I think it would be ok, if I just had to make sure the entire span was covered by smooth incremental steps. Note that these ‘checkpoint selves’ wouldn’t necessarily be active all the time, just unpaused for the sake of, for example, translation tasks.
So original self [0] → [1] → [2] → … ->[999] → [1000]. Each one only needs to be understood by the previous, and to understand the next.
I think it would be ok, if I just had to make sure the entire span was covered by smooth incremental steps. Note that these ‘checkpoint selves’ wouldn’t necessarily be active all the time, just unpaused for the sake of, for example, translation tasks.
So original self [0] → [1] → [2] → … ->[999] → [1000]. Each one only needs to be understood by the previous, and to understand the next.