Big O doesn't capture parallelism well enough and that's really been the push since Moore's Law started to hit diminishing returns on single threaded perf.