0:51
Eric Siu explains why defining success matters most with AI. He calls it "eval": encoding your judgment so machines and teammates can replicate your standards consistently, whether you're judging headlines, creatives, or proposals.
Eric Siu explains why defining success matters most with AI. He calls it "eval": encoding your judgment so machines and teammates can replicate your standards consistently, whether you're judging headlines, creatives, or proposals.
An eval, it's evaluation, right? Or you can just look at it as what is the definition of success? What is the definition of done? Because if you don't define that definition of done to the machine, how is it going to know? So when you read a draft and you think, well, that headline sucks, that creative sucks, or man, that interview with that person really sucked, or like all these things, right? Like what does a good proposal look like, right? All these things that you do over and over, you have an eval running in your head. And it's an eval, I mean, it's judgment, right? It's judgment that you have and you're trying to pass on this judgment. People are like, oh, humans are for judgment, the machines are for doing. Well, no, you need the machines to judge as well, right? You need the machines to think. But if you— the more context you give it, the more judgment you pass to it, the easier your life is going to be. And so eval is just— it just writes that judgment down so a computer or a teammate can run without you the same way every single time.
