You can build anything now. That's exactly the problem.

There's never been a better time to build a startup. And that's precisely why so many of them will fail for the same old reason.

Think about what used to take a team and six months. A landing page? Describe it to an AI and it's live before lunch. A pitch deck? One prompt. A working MVP? A weekend, maybe less. Market research, a GTM plan, a brand, a demo video — all of it has collapsed from months of specialized work into an afternoon of prompting.

This is genuinely amazing. It's also a trap.

Because when everyone can build everything, building stops being proof of anything.

The thing AI still can't do for you

Here's the uncomfortable part. AI can generate the deck. It cannot tell you if anyone wants what's in it. It can spin up the MVP. It cannot tell you if a single person will pay for it. It can write you a beautiful GTM strategy. It cannot tell you if the market you're aiming at actually feels the problem you think it does.

Every one of those artifacts — the deck, the site, the MVP — is a guess, dressed up to look like traction. And now that the guesses are free and instant, we're all drowning in them. The scarce thing isn't the ability to build. It's knowing whether you should.

That's the one part that still can't be faked, prompted, or shortcut. Someone real has to want it.

The old way, and why it's still the only way (for now)

For years, the only route to that answer was brutal: build the whole thing first, ship it, and then find out if anyone cared. Weeks — often months — of work before you got a single honest signal. Most founders never make it that far, and the ones who do frequently discover, far too late, that they built something nobody was waiting for.

AI made the "build the whole thing" part faster. It did nothing to make the "find out if anyone cares" part safer. If anything, it made the noise worse.

So build with AI — but end every step with the honest question

None of this means don't use the tools. Use all of them. Just don't mistake the output for validation. Here's the healthy version of the workflow, where every step ends with the question that actually matters.

Found a problem? Pressure-test whether it's real before you fall in love with it.

…but a convincing answer from an AI is still not a real person telling you it hurts.

Found a solution? Sharpen it.

…but an elegant solution to a problem nobody has is just elegant.

Need a go-to-market? Draft one.

…but a plan to reach a market that doesn't want you is a map to nowhere.

Need a deck? Make it.

…but a deck is a story about demand. It is not demand.

See the pattern? Every tool gets you a better guess. Not one of them closes the gap between "I built it" and "the market wants it." That gap is where good ideas quietly die.

That gap is where Few Good Ideas begins

Few Good Ideas exists for exactly that moment — after you've done the thinking, before you sink weeks into building the wrong thing. You put your idea in front of the market early: real founders, operators, everyday people, validators, and investors react to it. Not a panel of judges. The market.

What comes back is a single number — your EMV, or Early Market Validation score. It doesn't tell you whether you're a good founder. It measures one thing well: how much the market actually wants this, right now. It's the proof that a deck only pretends to be.

Building got easy. Being right didn't. The founders who win the next few years won't be the ones who can build fastest — everyone can build fast now. They'll be the ones who find out fastest whether they should.

That's the test. Stop guessing.