AI is changing how I think about validating a startup idea

I’ve been thinking about the usual advice for starting a software business:

Don’t build too much.

Validate first.

Put up a landing page. Find some beta users. Talk to them. Manually onboard them. Make sure people want the thing before investing in infrastructure, payments, onboarding, etc.

That advice makes a lot of sense when building software is expensive.

I’m starting to wonder if the economics have changed.

Over the past year, AI has dramatically reduced the amount of time it takes me to build software. Setting up authentication, infrastructure, a dashboard, Stripe, deployment pipelines—these aren’t month-long investments anymore.

Sometimes they’re afternoon investments.

I ran into this while building GameSignal.

I initially built an early-access process where someone requests access, I talk to them, manually register their repository, and then they can try the product.

I did that because I didn’t want to spend time building a complete self-service SaaS around something that hadn’t been validated yet.

But then I realized something:

I’m spending more time not building the infrastructure than it would take me to just build it.

So I’m changing the approach.

I’m building the self-service onboarding. GitHub authentication. Repository setup. Trial. Payments.

Then I’m going to put it in front of people.

If nobody wants it, that’s okay.

I’ll change it.

And that’s the other part I think AI has changed: the cost of being wrong has dropped.

If I spend three months building the wrong product, that’s expensive.

If I spend two days building the wrong experiment and discover nobody wants it, that’s information.

I can change direction on day three.

For a long time, one of the rules of startups has been:

Don’t build until you’ve validated.

I’m starting to wonder whether the new rule is closer to:

Build the smallest real version you can, put it in people’s hands, and let reality validate it.

Coding is increasingly becoming the cheapest part.

Waiting is becoming the expensive part.