Steven Bartlett

Now, I've been making social media content for 15 years.

So I have a bit of a sort of a mental history of like every year us doing this particular data analysis to see the variance between the worst thing we post and the best thing we post.

And this year we did the same.

We ran the same analysis with my data science team led by Austin and my team.

to see if the variance between the worst thing we post and the best thing we post is getting bigger.

Because what that kind of is a proxy of is, is the algorithm caring less and less and less about how many followers I have?

And is it caring more and more and more about whoever posts the most interesting thing today?

Which is if I was building a social media company and my primary objective was to retain people, I wouldn't care if you have two followers and Stephen has 500 followers.

I would just care about who posts the best thing.

The best stuff, yeah.

So we did that analysis and it's an almost perfect graph that looks like this.

And this is the variance increasing.

Yep.

Which actually means that, okay, Stephen's taken off from the sky, but every day I wake up increasingly, it doesn't matter what I did yesterday.

Interest algorithms, we call them.

Why is it fundamentally defensible if we think about first principles or human Maslavian needs?

Theoretically, though, I could ask an AI agent to think about the goal that will benefit me the most and make as many AI agents as it needs to to fulfill that goal.

And then it would start one business and then it would start another business.

Start lots of different businesses.

Start another business, build another website, contact a supplier.