of your business is non-replicable: your distribution, your data, your audience, your domain reputation.

How to integrate this into your validation process

I’m not telling you to abandon fast validation. Keep building the landing page, keep talking to potential customers, keep building the MVP in weeks instead of months.

But before you start, spend an afternoon answering these three questions in writing:

1. Will the problem I solve still exist in 2036? (Be specific about what would have to change for it to disappear)
2. What assets does this business accumulate over time that can’t be bought? (Data, distribution, community, domain reputation)
3. What protects me from being replicated when the technical barrier is nearly zero?

If you don’t have clear answers to all three, that’s not a signal to stop. It’s a signal that you need to design those answers before validating the launch.

Because you can reach year one with everything looking green and arrive at year five with nothing.

The 34.9% that survives ten years didn’t have more luck. They asked better questions before starting.

Are you validating an idea right now? Take the three questions and write me what you find. I’m interested in seeing the concrete cases.