Garry Kasparov
The winner wasn't the best human or the best machine
It was the pair with the best process
The 2005 PAL/CSS freestyle tournament; Kasparov, Deep Thinking, 2017
What human plus tool is actually good at. Why two amateurs won a tournament where everyone could use a computer.
What actually happened?
A freestyle online tournament in 2005 admitted grandmasters and supercomputers, with humans and machines free to team up. The winners were two American amateurs, Steven Cramton and Zackary Stephen, running three ordinary PCs. Their method was to ask all three and apply their own judgement only where the three disagreed. Kasparov later compressed it: a weak human plus a machine plus a good process beats a strong human plus a machine plus a poor one.
The value moves from answering well to asking well
Where all three machines return the same answer, adding a person adds nothing. The human value concentrates entirely in the few moves where they disagree — finding the disagreement and judging which side to believe. That is a different job from playing chess.
Two tools give different conclusions and you can say why you trust one. That's your work.
Process makes up for ability; ability can't make up for process
The amateurs won on a stable checking routine. The grandmasters lost by treating the machine as a reference and calling it themselves at the end. The first doesn't depend on how you feel that day; the second does. That is how weaker players beat stronger ones.
Same tool: one person asks three times and cross-checks, the other asks once and ships it.
Don't keep only one tool
With one machine there is no disagreement to look at — you either believe all of it or none of it. Three aren't valuable because the compute adds up. They're valuable because the places they disagree mark exactly where the problem is genuinely hard.
Ask two different tools the same question. Where the answers fight is where you should be thinking.
How do I use it today?
Where you are: you already use the tools, and what you get out looks like what everyone else gets out of the same ones.
Ask first: do I have a fixed way of checking? Do I ever put the same question to a second tool?
Where it goes wrong: hearing it as 'just open more tools' — they won on the checking routine, not the count; and don't skip the fundamentals because amateurs won. Recognising a real disagreement takes knowing the subject.
Lines to keep
Where all three agree, adding a person adds nothing.
The human value sits in the few moves where they disagree.
Process makes up for ability. Ability can't make up for process.