Ask HN: How much user validation is enough? When to build?

It seems that people have gotten a bad taste in their mouth with respect to building useless products due to lack of user validation. And that's fine. But it also seems that people are too afraid to build anything until users give them a green light every step of the way. I.e. we've gone too far the other way.

I'm literally stuck in meetings sometimes debating whether or not to put a hold on the next electronic prototype of a product because we're unsure how to position the UI: pixels on the goddamn screen, of the old one. Because we can't move forward until we know exactly what we're going to build... Excuse my frustration, but if this is your legitimate source of contention as a startup, then you're probably building a commodity rather than a technology. Which is fine, but definitely isn't the case for all companies.

How many user surveys are enough to show interest in a concept? How many email signups on your landing page should it take to move forwards and build out a product? Here's a harder question: you have a prototype, you've done some user surveys - when to direct the team's attention to finishing out the product rather than market research (a serious issue for say a 3-person team).

The way I see it, anyone can make a decision with perfect information or someone telling you just do it, while the opportunity for a new venture lies in moving fast in the general right direction. There's a trade-off: by the time you're certain people want something, someone's probably already made it. I like to believe that logic and first-principles reasoning can be an efficient alternative to never-ending user validation.

Thoughts?