Exactly - the possible improvements may compound, but they converge logarithmically towards an upper limit absent new insight that establishes a new upper limit.