Matthew Prince

If you have fewer users, you pay less.

You also want a bunch of sellers that are involved.

You want the local newspapers.

You don't want massive conglomerates to exist.

And you want that to happen between those various things.

So it could be that it's sort of a price per transaction.

The other model that it could be is much more like what you described, you know, with the music licensing services, with something like a Spotify, where effectively they collect a pool of capital.

And then they distribute it out in a way which is fair on both sides.

And ideally you'll have multiple different players who will say, Hey, here's how we distribute it that we think is fair.

Here's another company that says, here's how we distribute that you think it's fair.

And each of them, you know, tries to negotiate with the big AI companies to get the best possible deal for their, you know, members or for their, for their end customers.

But step one has to be you have to have scarcity.

A quick note.

Or you let every buyer buy.

define their own rating algorithm in the bond market it turns out that yes you've got fitch and moody's but you actually have the banks that create their own models on what they think the default risks are and and those sorts of things at some level like in this case like how we're thinking about it at cloudflare is we're building what is effectively like a pluggable framework

where the different AI companies can say, here is my algorithm.

And you think that's going to be like a bunch of work, but they've actually kind of already built it.

They think of it as a pruning algorithm because the way these things work is they're building giant trees.

But one of the most important things that you have to do in order to make these trees not be infinitely large is prune off the information that's already there.

A way of thinking of that is back to the analogy of Swiss cheese.