[api]

"And the only way to figure out what those improvements are is to build them and see if a massive number of new users show up."

That's sort of what I wrote here:

https://www.zerotier.com/blog/?p=74

You can't test those markets with mockups and quick hacks, at least not easily. Those markets have a tendency to sit on the sidelines for a long time, since only a lunatic would actually build a real product just to test a market.

Despite having the business school dogma of "test the market before actually building something" banged into my head with a hammer by every MBA-ish type I've run into, I am still skeptical of it. I don't think it's always possible to test without building.

YouSendIt and all the other file sharing sites were really pretty simple hacks, exactly the kind of "MVP" that a b-school sort would suggest tossing together to determine if anyone cares about sending each other files. Dropbox on the other hand was a seriously hard piece of engineering. The first time I used it, I immediately thought "whoa, that was not easy to build." You see, the file sharing problem is not the file sharing problem. It's really a distributed database problem in disguise. All the others built file sharers, but Dropbox built a centrally-supervised distributed database ala BigTable or GPFS.