There's this idea in computer science wherein the maximum theoretical speedup that can be acquired with an arbitrary number of processors is related to the percentage of the program which can be parallelized. If we have two segments of code that take the same amount of time to execute with one CPU core in which the first segment can't be parallelized at all and the second segment is perfectly parallelizable, we can only run the program twice as fast, no matter how many CPU cores we have.
There's a similar idea in economics. It seems like the most powerful and civilizationally relevant feature controlling the medium to long term change in the price of goods is the extent to which the production of that good can be decoupled from the expenditure of man-hours. Some economic activity isn't "parallelizable" using current technology- we can't practically make that activity much faster without building powerful substitutes for humans, technology which is (for now) mostly out of our reach...
For example, it turns out that moving stuff over land is not easy to decouple from human labor without self-driving cars. There are two methods of overland transportation worth noting here: Cars and Trains.
Due to the nature of our road infrastructure, there's a pretty clear upper bound on how efficient car-based transportation of goods can get. There are legal limits on the allowed speed and size of vehicles, so without self-driving tech, we can't change how many man-hours need to be spent per cubic meter per kilometer.
Train-based transportation has its own problems which limit its ability to dominate overland transportation. Namely, our train transportation network is incomplete in that goods must still be ferried to their final destination by cars, so overland transportation is only some % "parallelizable".
Using this we can predict that overseas transportation would be really efficient compared to overland transportation. I suspect that Nautical miles per cargo container per hour per person can be pushed extremely high using current technology, laws, port and canal infrastructure, etc.
And indeed this is true:
Be warned that overland freight for even short distances can often be almost as much as ocean freight for thousands of miles (I recently paid $2000 for a 20′ container to be shipped from Shanghai to Los Angeles and then $1100 for it to be shipped 30 miles from Los Angeles to San Moreno).
Our ability to recursively reinvest our production is most strongly limited by these required industries which are mostly bottlenecked by the number of humans and how long they're willing to work, neither of which are easy to manipulate and neither brings the sort of powerful prosperity that characterizes modernity.
EDIT: Someone commented that other factors are at play that makes cargo ships efficient. I do not disagree- this was just an example of the sort of weak estimate which could be made using this idea. I am interested in determining how important this effect is in the case of cargo ships, so I will do a short analysis.
Compare to semi-trucks which can carry roughly 40,000 kg of material. We will say these trucks move at 100km/h. The wage of a semi-truck driver in the U.S. is roughly $20, so combing we have, so it costs $1 to move 200,000kg one kilometer- or rather, that is the wage component.
The distance between Shanghai and Los Angeles is ~10,000km, and the limit weight of a 20ft shipping container is about 80,000kg. Assuming wage costs for ship crew are negligible (this should cancel out with previous generous estimates), we have 800,000,000kg * km for this trip for this container. If the same wage were required per kg per km as in the semi-truck case (note much is ignored here for the sake of approximation), this would cost an additional $4000. For comparison, according to this site, the cost of shipping 80,000kg from Shanghai to LA is $14000.