Also they have a firehose of revenue which makes it possible to hire hire hire, push compensation levels up, and suck the air out of any other interesting companies.

That’s a very small subset of the STEM population.

Small anecdote : I’ve noticed that Ivy leaguers tend to be very polished and particularly good at the game.

Interpersonal skills is a necessary component of playing the game, but the end game is power and influence and winning.

It’s awful when the wrong tool/stack/architecture/etc. is chosen due to someone’s powerplay.

There are little emperors at every firm trying to lobby their influence; then the ones who gain start choking off anything that smacks of competition.

I see this as a huge negative.

"We only hire the smartest folks with the highest GPAs from these schools."

That was a hiring strategy that was famously debunked by Google itself. Now they hire folks that never went to college or were not CS majors. As in they can also be physicists, mathematicians, artists.. etc...

However, I guess they still though think leetcode is a useful signal to identify talent. Or talent that can be trained. So it is more of an aptitude test. Like how a college degree is a signal that you completed something. And a PhD is a signal that you can do research. It doesn't mean other folks can't.

We all know this fails to assess actual work performance (which is subjective anyway) but that is presumably proven by your work history itself. Work performance is a better measure of actual performance than the leetcode proxy any day.

It also most certainly does not select for diversity and different problem solving perspectives. This is a big blind side for any corporation. But you have to be an interested and motivated hiring manager to identify those people. Those are probably the best to hire though. I imagine YC is looking for those types since it is sort of a required trait for a startup.

Some of the best software engineering managers and software product types I've worked with are not CS majors and cannot program.

By now Google has the data to run the analysis for these types studies across the board. They may even have natural controls (people who've transferred roles internally etc...) if they don't have proper randomized controls.