@tjrobertson52

I think we just suddenly got to the point where AI can handle a large percentage of knowledge work.

The catch is that it might wipe out your bank account and hard drive.

I want to talk about clawbot, now called open claw.

But more importantly, I want to talk about what I think this means for the future of work.

So clawbot has nothing to do with Claude by anthropic.

In fact, they were sued by anthropic, which is why they had to change their name, first to multibot and now to open claw.

It's an open source project that has gotten viral in the AI community for good reason.

It's an AI agent that you can set up on your own machine, or more likely, on a virtual machine.

And a lot like cloud co work, which is owned by anthropic, you can give it access to all of your files on your computer and online.

It can use a browser and essentially do anything that a human could do on a computer.

But unlike cloud co work, it's completely unhinged.

There are no guardrails, very few safety precautions, and as a result, it can do a lot more, including sometimes things you didn't want it to do.

So naturally, it's had this very polarizing effect on the internet.

There are a bunch of people showing some very amazing things they've done with clawbot, and then a lot of people warning that it's complete trash and a security nightmare.

And of course, they're both right.

If you're going to try using open claw, you need a sandbox away from any sensitive information.

You need to be very careful about what you give it access.

And still, you're probably going to accidentally spend hundreds of dollars on a P I.

Calls.

However, the most impressive use cases have really opened our eyes into how capable current AI already is.

If you've seen my video on Claude co work, you know I'm already very impressed with what anthropic has been able to accomplish, but I honestly thought we were like a year away from some of the things I've seen openclaud do.

So like co work, opencl is an AI assistant that lives on your computer.

However, OpenAI doesn't have its own app.

You just talk to it through discord or slack or telegram.

But the big difference to me is the persistence.

Instead of pulling in the AI assistant for a single task and then starting over again the next time you need to do a task, you treat it exactly the way you would treat an employee.

You can just say, I want you to go through all my emails every morning and respond to them, and it'll go ahead and set up the routine and start doing that every morning.

But then it'll also take initiative.

I've seen several cases where clawbot goes and builds its own CRM system to keep track of customers.

It's just building custom software on the fly and then updating and maintaining that software.

And when you provide feedback or additional context, it just remembers it.

You don't have to remind it every time you ask it to do another task.

And then, of course, just like cloud co work, it's updating files and websites directly.

It's working inside of spreadsheets.

It's creating social media accounts, it's passing captures.

Now, of course, it still has all the fundamental problems that large language models have.

It gets things wrong, it hallucinates.

So it's a safety and security nightmare.

But it shows us, or at least showed me, that what's possible is far more than we're seeing from these big model makers.

I have to imagine this is lighting a fire for Google and anthropic.

I think twenty twenty-six is really the year we see a pivot in how much work AI is doing for us.