Intuitively: if non-code is a strict bottleneck, no amount of coding speedup can raise total code output by more than
Phrasing confused me for a minute here. I think you mean assuming pre-speedup timeshare of non-coding is half, and assuming optimal reallocation of time toward research progress (which is maybe the same as assuming no wasted code, if code is not literally free), then under Leontief substitution you won’t observe more than double the code output.
Phrasing confused me for a minute here. I think you mean assuming pre-speedup timeshare of non-coding is half, and assuming optimal reallocation of time toward research progress (which is maybe the same as assuming no wasted code, if code is not literally free), then under Leontief substitution you won’t observe more than double the code output.