Amazon has a leadership principle for this: invent and simplify. 37signals has a name for the second half: the wind tunnel test. AI has made the first half easy. Nobody has solved the second.

You can ship a working prototype in days now. Test the idea, get real signal, move on. That part is basically solved. But "we shipped it" and "we're ready to build on it" are not the same milestone — and AI is widening the gap, not closing it.

To get a prototype working, you take on complexity you don't fully understand yet. Dead ends, overbuilt abstractions, trade-offs made under time pressure. That's normal — it's how you find out where the value actually is. The mistake is what happens next: teams keep building on top of the proof instead of stripping it back first.

Skip that step and you're not just carrying technical debt. You're carrying conceptual debt ("why does this exist"), operational debt ("only one person understands it"), cognitive debt ("every change feels risky"). Velocity drops. Onboarding gets harder. Small changes start to feel expensive.

That's what the wind tunnel test is for. Put the prototype under pressure. Anything unessential flies off. What's left is the part that actually mattered.

Agentic tools make this more urgent, not less. They're extremely good at inventing. They'll also happily hand you unnecessary complexity — some visible, some nobody fully reasoned through. AI can build it. A human still has to decide what's worth keeping.

Let complexity happen while you're discovering the truth. Prove the value. Then stop, and simplify ruthlessly before you scale.

AI helps you ship. Simplification helps you scale.