Andy Halliday

Yeah, but this wasn't even the people who had been incorrect being told by the AI.

This is just a Reddit post that was wrong.

And the user being told by the AI after some conversation about that post that the post was right when it was objectively wrong.

An example would be useful in this case of the 2,000 Reddit posts that were used in the study.

Right.

But they had to first go about finding 2,000 posts where in Reddit the comments basically successfully disputed and refuted posts.

the position of the original host, right?

So now you present that same Reddit post to users and give them the assistance of an AI to evaluate the validity or the invalidity of that particular post.

And half the time,

The models, after some conversation with the user who had a particular bent, you know, ended up agreeing that the post was correct.

Anyway, enough of that one.

Carl, give us some real AI news here.

It's interesting that they've chosen OpenClaw as the centerpiece of what that is, as an initiative for, let's call it Chinese manufacturing and operational productivity for its entire economic system.

And, you know, Open Call has truly created a platform which will allow, you know, inexpensively allow corporations and or startups to provide solutions.

And I know of examples, I'm sure, Carl, you know of examples where that's being done today.

And it's being now encouraged by China that, you know, here, let's all gather in this kind of,

incubator space in one of the tech cities and let's all use OpenClaw to provide solutions that solve this in farming and solve this in logistics and solve this and then operate on a continuous basis to optimize.

that makes a lot of sense and it should be done in corporations as well.

Sure.

So, apparently, Claude Code's source code has been leaked.