Hacker News Reader

A "10% better" tool could make no difference, or it could make the work 100% better. The impact isn't linear.

It's likely probabilistically linear... like speeding on a street with random traffic lights.

Right, I should have put a "necessarily" in there.