Looks like its best not to enroll in classes with more than 5 students.

Cynicism aside, seems like a good step.

Actual grade meaning would require criterion-referenced assessment: define what mastery looks like, grade against that standard, and let the distribution fall where it may across years and cohorts. That's hard and unsexy, so instead we get an administrative quota that launders the appearance of rigor while the underlying problem, that Harvard's admissions process selects heavily for wealth and legacy, goes untouched.

Harvard A's will now tell employers that a learner beat ~80% of a nepotism-filtered, endowment-curated cohort in a single semester. That's a relative rank, not a measure of exceptional work.

It's a bit alien to me. Where I went to school, you used get scores from 1(lowest)- to 10 (highest) where 6 is "Acceptable". You could curve the questions, but not the students. So theoretically the whole class could all score 1s, or all could score 10s. This makes more sense to me, if everyone works hard, they should all succeed, and if they're all lazy they should fail.

You couldn't arbitrarily decide that exactly 20% gets -say- an 8. I've always wondered what the steelman is for curve grading.

What if less that 20% of the class do "exceptional" work? What if more do?

Those pushing this either haven't thought it through, or simply want to be seen to be doing something to address grade inflation, and this is something (just not something useful).