As a tiny example, look at phone calls. They used to be absurdly expensive. In college I remember having phone bills costing ~30 hours of (minimum-wage) labor. Now it would be hard to explain to an 18-year-old what a long-distance call even was. These days I have effectively unlimited calling from anywhere to anywhere via a handheld device that costs ~7 hours of (minimum-wage) labor/month, and I see lower-cost vendors that provide it for ~4 hours/month.
If we had taken a regulation-first approach, where each new service had to get regulatory approval, I could imagine us still being stuck in the old paradigm, where each phone call had to go through a monopoly operator, and things like Skype were illegal. Or maybe we'd be part-way along the curve, but with incumbents pushing to increase regulatory burden and hobble startups.
So I agree the problem with a default-permit model is that you have more problems to fix, and some can be big. But the problem with a default-deny model is that you miss out a lot of gains. And those, being hypothetical, are easy to underweight against the benefits of the status quo.