That's looking at it from a purely technical perspective. A 20% better Twitter or Facebook (from a technical perspective) is certainly comparatively easy to do, but would still have an approximate value of nil. The value of Twitter or Facebook is not technical (though running services at this scale is technically very challenging): it comes from the people using it. Building another Facebook, even with marginal improvements, is never going to be enough to make enough people switch.

So is the case for AT&T: even if you could build a network competing with the AT&T one it would still have been a largely useless endeavor, unless you could also force AT&T to be interoperable with your network.