State school here: the Harvard grad is lower variance. If you need to reliably fill roles with people who won’t embarrass you, they’re polished and predictable. If you’re looking for diamonds, they certainly are more frequent among the Ivy League graduates, but you’ll pay up for them unlike sifting for them in the rough.

The absolute best software engineers that I've ever worked with most typically have music degrees, by a wide margin. Next most typically they went to Waterloo and then MIT, RIT or CMU. Next group are the self taught ones and then everywhere else.

> Waterloo

The fact that Waterloo maintains a monopoly on its internship/coop model blows my mind every year we do recruiting.

It's the key to their success and other schools are crazy not to do it. But also I think their program is just straight up better.

OTOH, after successfully recruiting from Waterloo for over a decade, my company's recruiters suddenly decided to stop because it was more work for them even though it has consistently given great results.

It's a decision we've been fighting but our recruiting org won't budge.

As someone who’s familiar with both state and Ivy, as both a student and an employer, the variance is most pronounced in the lower quarter than it is in the top quarter. Even the bottom 25% at Harvard are (mostly) competent. With the bottom quarter at a state school, you’ll find some who would have been better off if they had chosen a different path in life. Your job as the employer is to be able to reliably identify that bottom quarter, which isn’t difficult with experience.

I found little practical difference in the top ten percent between an Ivy and a decent state school. In fact, once someone is in the top half, other factors such as organization, drive, and the ability to get along others will dominate. As such, I save money and no longer hire from Ivy’s.

One of my top students at our state school got $400k+ to start as a dev.

When you're good, you're good.

I couldn't help but hear this in my head in the voice of Montgomery Burns.

Excellent.

"Legacy admissions" are fine because intelligence is 85% heritable.

That’s beside the point. Look up ‘regression to the mean’. If two highly gifted parents each with an IQ of 150 have children, their children will very likely be in the upper half or even perhaps upper quarter of the IQ distribution, but very unlikely to be the top 0.025% ~ IQ of 150. That is, their kids will be above average, but likely not Ivy quality.

There’s a reason legacy admissions need a thumb on the scale.

“Not ivy quality” - Only people with 150 IQ get into Harvard? News to me.

A hallmark of the argumentative mind is nitpicking and missing the larger point.

It’s just an internet comment, but thanks for the tip, Freud.

This is fallacious.

If intelligence is so heritable, why do children of Harvard grads need preferential treatment?

They don’t do the hard work part?

* Unless they happen to be scholarship students. Scholarship students traditionally have to work harder than everyone else because they can't afford to buy homework or tests, or to pay people to take tests for them. One potential mark against them is the possibility of grade inflation endemic to the institution, but this maybe more or less variable in a particular STEM discipline.

Also, state research universities didn't/don't have as much grade inflation because they're intent on "competing" in the ranks without having the legacy, endowment, or name branding that Ivy's and Pac-NN's, so typically all they have (outside of NCAA sports $$$). University academic department staff are worried about bolstering their public image and of winning the approval of accreditation organizations.

Frame of reference: Most all of my college-prep public high school friends went to Harvard, MIT, Stanford, or CalTech while fewer, like me, went to University of California universities (UCs) like Berkeley or UC{LA,SD,SB,D}. I'm only certain of a few certain universities of that era were more or less more rigorous by comparing notes on specific courses and how team participation was regulated fairly or not. In 2001, we had a intro networking course requiring creating a forked caching HTTP/0.9-1.0 network web proxy server as one project of many.