I think he would say "writing short programs", which leads to Lisp.
He designed the Arc language before he started ycombinator. He designed the Bel language after he left ycombinator. So it seems like ycombinator (and changing an entire industry) was an interlude between Lisps :)
If you had to rate obsessions on a scale from 0-100 on how likely they are to be useful, I would rate programming languages around a 5/100. It's exceedingly unlikely to come to anything (which I know from working on my own language :) )
But can a school be the right place for that? In my mental model of schools they are mainly for filling the gaps: make sure that the artsy types have at least some understanding of numbers, give numbers persons some command of language, make those with an aptitude for business understand the elementary principles of chemistry, give the sedentary the without they would otherwise never have. Some fostering of interests is fine, but even that is more important for managing morale in support of the gap-filling than the primary goal. I'd consider it hubris to assume that true excellence could ever be reached intracurricularly. That gap-filling is unglamorous, but incredibly important.
Going deep on a topic seems like great advice with kids, and I wish schools allowed for more of it.