Hey there! I'm a Sophomore at Harvard who's helping run HAIST. Congrats on your acceptance to Harvard! :) Some initial thoughts, which others might disagree with:

• Harvard being inflexible seems wrong. My impression is always that all "prerequisites" are merely suggestions; to get around them, simply shoot the professor a very quick email. Taking grad classes is pretty common; I was taking one this semester, and a (Harvard) friend of mine has taken a few grad MIT classes in his Freshman and Sophomore years (both my and my friend's classes were CS classes). My understanding is that MIT has more requirements, but I'm unsure of this. Note that at Harvard, if you don't pass out of your language requirement, you need to take two semesters of a language.
• MIT has somewhat better classes to prepare you to work on technical alignment, I think. On the other hand, MIT classes are more time-consuming.
• I expect Boston/Cambridge to become an alignment hub in the next couple years (and I intend to help make it one). Currently, there's HAIST, MAIA, CBAI (https://www.cbai.ai/), and various independent researchers around the area.
• I suspect that it's worthwhile for you to put some minimal effort to apply to MIT/Stanford/maybe Harvey Mudd. Seems possible and bad if you realize, in a few months, that one of these colleges is much better suited for you than Harvard. But also seems fine to not apply, since I expect Harvard to not be a worse option in general for technical alignment.

If you want to talk to Harvard/MIT/Stanford students working on technical alignment, I'm happy to connect you with some people!

And if you end up deciding on Harvard, I'd absolutely love to introduce you to people in HAIST & to give you advice on anything and everything Harvard-related. :)