Don't Tread on GPT

It's been a story that we've dipped in and out of the past few weeks.

But to get you all up to speed, including a recap for people who have been listening and a crash course for those of you who haven't been listening.

A couple months ago, Anthropic announced the launch of Claude Mythos, which was a giant LLM that they had trained.

For the hell of it, I assume.

Yeah.

That's what they do, those nerds.

They just be doing stuff.

No, because, you know, they still operate on this scaling law hypothesis, all these AI labs, where it's like if you throw more compute at these LLM transformer model, they're going to get more and more capable.

And this is the latest and biggest.

And they announced after they trained it, hey, we've trained this massive model, which we're calling Mythos, And it's really good at cybersecurity.

Scary good at cybersecurity to the point that we're scared.

Everyone should be scared.

Yeah.

We're scared for you.

We're scared for you.

They announced, we're not going to release this thing, at least not imminently, but...

We're going to start this thing called Project Glasswing, very, very intimidating name, where we're going to partner with some of our best friends in the AI world, in the tech world, in the cybersecurity world.

So to basically say, get your shit in order using our tooling.

Yeah, because

otherwise all hell will break loose.

All hell

will break loose.

These models are coming in red hot, folks.

Anyway, so they did that.

That was working, but the lure of... putting stuff out in the market ahead of their IPO became too great.