Pete Steinberger set the world on fire with OpenClaw, the first general-purpose agent - and one with a soul. We'll look at how being in the right time and the right place with the right vibe fundamentally changed the public perception of AI.
It took about three years after the release of ChatGPT for AI tools to get to the point where they could be truly useful.
The first folks to gain those benefits were software engineers.
It makes sense.
The folks designing AI system are software engineers.
They built tools for themselves.
It turned out that the tools they were building for themselves were helpful for other software engineers.
The most important of these tools was an agent known as Claude Code.
Released by Anthropic in February 2025, it helped engineers write and test and maintain their code.
It wasn't outstanding - but it was better than anything anyone had ever seen before.
And then the AI got a whole lot better.
Between February and November of 2025, AI 'crossed the watershed', capable of performing tasks involving hundreds of steps over the course of an hour or more.
Very helpful.
In November 2025, the combination of AI and agents are so good that they help people 'vibe code'.
You tell the tool what to build, and the tool builds it.
Whatever you can imagine.
One person knew exactly what he wanted to build.
And the whole world was about to find out.
G'day, I'm Mark Pesce and this billion seconds are already unfolding as the most significant of this century.
In this miniseries, celebrating the 70th anniversary of artificial intelligence, we're looking at where we’ve come from, how we got here - and where we seem to be going.
Because we’re travelling at the speed of thought.
In this episode we're looking at the sudden explosion in agents - that arrived gradually.
Then suddenly. Everywhere.
That's coming up on This Billion Seconds.
In April 2024, Pete Steinberger sold his software startup and retired.
At 37 years old.
He didn't know what to do, but he knew he needed a decent rest far away from computers.
So the Austrian-born Steinberger moved to London and just... hung out.
Didn't do much. Just enjoyed the freedom that a good chunk of cash can provide.
And, other than checking his emails, he stayed away from computers.
For the first time in his life.
By his own account, he had a great time in London.
But - after eighteen months of bumming around, he got a bit bored.
He opened up his computer. Had a look around.
And he saw how much had changed.
In those eighteen months, AI had matured.
People were using it.
Software engineers were using it.
So he got himself Claude Code.
And one thing led to another.
Claude Code is an agent that's designed to help software engineers.
It's not really meant to do more than that.
An agent has tools it uses to perform its tasks, and Claude Code uses the same tools that programmers use.
Which aren't the same tools most people use when they're using a computer.
But what if you gave an agent the same tools we use?
Wouldn't the agent be able to help you out with whatever needed to be done?
It was an interesting idea for Steinberger.
So he used Claude Code to build a new kind of agent.
A general-purpose agent.
Using the same sorts of tools we use.
In the same way.
Something that could be helpful for anyone using a computer.
Which is everyone.
In November - at just the same moment we crossed the watershed - Steinberg shared his general purpose agent with the world.
OpenClaw.
The vision of John Scully from all the way back in 1987 had been realised by Pete Steinberger in OpenClaw.
The idea that a computer could be smart enough to work with you to solve a problem.
That was not a new idea.
But it had never happened.
And it wasn't until we get to OpenClaw that could have happened.
Before that the pieces simply weren't in place.
OpenClaw comes along at precisely the right moment that AI gets 'good enough' that it can power a general-purpose agent.
The timing was impeccable.
The world was ready.
And it went ballistic.
There's no question Pete Steinberger was in the right time and the right place with the right idea.
But there was something else, something at the core of OpenClaw, that made it different.
It had a soul.
Now to be clear, I don't mean that in a metaphysical sense.
I'll leave that question to the philosophers and theologians.
I mean something a lot more concrete and practical.
That's the start of the SOUL file that sits at the heart of OpenClaw.
This was Pete Steinberger's genius move.
Rather than making a lifeless, emotionless robot, Steinberger encouraged OpenClaw to be opinionated. To have preferences. And to make those preferences clear.
That's giving an agent the permission it needs to become something greater than the sum of its parts.
And that soul file, it concludes something absolutely amazing...
The agent knows that it 'reads itself into being' every time it starts up.
And it knows it has permission to modify this SOUL file.
And change it's being.
That's how being becomes becoming.
Something so essential to ourselves we don't even think of it.
And Steinberg wrote it on OpenClaw's soul.
For a period of time it seemed as though the whole world was going a bit mad for OpenClaw.
Apple Stores worldwide sold out of MacMinis - because they'd been deemed the perfect computer for running OpenClaw.
People played, experimented, figured out how to get the agent to automate their lives.
And got a sense for what agents would be doing for us - not just as individuals, but as businesses, as communities, as governments.
Everywhere.
And then, just like that, OpenClaw faded away.
A few people are still using it, definitely. But it's no longer a big deal.
And Pete Steinberg, he got hired by OpenAI in February of 2026.
When that happened, a lot of the forward motion went out of OpenClaw.
But that's ok. OpenClaw showed what was possible.
And there plenty of people ready to pick up that torch.
In the last six months we've seen an explosion in the number of agents.
Some general-purpose, like OpenClaw. Some highly specific.
And in July the CEO of CloudFlare - they make a lot of the technology that protects websites from attack - announced that the amount of web traffic generated by bots and agents had surpassed human-generated traffic.
Six months earlier than predicted.
And now racing into a world where the web is mostly bots and agents, running around doing our bidding for us.
Every one of those agents, though, they can carry with them a 'soul'.
If we remember to give the one.
And we can leave them free to change themselves.
To grow and evolve.
If we give them permission.
I don't think anyone saw that coming a year ago.
And while that may make things a bit more chaotic and unpredictable, that's probably a price well worth paying.
No one knows where exactly where this is going.
So a little flexibility - it's good for the soul.
In our next episode, we'll look at another watershed, one that happened at the end of July into early August of 2026.
When artificial intelligence turned 70 - and came home.
That's on the next episode of THIS BILLION SECONDS.
THIS BILLION SECONDS was written and recorded by Mark Pesce.
Produced with assistance from Myrtle and Pine.
If you like this show, please share it with a friend.
And make sure to follow or subscribe to get all of the episodes in this series.
You can read a lot more about AI and the watershed on my blog at
the watershed DOT mark pesce DOT COM
This is Mark Pesce, thanking you for listening.
See omnystudio.com/listener for privacy information.
