given that i don't really agree with his talk, i'm going to defend it anyway.

what has happened with automobiles? when first introduced, they were the domain of enthusiasts only. then they spread out to a wider audience, an audience that was frequently irritated by the maintainance requirements, breakdowns and poor behaviour of the machines. that gets us to the 1980s.

and now .... cars are almost all of a near-uniformly high quality, and that quality exceeds the levels attained even in 1970's era elite vehicles. you need to know almost nothing about them to use them (other than knowing how to drive). their reliability and longevity have improved dramatically, as have their safety qualities. maintainance is generally simple - regular oil and other fluid changes, less frequent tire replacements, occasional body work due to damage. oh, and much better fuel mileage too. of course, every single one of these improvements isn't the end of the road (no pun intended), and new engine and propulsion systems still offer huge new areas of potential improvement, along with self-driven cars for some situations.

if i give this presentation the benefit of the doubt, what he's arguing for is that *consumer* software, having arrived at feature parity, should be planning to evolve in the same general direction that automobiles have. that is: away from an object only maintainable by an enthusiast who is willing to take the time to learn about it, passing through the phase of an unreliable, short-lived object that gets fixed by others, to reach a state where fundamentally, the owner/user doesn't need to know anything and the thing just works today, tomorrow and the next day.

i don't think i want to write software for people who think software should be this way, but i will concede that if you accept the premise, it has its own compelling internal logic.