[scott_s]

That's his whole point. But he goes further, by showing via experiment that a lot of distributed systems are solving problem sizes in benchmarks that can be solved quicker in a single thread.