Nilay Patel

One thing I'm really curious about in this dynamic is sort of the value of data.

And I promise this will get to AI in a roundabout way.

I'm previewing that we're going to arrive there in this roundabout way.

I do my best to bring up AI sort of like late middle.

Because people's actual businesses are interesting to me.

And the idea that we're all just going to yell about AI in some mythical way, it comes up.

I promise it comes up.

But I like to understand the actual business first.

And the reason I'm saying I'm going to get to AI with this next question is

is because real estate is such a local, data-poor business.

You're just in whatever market you're in, and those are the houses, and every transaction is different in a way that maybe is not even remotely comparable to the next-door transaction that's going to take place.

Or maybe it is, but that's why everybody's mad about this estimate.

There's just something about this market, this business, where the presence of additional data might not actually be helpful.

especially if you have to collect that data on the national scale and then tell some local broker or some local agent or some local buyer or some local seller, hey, we know everything.

We can definitely close because the data says this national broker that we've partnered with, this national mortgage company we've partnered with is going to close.

How do you reconcile that?

Because this feels like in order to make the AI tools work, you definitely need to feed them lots of data, but the presence of the data might not actually matter at the local level.

Yeah, I'm just thinking about when you say we're going to build a bunch of software for the agents, or we're going to originate a bunch of mortgages, the value of the agent is perfect knowledge of their locality in many ways, or being able to stage a house.

It's in the atoms.

It's not in the bits.