It may actually improve mean outcomes, but harm societal outcomes, as the scientific impact of educated individuals may tend to be power law distributed (e.g. the most important breakthroughs come from a small sector of the population with wildly disproportionate impact)
Forget the fact that he's now a 50 year old dude, this kind of stuff started in high school
There are multiple examples of highly citied ML papers coming from essentially people in the middle of undergrad, meaning they essentially learned all their shit in high school. The first protein diffusion paper, the single cell autoencoder paper, diffusion autoencoder paper... these were all from essentially high school prodigies publishing in undergrad or first year of graduate school.
Just because the top X% is guaranteed admission, that does not mean all (or even most) of the school is from the top X%.
Many good cases to have a strict zero sum competition but you won’t see collaboration in these (eg olympiads, competitions). That’s fine for a short term event but for long term learning in a persistent social group it seems good to encourage collaboration
The bad thing about UT's policy is that it encourages well-off students to move to a less-competitive school district (usually rural) in order to improve their chances.
