The Man Who Built the Tool That Took Over the Internet
In November of last year, Peter Steinberger sat down with some AI coding tools and started tinkering. He wanted to organize his digital life. What he built instead was OpenClaw, a tool that can connect to any AI model and take real-world actions through messaging apps as if it were a personal assistant you could text.
Within months, OpenClaw became the fastest-growing open-source project in history. Jensen Huang of NVIDIA called it "the next ChatGPT." Sam Altman hired Steinberger to "drive the next generation of personal agents" at OpenAI. Cities in China started offering grants worth over $700,000 to businesses building on the platform.
Today, Steinberger stood on stage at ClawCon in Tokyo, a conference for OpenClaw enthusiasts where many attendees showed up dressed as lobsters, and told AFP something that anyone running a business should hear: this year, things change.
"You'll see much more of that this year because this is the year of agents," Steinberger told the news agency. He was talking about AI becoming a genuinely useful personal assistant for ordinary people, not just developers.
He added: "There are still some things we need to do to make it better" - but demand is already "ramping up," with more developers now making the future happen.
What "The Year of Agents" Actually Means for a Business Owner
The phrase "year of agents" sounds like the kind of buzzword-heavy prediction tech people make every January. But there is something different happening here. This is the creator of the tool, speaking at a global conference in Tokyo after watching his project go from a weekend experiment to an international phenomenon in under five months.
Here is what it means in plain terms.
An AI agent is different from a chatbot. A chatbot answers questions. An agent takes actions. It can check you in for a flight, draft and send an email, update a spreadsheet, pull data from a website, and file a report - all from a single typed instruction, without you clicking through a dozen screens.
OpenClaw is the tool that made this accessible to regular people. You install it on a computer, connect it to an AI model like Claude or GPT-4, and then interact with it through a messaging app or terminal as if you were texting a very capable assistant.
When Steinberger says 2026 is the year this goes mainstream, he is saying that the gap between "tech demo" and "something your employee can actually use" is closing fast. The demos at ClawCon Tokyo showed real workflows on stage, not curated prototypes. Hundreds of attendees came to learn how to install and run their own agents.
That is a signal worth paying attention to. Not because you need to run out and install OpenClaw tomorrow, but because the window for figuring out what role this technology plays in your business is shorter than most people realize.
NVIDIA Stepped Up the Same Day
On the same morning that Steinberger was speaking in Tokyo, NVIDIA's security team published new details about NemoClaw, their enterprise security layer for OpenClaw agents.
NemoClaw, launched at GTC 2026 in mid-March, is what NVIDIA built specifically to make OpenClaw safe for business use. It wraps an OpenClaw agent in four layers of security controls that prevent it from accessing files it should not touch, making network calls it has not been approved to make, or running system commands outside of a defined sandbox.
The new details from NVIDIA specifically addressed the agentic execution model - how NemoClaw handles an agent that is running autonomously for hours or days rather than responding to a single prompt. The security team described how the guardrails stay active during long-running tasks, not just at startup.
This matters because one of the most common concerns business owners and IT teams raise about AI agents is the question of control. If an agent is running in the background taking actions on your behalf, how do you know it stays in bounds? NemoClaw's answer is: hardware-level enforcement at every step, with logs you can audit.
You can read more about how NemoClaw's security layers work in our NemoClaw security guide.
Why OpenClaw Could Not Have Come From a Big Company
Steinberger made a remark in Tokyo that cuts to the heart of why this tool became what it is. He told AFP: "What you have to know about OpenClaw is, like, it couldn't have come from those big companies. Those companies would have worried too much about what could go wrong instead of just - I wanted to just show people I've been into the future."
This is not a knock on the big companies. It is a description of how innovation actually happens. The large labs - OpenAI, Google, Anthropic, NVIDIA - move carefully because they have enormous user bases, regulatory scrutiny, and reputations to protect. They build cautiously, test thoroughly, and launch responsibly.
Steinberger built something in a weekend that showed people what was possible. That is a different job. And because it was open-source, the community picked it up and started building things he never anticipated.
Now the two worlds are converging. NVIDIA built NemoClaw to wrap Steinberger's tool in enterprise-grade security. OpenAI hired Steinberger to help them build the next generation of personal agents. The wild experiment is growing up, with the resources of some of the most valuable companies in the world behind it.
For business owners, this convergence is the story. The experimental phase is ending. The "should we pay attention to this?" question is being answered by hundreds of attendees dressed as lobsters in Tokyo, by NVIDIA's security engineering team, and by the creator of the tool himself.
The ClawCon Tokyo Moment
ClawCon is not a corporate conference. It is a gathering of people who are genuinely excited about what OpenClaw can do. There were demos on stage. There were experts helping attendees install their own agents. There were hundreds of people who flew to Tokyo specifically to learn about and celebrate an open-source project.
That kind of grassroots energy is rare. And it is the same energy that powered the early adoption of tools like Slack, GitHub, and Notion - tools that started as developer favorites and became company-wide standards within a few years.
Steinberger described the participants as developers "making the future happen." He noted that similar scenes are unfolding across China, where users have been particularly fast to embrace OpenClaw's ability to organize emails, help with coding, and handle a wide range of automated tasks.
We covered the China phenomenon in detail in our earlier piece: OpenClaw Goes Global: The China Phenomenon Every Business Owner Should Know. The short version: Chinese cities are offering six-figure grants to businesses building on OpenClaw. That is not a trend you ignore.
What Business Owners Should Actually Do Right Now
This is not a "drop everything" moment. But it is a "get informed and start forming a view" moment. Here is a practical breakdown:
If you are curious but not ready to act: Read our plain-English explainer on OpenClaw and our use cases guide. These will give you enough context to have an informed conversation with your IT team or your software vendors when the topic comes up - and it will come up.
If your IT team has been blocking OpenClaw for security reasons: Point them toward NemoClaw for Business. The security blockers that kept OpenClaw out of enterprise settings - unrestricted file access, uncontrolled network calls, no audit logs - are the exact problems NemoClaw was built to solve. The conversation is changing.
If you are actively evaluating AI tools for your business: OpenClaw is worth putting on the list. It is not a SaaS subscription; it is an open-source platform that runs on your hardware or your cloud environment. That means no vendor lock-in and full control over your data. The tradeoff is that you need someone technical to set it up and maintain it. Our install guide covers the basics.
If you are a decision-maker with AI skepticism: The question is no longer whether AI agents will be part of how businesses operate. The question is when, and whether you want to be ahead of that curve or catching up to it. Steinberger did not build OpenClaw for enterprise customers. He built it for himself. The fact that it resonated this broadly is the most honest signal possible about where things are heading.
A Note on the Security Story
One theme that runs through both today's stories - Steinberger in Tokyo and NVIDIA's NemoClaw details - is the tension between speed and safety.
Steinberger was explicit about this. He said big companies would have "worried too much about what could go wrong." And he is right that the product never would have launched if it had to clear every enterprise security review first. But the concerns are real. AI agents that can access files, send emails, and make API calls are genuinely risky if they are not properly controlled.
NVIDIA's NemoClaw announcement today was a direct response to that concern. By publishing specific details about how the guardrails work during long-running agentic tasks, NVIDIA is making a case to IT security teams that they have engineered answers to the specific failure modes that worry them.
This is the mature conversation. Not "AI agents are too dangerous" and not "security concerns are overblown." The mature conversation is: here are the specific risks, here are the specific controls, here is how you can verify they are working. That conversation is happening at an enterprise level now in a way it was not six months ago.
If you want a plain-English breakdown of how NemoClaw's security architecture actually works, our security explainer walks through each layer without the technical jargon.
The Bottom Line
Peter Steinberger built OpenClaw in a weekend because he wanted to organize his email. Today he is at a conference in Tokyo where hundreds of people dressed as lobsters came to celebrate what that weekend project became. NVIDIA's enterprise security team published details on the guardrails they built specifically to make it safe for companies like yours.
He told AFP that 2026 is the year AI agents stop being a demo and start being a daily reality. Given where OpenClaw was in November and where it is today, his track record on calling these things early is pretty good.
You do not need to act this week. But you should probably have a plan by next quarter. The technology is getting safer, the tooling is maturing, and the grassroots momentum that started in a living room in Austria is now being amplified by the largest chip company in the world.
That combination does not stay "emerging" for long.