I think the point is that to go faster, you have to throw disproportionately more people at something, and probably build it in a different way to accommodate that, and test the integration of their work more. Cost is people x time, not just time. Three people working steadily for a year will be far more efficient than 20 people working for 6 months.