I take it bad founders means the founders that run less than billion-plus companies.

I would absolutely believe that founders who (go on to) run billion dollar companies make a point of replying to YC partners very quickly. "Successful executives are highly responsive to investor and advisor emails" seems eminently plausible. It doesn't suggest that they're equally responsive to all emails, but they've got a sense of who's important/needs to feel important.

I'm sure they do. But you can also interpret that as 'bootlickers are the kind of people I like'. And that is not necessarily equivalent to 'good founders'. So I think it is a bit of a thin element to judge people by.

I think it is maybe best reframed as "good founders from the perspective of those who control capital".

Whether these people ultimately improve society, or create a better sense of purpose for their employees, or provide visionary direction for the company at a higher rate than other founders is kind of orthogonal (or perhaps anticorrelated) to being good stewards of invested capital.

Founders can, to some extent, get help with (or succeed financially despite the lack of) the other things, but I think it can be reasonably argued that if they're not personally regarded as good stewards of capital, then the whole enterprise is in doubt.

Precisely. Good founders, bad founders, in the eyes of the beholder.

I'm pretty sure I have a completely different opinion on what constitutes a good founder and what constitutes a bad one compared to Sam Altman, fortunately I don't have enough clout to make authoritative statements on the subject.

Except that he only found out there was a correlation between startup success and responsiveness by doing a historical analysis.

Or that's the bit that sounded like a nice story.

Startup success is random enough that it'll be correlated with a great variety of interesting things. A good scientist makes a hypothesis based on a casual model first, then constructs a test for it. Data mining for correlated things is a recipe for apophenia.

Goodhart's law would tell you what any metric-driven understanding of the cause of success would cease to be useful as soon as it was announced.

It remains plausible that the connection is neither causal nor spurious, but after the announcement of the correlation, the correlation becomes useless

P|Q != Q|P

No, but P(Q|P) = P(P|Q)P(Q)/P(P)