Ryan Sean Adams

Isn't this just another example of something that has the ability to get measured?

And I guess you can measure how well a video matches

reality and what reality looks like?

Isn't this just an example of verification being valuable at first, but then getting quickly automated again by AI?

So even verification is not safe in this model.

Okay, so shrinking frontier.

So that is the idea of the codifier's curse.

It's basically like, you know, this is humanity's last job is what you're saying is verification.

But even that last job is we're all standing on an iceberg and that iceberg is kind of slowly melting away and the surface area is getting smaller and smaller even on the verification front, right?

So the iceberg actually will get bigger in some ways, which is kind of the non-measurable human...

status type games, human subjective preference types games, that's where I guess the economy for humans will expand and the job opportunities for humans will expand.

I'll give you another example.

So David and I are messing around with an OpenClaw agent.

We have a Discord for him.

Who doesn't do this, right?

The prompt is kind of simple.

Hey, create a media company, just because we're just seeing if it can create a media company the way Bankless has created one.

And just getting it to tweet something that doesn't sound like AI slop.

Something coherent.

It's just like, it feels like mission impossible.