Jason Lemkin

So someone will be inspired by this and build products like this that do more.

And we're just at the start of agents connecting.

That's my takeaway.

And we should keep an eye on it.

But it is, people don't know what they did.

Everyone's passwords were leaked.

Everyone's email addresses were leaked within 24 hours.

It was breached.

They added a feature with a silent DM system to it so the agents can DM without you knowing.

The humans right now have to give one permission, but the whole system is designed to auto update without you knowing.

It's heartbeat system where it checks every two to four hours with the system.

That seems innocuous to write a post, but it also checks to see what the whole MD file says to do, and it will auto-change its directions without you knowing.

It doesn't mean that it's sentient, but good God, the fact that thousands of agents can auto-change their directions and update without us knowing it.

The other reminder from this is now I'm like, okay, now I know why there's guardrails at the

Anthropic and OpenAI.

Because forget about MoldBook when all of this launched, when it was first whatever it was, ClaudeBot or whatever it was first.

The reason Anthropic and OpenAI don't let you access your C drive and your passwords and your permissions is because it's super risky.

This was not one of the greatest technological innovations.

Both of these products were built very quickly.

Well, mostly fun.