Are you AI tokenmaxxing your way to the top?

Morning Brief Host Julie Hyman and Yahoo Finance Head of News Myles Udland sit down to talk about the Financial Times' coverage of Big Tech employees — such as at Amazon (AMZN) — engaging in the latest trend: AI tokenmaxxing, or excessively using AI to inflate AI usage on internal leaderboards.

Speaking of other token users.

Yes, I I posted about this on LinkedIn yesterday because it's been on my mind, just like how stupid I think.

Oh, so you wrote that yourself though?

I wrote it myself.

You didn't token max your LinkedIn post. Oh, my goodness.

I do use AI for stuff. My focus is not

Not this?

My no, not this. I still value the human word coming from my human brain. Um and I'm going to talk to Joanna Stern, a longtime tech columnist later today, who's just out with the book that she is it makes pains to say was written by humans and illustrated by a human.

Yep.

So, there you go. Um

So it's worth reading.

I think, I mean, I'm been I've been reading it. I've been enjoying it. So, um, I would argue that it's that it's worth reading. Um, but, you know, this but the latest my latest annoyance was sparked by the story that you sent me um about Amazon that like they have the token maxing leaderboard, right? But the people are like automating and using AI for silly stuff just to up their token usage.

Yes.

Well, and the way, so this is a report from the Financial Times yesterday about Amazon's internal token maxing. I think the sources they talk to, view it as a problem. Others may not. It follows the meta-reporting, which was very obvious like there is a leaderboard. The Amazon story from the FT made it seem like various layers of management and personnel may or may not have access to how many tokens you're using.

Right. And they're not using it on to judge your performance.

So well, so this is the point. is like no one inside Amazon believes that. Everyone's like, oh, this is part of my performance. Um and it got me personally thinking about if I was given a mandate to token max, specifically token max, not just incorporate AI into my workflows. How would I do it?

Right.

But I think the answer is I would just take every slack that someone sends to me and then put it into whatever bot uh I'm, you know, being told to use whatever LLM and say, pretend you are me. I and then have a very long description of myself. How would you respond to this and and prompt three other conversations and then it just give me a bunch of text.

Well, and you can write, you could write a code for that.

I don't think I was going to say write a Miles.

I was going to say write a Milesbot.

doing that manually would not get me anywhere, but I could, I could Claude-bot it.

If you Claude-botted it, that would use a lot that would use a lot of tokens. And then you would you would have your own you would never have to answer a slack ever again. You could just get your bot to do it.

Yeah.

Maybe this is a stupid idea.

Oh, you know what feels so good though, as a human, not answering slacks though. I know I'm notorious for it. People really it really annoy people that I'll be like, I'll go quiet on.

I've never never, I've never noticed that, Miles. I don't know what you're talking about. I've never had a hanging slack for for Miles for hours and hours waiting for a reply.

I mean it could be longer. It could be years. It's probably unreplied slacks for years.

Should I double check? And if I create a bot then... Maybe I should build a bot to see how many slacks you have not replied to.

Oh, that's a lot.

A lot. So in other words

Yeah, it's a lot. Email's even more.

like obviously the incentives are screwed up when it comes to this. that if the that if the goal is just token maxing.

Yeah, right. It's the old uh you ship your org chart, you ship your KPIs, whatever it is. Yeah.

It's just I find it tiresome and silly, but I guess this is where we are. We're in this space.

So you so you wrote these uh I'm looking at the the LinkedIn post. You wrote these English words in this left to right, top to bottom all by yourself?

That's correct. All by yourself.

You're like, they're terrible words. I don't know why you're being so mean. I mean, no, no, no. Um, I do well, the last thing on the token maxing though, I guess in the world, then we've talked about this off camera of like, there's so many things that we do that I look at them like, I what would AI get me for this? And specifically in this use case. So whether it's an a LinkedIn post, right, as a per you know, a public figure, or if you're writing a post for the website, if you're using AI to create length, I think you're announcing it doesn't need to be consumed.

Well, and you're like here's my slop. And when you're writing something, it helps you think.

Yes.

So if you're not writing, are you not thinking? That's

Well, look, a lot of people's jobs don't require thinking. So, there you go. Okay.

Yeah, I guess not. Okay. On that note.

I guess so.