1590 SAT scorer, working at PhD level role at Google at 18, sues Cornell and other schools over racial discrimination.

Nope look at what happened at the UCs. They’ve been race blind for decades

Shhhh you’re not supposed to point out that his essays or interview have gotten him rejected.

let's see here.

1) dad did the paperwork for the e-signature company

2) dad worked at google and he magically gets a job at google

3) he gets promoted at google after he's been turned down for college

4) he doesn't look like he ever went to college

5) he took no AP or honors frosh fall, and few of them until junior year

6) he makes a B+ in an elective class as a soph hence no 4.0 U/W, tripping up his cute game

7) they oddly don't know his class rank

8) they oddly give out his junior year GPA and grades and not his senior ones

9) his school does offer at least AP language classes

10) he doesn't say when he applied, whether any were early decision or normal senior spring admissions

11) he doesn't say if his senior grades changed

12) he doesn't say if he applied to multiple early decision programs where the rules maybe told him you have to do only one, and where they might turn you down for that alone

Fencing doesn't have to be that expensive. Absolute nonsense. That's just a smokescreen to weed out people without the jewel-adorned-power-of-belief cojones of fencing prospects

i do think some of the denials are odd but the consistency of it suggests some systemic problem. they are trying to say it's race but most of his applications are either elite schools, top CS programs, or UC where it's been illegal a while. so to me it's something tripping him up over and over. maybe the rank, maybe the lack of many honors/AP, maybe typos in an application, maybe bad essays or interviews.

something is off where everyone says "nah" except a couple out of state publics.

the funny thing about the merit folks weighing in is they don't seem to want to hear details. part of merit is details.

they also seem to want to ignore that his list of applied schools reads like a who's who of CS programs, where almost every kid applying is some smart, high test score, high grades, programmer. you're then the kid who avoids honors classes.

It’s “banned” but they use other proxies for it. As a trend, schools with LOWER average AP scores have higher acceptance rates into UCLA.

Why? The reason is obvious.

re merit and interviews, my brother had similar GPA and scores to me, and had played sports in HIS but wasn't recruited like i was. applies same school i got in easy. we were both top half of the profile. goes to interview and apparently in naive honesty keeps chatting about his interest in a major the school doesn't have.

reject. makes me angry to this day.

he then goes to another school with that major, graduates with honors, and he'd say it worked out. as the one poster said.

but, yeah, you can blow it in interviews. it's not just a points exercise.

no, it's not so obvious. in my current red state, anyone who graduates with x% rank gets in the flagships.

if you think about it, that does benefit students at weaker schools in poorer areas.

but it is also an earned distinction.

one thing i hear can be a problem is if "y" number kids from your school get in some selective schools, they may move on to accepting from someplace else. they don't want 20 kids from Gunn HS at cornell. and so his weaker class schedule and rank at a school with a nearly 1400 SAT average may mean they take others first, and get sick of kids from there when his application comes up.

Yes, your article is also evidence that over optimizing for any standard test (be it the SAT or AIME or IMO) results in better outcomes on that test than having general knowledge and actual understanding, and that the AI can do well on these despite having extremely limited general knowledge or actual understanding.

There are thousands and thousands of practice problems for these exams online, and some of them (like AIME) have explicitly become common benchmarks that the various AI companies optimize for.

But once you step outside of those and ask questions that are common, but not outrageously common, the AI completely falls apart and usually produces answers that don't even parse properly. I'd take this as pretty good evidence that its the exam specific practice and not the general intelligence or actual subject understanding that produce the good test results.

For example, here are a few problems that would be easy for the good but usually not amazing first and second year math/CS undergrads I teach. The first two are minor variants of problems I've explicitly given as homeworks and have solutions that can be found with a quick google search and appear on places like wikipedia. The third is completely trivial. I just tested them with deepseek r1 (which does way better than gpt-4 on math benchmarks like AIME) and it gave completely nonsensical responses to all of them.

1: Suppose that we partition the natural numbers into three disjoint sets. Show that there are two distinct numbers x,y such that {x,y,x+y} belongs to one set of the partition. You should answer this question from first principles.

2: Let x_1,x_2,... be an infinite sequence of natural numbers and let n be a large enough natural number depending on the sequence. Show that if A is a subset of {1,...,n} with |A|>.01n, then there is an element a of A and i<j such that a+x_i+...+x_j is in A. Your proof should be from first principles.

3: Let T be a tree (in the sense of graph theory) with minimum degree 3. Must T have more leaves than other vertices?

It's also pretty easy to demonstrate that it's lack of ability here is really coming from not being able to reason as opposed to not knowing the correct math. For example, in the reasoning steps on the first problem deepseek recognized the problem from its training data enough to see that it's related to Schur's theorem (which is the generalization of problem 1 to any number of pieces instead of just 3) but ultimately decided that the results were unrelated before going on to do some random nonsense. But after I went back and pointed out the relationship with Schur it could produce the wikipedia proof. Similarly, if you slightly rephrase question 2 then it recognizes it as Poincare's theorem and can copy down the standard argument. The third problem on the other hand is too easy to really show up in any homework databanks or note sets, and so deepseek couldn't make much progress and in fact gave worse and worse answers as I pointed out mistakes and asked it to try again.

As an example - Lowell (1350 average SAT) vs Mission High (1020). And before I hear woke whining about how they’re test optional (which also was done in order to discriminate, mainly against Asians but also whites), it’s pretty obvious that students with 1350-1370 SAT averages are better than those with 1020.

Yet Mission has a 43% acceptance rate to Berkeley while Lowell has a 14% one? Lunacy

Based on...?

put differently, if the focus is more on securing the top kids from each area, via rank, then his theorized 9% -- and he's using junior numbers and guessing -- says there might be not just 9% at his school better but also that he'd be seen as behind 9% at every other school as well. the state would justify it as we believe talent is evenly distributed but inefficiently harvested. since they are using rank and not race, it's not racial. they don't care what color ranks how.

they would then see it as his fault he chose a less challenging program that weakened his rank.

Why should it be the top kids from each area? Take the top kids overall - if one school has an average SAT of basically 1400 compared to just over 1000, the former will have more talent. If there's basically no talent in inner city LA, SF, Oakland, or extremely rural areas, it is what it is.

So how is it "blaming minorities"?

He's suing over being discriminated against based on being a non-favored race.

here's the deal. in my current state the first thing is are you x% class rank. if you meet that, those are auto admits. that is apparently thousands of the frosh alone. those slots are gone before your 9% rank kid gets holistic review or even a chance to wave around his big SAT score.

and then at elite schools, particularly those admitting by individual programs, he might holistic review but neither his GPA nor his SAT are that amazing among those applicants. for cornell or MIT, for example.

and i still think he'd get dragged for not so many honors classes.

to me there is also not much gradation in his list of schools. there's no pomona/claremont/mudd. etc. it's the best CS in the nation, and then UC. he's kind of missing a mdidle layer where one can get admitted in between those.

The existence of nepotism that benefits a small number of wealthy white men does not justify discriminating against all white men on the basis of their skin color. Your anecdotal accounts of Trump's cabinet members' qualifications have no relation to some poor white kid from West Virginia or Mississippi and his qualifications.

Just STOP discriminating against children on the basis of their race. This is not hard. You are not resisting Trump. You are not proving some point about your dislike for the defense secretary. You are not righting the wrongs of slavery or jim crow. All you are doing is taking away opportunities from children who did nothing but be born a skin color you deem makes them not worthy of college admission.

just go ahead and say you believe in Bell Curve for us.

what the state is thinking is they can't all possibly be stupid, but maybe the school is lower quality or a mess. at which point some kid with a 1400 with zero test prep from an inner city school with 900 average is probably a genius, compared to some kid with a 14xx or 15xx from some prep school or amazing suburban high that churns out 1400 as its average.

on this argument i think the running training argument is apt. that a coach might want the kid with a 4-flat on 30 miles a week over the one doing college level miles to get the same time.

at least some of what we're wandering into is whether you're really seeking true merit or instead wanting to reward privilege. because if you both end affirmative action AND end financial aid, then say "we're picking on merit now," in reality the poor kids have a harder time getting in or affording it. which benefits money.

my parents were broke and never finished college. i made grades and tests. i could use financial aid to go places where i didn't have the money then.

the more you carve back any attention to affordability or the fact that in an area a private might average 1400, a suburban public 1100, and a city school 900, you're just rewarding privilege. and you're pretending people are just smarter at private school, when they often live next door to the kids in suburban publics who will average 200 points lower. which is my response to is it kids or is it schools. if attending a private buys your neighbor 200 points, that's not IQ anymore. ditto whatever 10s of points i gave up with zero test prep.

if you want to play this game you can resource up inner city schools like an elite private, or at least a normal suburban. the one i had to go to to take the SAT looked like it had been the same since 1930.

If the parent (well, the single mother) and the culture doesn't value education, there's nothing you can do. Baltimore spends the 4th most per student per year. 40% of its high schools have zero students proficient in math.