a boot on your work but they have to say it if they are going to get you to change something.

So this is from the earliest days where my editor was welcome to tell me that something worked or didn't work because I, you know, that was useful to me. But how do you tell somebody who's an artist that they worked all day on something and it's bad and it's not worthy of being published? Have you ever thought about how would you word that? Because you don't want to crush somebody's spirit, right?

So here's the reframe when you don't want to crush somebody's spirit, but you really have to tell them this wasn't good enough. And it went like this. The usual frame, so this would be the wrong way to do it. The old way to do it is that you did this wrong or it's not funny, right? You did it wrong. It's not funny. You did a bad job. That would be the old way.

Now listen to this reframe. This was for my first editor. "Your other work is stronger." Boom. Now is that brilliant or not? "Your other work is stronger." That's all she had to say. If you tell me my other work is stronger, I'm competing against myself and my other work and I'm not competing with her. So she basically takes it down of, you know, you and I have a disagreement about whether this is good and she turns it into a disagreement with myself. "Your other work is stronger." Damn it. I can make this stronger too. She was a genius.

All right. I wonder if there's any science that they didn't need to do because they could have just asked me. Oh, here's some. Oh, by the way, before I forget, Owen Gregorian will have his Spaces event after this is over. All right. So if you want to get a little extra talking about this stuff or maybe some other stuff a few minutes after we're done today, Owen Gregorian will fire up his X Spaces which is the audio-only feature on X. You can just Google him Owen Gregorian and you'll find it easily.

Anyway, so there was a test of AI capabilities. So there's a new paper, meaning a scientific paper, where they tried to test AI's ability to do actual online freelance work. Have you ever heard me talk about how capable AI is to do actual real useful things? Do you understand that from the very beginning I've been probably one of the biggest skeptics of AI being able to actually do something without a human or even helping a human because the LLM model to me looks like an amazing user interface and that's about it. I just don't see how it could do real work. That's a longer conversation.

The paper was to test exactly that to see if we're at the point where the AI could replace a person and be like an AI agent, do actual freelance tasks. And so they gave it a bunch of tasks and they found out that it could do about 3% of the things, but it didn't make anybody faster at anything. Essentially, it found it was worthless.

Now, I know what you're going to say. Scott, Scott, Scott, you don't understand how adoption curves work. First, it's useless, useless, useless, useless, followed by useless, useless, useless. But boy, when it kicks in, whoa, whoa. Soon as it kicks in, it's going to go to the next level. What you need is just more training. What you need is a bigger data center. What you need is another trillion dollars. And then we got something.

Well, d