It's an acknowledgement that the early days of a startup don't need to look like the later days. At 42Floors, we met with every landlord in person and then manually typed in every listing. Obviously, we can't do that at scale, but it was an important step to getting our startup off the ground when no one had yet heard about us. Later on, when we had built up credibility, we built web scrapers and integrated data feeds

The classic YC example is AirBNB. Brian & Joe flew to NYC every week to personally recruit hosts, take pictures of their apartments, etc.

Clearly recruiting hosts in person, one by one, doesn't scale. But it's okay to do these kinds of things at first because they help you understand your users & help create super-users who will go out an evangelize your product.

I believe that this important mantra is illustrated by the use of a Minimum Viable Product (MVP) technique. This is described in detail by Eric Ries in The Lean Startup (pgs. 99-102), wherein he explains the tremendous amount of learning and customer discovery that a startup founder/team can experience first-hand by making the effort to personally provide initial customers with customized one-on-one "concierge treatment"...

I like to view it as in the beginning do as much as you can get to get open-ended discussion with prospective customers. Later on use that learning to build a more scalable/product-focused approach optimized for acquisition/growth.

The first thing to do is actually validate that people need your product. If you can't win in the short term then there won't be a long term.