“They laughed at Columbus, they laughed at Fulton, they laughed at the Wright brothers. But they also laughed at Bozo the Clown.”

Note also that Graham tells us what the “simple solutions” of Viaweb and Y-Combinator were that made them successful, but fails to do so for Arc. I wonder why.

I don’t read Paul’s essay that way. The way I saw it, he was making an observation about a pattern he’d noticed. You’re right, it doesn’t amount to a defense; but that’s only a weakness if it was intended as a defense. I don’t see it the same way: I think he said (roughly) the first two sentences of your summary, but he didn’t say the third. Rather, he postulated that new solutions that have the six common attributes he noticed, which may (or may not) be successful in the long term, will *always* produce criticism. The conclusion I draw is that the information content of such criticism to those particular kinds of solutions is lower - it needs to be examined in specifics (to see it isn’t *solely* a reaction to the first five attributes), as well as taking time into account (for the sixth attribute, the rapid iteration).

This conclusion - that the information content of the aggregate criticism is lower - is ineluctable, because it has a measurable historical false positive rate.