We can also figure out how to get useful results out of inherently unreliable algorithms – it's definitely possible, I've been finding useful ways to put this stuff to work for a couple of years now

→ View original post on X — @simonw,