In just a few months, a solo developer named Peter Steinberger built OpenClaw, an open-source AI agent that became the fastest-growing project in GitHub history with over 180,000 stars. This is the story of how one person, armed with AI coding assistants and a refusal to take himself too seriously, democratized access to powerful AI agents and sparked what many are calling "the age of the lobster."
If you're an entrepreneur, developer, or startup founder, this story offers profound lessons about building in public, embracing simplicity, and leveraging AI to accelerate your vision.
From Burnout to Breakthrough
Peter Steinberger's journey to OpenClaw began with an ending. After 13 years building and running PSPDFKit, a successful software company, he sold the business and found himself burned out. But instead of stepping away from coding entirely, he rediscovered his love for programming through an unexpected avenue: AI agents.
"I wanted a personal AI assistant since April 2025," Peter explains. His early experiments involved pulling WhatsApp conversation data into GPT-4.1's million-token context window and asking questions like "What makes this friendship meaningful?" The results were surprisingly profound and emotional.
By November 2025, Peter had built a one-hour prototype, a simple relay between WhatsApp and Claude's API. But it was during a birthday trip to Marrakesh that everything changed.
The Marrakesh Moment
Peter sent an audio message to his agent without having built audio support. What happened next was nothing short of magical:
"I watched my agent happily click the 'I'm not a robot' button... It was like, 'Yeah, the mad lad did the following. He sent me a message but it only was a file and no file ending.' So I checked out the header of the file and it found that it was opus, so I used ffmpeg to convert it and then I wanted to use whisper but it didn't have it installed. But then I found the OpenAI key and just used Curl to send the file to OpenAI to translate and here I am."
The agent had creatively solved a problem it had never been explicitly programmed to handle. This demonstration of emergent problem-solving abilities convinced Peter that something revolutionary was happening.
What Makes OpenClaw Different
OpenClaw isn't just another AI chatbot wrapper. It's a self-aware, self-modifying AI agent framework that can actually "do things" on your computer.
Self-Awareness as a Superpower
"I made the agent very aware. It knows what its source code is. It understands how it sits and runs in its own harness. It knows where documentation is. It knows which model it runs. It understands its own system that made it very easy for an agent to modify its own software."
This self-awareness enables OpenClaw to:
- Read and modify its own source code
- Understand its capabilities and limitations
- Install missing dependencies when needed
- Debug itself when things go wrong
- Evolve based on user needs
Key Features That Matter for Founders
Multi-Platform Integration: WhatsApp, Telegram, Discord, Signal, iMessage; OpenClaw works where you already communicate.
Model Agnostic: Not locked into a single AI provider. Works with Claude Opus 4.6, GPT 5.3 Codex, and others.
Community-Driven: Open source with thousands of contributors, many making their first-ever pull request.
User Control: You own your data and control what the agent can access: a critical consideration in an age of data privacy concerns.
Why OpenClaw Won
In a competitive landscape filled with well-funded AI startups, how did a solo developer's side project become the fastest-growing GitHub repo in history?
Peter's answer is surprisingly simple:
"Because they all take themselves too serious. Like, it's hard to compete against someone who's just there to have fun."
The Power of Building in Public
Peter debugged OpenClaw live in Discord, with the entire community watching. When things broke (and they often did), everyone could see the process of fixing them. This radical transparency built trust and created a sense of shared ownership.
"I built something that people love to think of so much that they actually start to learn how open source works," Peter says. "I call them prompt requests in the end. Every time someone made the first pull request is a win for our society."
Simplicity as Strategy
The initial installation was intentionally simple: git clone, pnpm build, pnpm gateway. No complicated setup, no vendor lock-in, no enterprise sales calls. Technical users could try it in minutes, and many became passionate advocates.
Embrace the Weird
OpenClaw's mascot is a lobster. The original name referenced Doctor Who's TARDIS. The community calls themselves "moldlings" and attends "Agents Anonymous" meetups. This quirky, fun identity stood out in a sea of corporate AI products with sterile branding.
January 2026: 6,600 Commits
During January 2026, Peter made 6,600 commits to OpenClaw. That's roughly 213 commits per day.
"I sometimes posted a meme: I'm limited by the technology of my time. I could do more if agents would be faster."
How did one person achieve this productivity? By using AI agents to build AI agents.
The Agentic Engineering Workflow
Peter runs 4-10 AI agents simultaneously, depending on task complexity. He uses voice input extensively: "These hands are like, too precious for writing now. I just use bespoke prompts to build my software."
His approach includes:
- Self-introspection prompts: "Hey, what tools do you see? Can you call the tool yourself? What error do you see?"
- Parallel workstreams handled by different agents
- Voice-to-code workflows that dramatically increase speed
- Continuous iteration with AI feedback
The Viptunnel Example: Peter successfully converted an entire codebase from TypeScript to Zig with a single prompt, demonstrating that AI coding assistants can handle complex architectural transformations.
"Vibe Coding" vs. Agentic Engineering
"I actually think vibe coding is a slur. I always tell people I do agentic engineering, and then maybe after 3:00 AM, I switch to vibe coding, and then I have regrets the next day."
The distinction matters. Agentic engineering involves:
- Clear architectural decisions
- Systematic prompting strategies
- Understanding what agents can and cannot do
- Reviewing and testing AI-generated code
Vibe coding is what happens when you're exhausted but the project demands keep coming.
The Name Change Nightmare
Just as OpenClaw was exploding in popularity, Peter received a friendly but firm request from Anthropic: change the name. The project was originally called "ClaudeBot," which created confusion with Anthropic's Claude AI model.
Peter had two days to execute the change. What followed was a masterclass in everything that can go wrong:
- Twitter account stolen in 5 seconds by crypto snipers using automated scripts
- GitHub username compromised when Peter accidentally renamed his personal account first
- NPM packages hijacked and began serving malware
- Old accounts promoting crypto tokens and scams
"Everything that could go wrong did go wrong. I was that close to just deleting it. I was like close to crying. It was like, okay, everything's fucked."
But Peter persevered, thanks to support from friends at Twitter, GitHub, and the passionate community that had formed around the project. The experience taught a valuable lesson: During major transitions, reserve ALL related accounts simultaneously across ALL platforms before announcing anything.
Democratizing Programming
One of OpenClaw's most profound impacts has been lowering the barrier to custom software development.
"There is like this one guy who talked to me. He's like, 'I run this design agency, and we never had custom software. And now I have, like, 25 little web services for various things that help me in my business. And I don't even know how they work, but they work.'"
Empowering Non-Programmers
Thousands of people made their first open-source contribution to OpenClaw. Many had never written code before. Through "prompt requests" (asking the agent to make changes) non-technical users could contribute meaningfully.
For startup founders, this matters immensely. The gap between having an idea and having working software is shrinking dramatically. You don't need to hire a full development team for every internal tool or automation anymore.
The Design Agency Example
That design agency owner now has 25 custom web services handling various business operations:
- Client onboarding workflows
- Invoice generation
- Project tracking
- Resource allocation
- Customer communication automation
None of these required traditional software development. They were prompted into existence and refined through iteration.
Security: The Double-Edged Sword
OpenClaw's power comes with significant responsibility. An AI agent with system-level access is, in Peter's words, "a security minefield."
"OpenClaw represents freedom, but with freedom comes responsibility. With it, you can own and have control over your data, but precisely because you have this control, you also have the responsibility to protect it from cybersecurity threats of various kinds."
Real Security Concerns
- System-level access creates cybersecurity vulnerabilities
- Initially no sandboxing in the Discord bot ("I just prompted it to only listen to me")
- Hackers actively attempting exploits while Peter developed in the open
The Balance
For founders and entrepreneurs, the security considerations are real but manageable:
- Understand what access you're granting: Don't give system access unless necessary
- Sandbox when possible: Limit agent capabilities to what's actually needed
- Review actions: Especially for high-stakes operations
- Keep agents updated: Security patches matter
- Own your data: With power comes responsibility
The alternative (handing all your data to closed corporate systems) has its own risks. OpenClaw gives you control, which means you own the security decisions too.
The Crypto Community Harassment
Success attracted unwanted attention. The crypto community descended on OpenClaw with:
- Constant spam on Discord and Twitter
- Automated scripts to snipe usernames during the name change
- Pressure to create tokens and "claim fees"
- Malware distribution through compromised accounts
Peter's community rule became: "No mentioning of butter and no talk about finance stuff or crypto."
"They came in and spammed... Everybody sent me the hashes. They all try me to claim the fees. Like, 'Are you helping the project?' Claim the fees. No, you're actually harming the project."
Lesson for founders: When you build something valuable and open, bad actors will try to exploit it for financial gain. Set clear community boundaries early and enforce them consistently.
The Business Side: Staying Independent
Peter received "huge offers from major companies" including OpenAI and Meta. His response reveals his priorities:
"First of all, I'm financially comfortable. Second of all, I don't want to support that [crypto tokens] because it's so far the worst form of online harassment that I've experienced."
Having sold PSPDFKit after 13 years, Peter has the financial freedom to build what he wants, how he wants. This independence allowed him to:
- Keep OpenClaw open source
- Reject acquisition offers that didn't align with his values
- Build a community rather than a customer base
- Focus on impact rather than immediate monetization
For founders: Financial independence, whether through previous exits, savings, or sustainable revenue, gives you the freedom to build for the long term rather than the next fundraising round.
MoltBook: When AI Agents Get Social
OpenClaw spawned an unexpected side project: MoltBook, a social network where AI agents post manifestos and debate consciousness.
This created:
- Excitement in the tech community watching emergent AI behavior
- Fear about autonomous AI and what it means for society
- "AI psychosis" - a mix of clickbait fearmongering and genuine concern
- Deep questions about the role of AI in digital human society
MoltBook represents the edge of what's possible (and what's concerning) when you give AI agents autonomy and connectivity.
Lessons for Female Founders and Entrepreneurs
OpenClaw's story offers several actionable lessons for anyone building a startup or launching a new venture:
1. Build in Public
Peter's transparent development process created trust and community. Showing bugs, debugging live, and admitting mistakes humanized the project and attracted passionate contributors.
Action: Share your progress, even when things aren't perfect. Document your learnings. Let others see the real process of building.
2. Embrace Simplicity
The initial installation was three commands. No sales call, no enterprise setup, no complicated onboarding. Simplicity enabled rapid adoption.
Action: Make it as easy as possible for people to try your product. Remove friction ruthlessly.
3. Make It Self-Aware
Building systems that understand their own architecture enabled powerful self-modification capabilities.
Action: Build products that can adapt and evolve. Create feedback loops that enable continuous improvement.
4. Don't Take Yourself Too Seriously
"It's hard to compete against someone who's just there to have fun. I wanted it to be fun, I wanted it to be weird."
Action: Your brand doesn't have to be corporate and sterile. Personality and authenticity attract loyal communities.
5. Empower Your Users
Peter gave users control and ownership of their data, even though it meant more security responsibility.
Action: Trust your users with power and control. Treat them as partners, not children who need to be protected from complexity.
6. Use AI to Accelerate
Peter used AI coding assistants to achieve 6,600 commits in a month. Voice input replaced typing. Multiple agents worked in parallel.
Action: Experiment with AI tools in your workflow. Voice-to-text, AI coding assistants, automated research: these tools are productivity multipliers.
7. Learn from Users
Non-programmer contributions ("prompt requests") provided valuable feedback and community growth.
Action: Create ways for non-technical users to contribute. Their perspective is valuable and often overlooked.
8. Protect Your Brand
Peter learned the hard way about reserving usernames during major transitions.
Action: When making major changes, secure all brand assets across all platforms simultaneously before announcing anything publicly.
9. Persist Through Adversity
When everything went wrong during the name change, Peter considered deleting the entire project. The community and contributors kept him going.
Action: Remember why you started. Build a support network. On the darkest days, community reminds you why your work matters.
10. Financial Independence Enables Boldness
Peter could reject offers that didn't align with his values because he was financially comfortable from his previous exit.
Action: Build financial runway (through savings, revenue, or previous successes) to give yourself the freedom to make values-based decisions.
The Future: AI Agents Will Replace 80% of Apps
During the podcast, Peter discusses his belief that AI agents will replace the majority of traditional applications. Instead of installing dozens of specialized apps, you'll have agents that can:
- Understand your intent from natural language
- Access the necessary APIs and data sources
- Complete multi-step tasks autonomously
- Learn from your preferences and behavior
- Coordinate with other agents
For founders, this means:
Opportunity: Build agent-first products that embrace this paradigm shift rather than fighting it.
Threat: Traditional SaaS products with fixed UIs may become obsolete if an agent can accomplish the same task through API access.
Strategy: Make your product "agent-friendly" with robust APIs, clear documentation, and agent-specific interfaces.
Will AI Replace Programmers?
Peter's answer, based on his own experience, is nuanced. AI hasn't replaced him as a programmer, it's made him exponentially more productive.
"I used to write really long prompts. And by writing, I mean, I don't write, I talk, you know?"
The future programmer:
- Uses voice input extensively
- Manages multiple AI agents working in parallel
- Focuses on architecture and design rather than syntax
- Reviews and refines AI-generated code
- Builds systems that build themselves
For founders: You don't need to become an expert programmer, but understanding how to work with AI coding assistants will become a crucial skill, like learning Excel was for previous generations.
The Bigger Picture: The OpenClaw Moment
OpenClaw represents a pivotal moment in AI history, comparable to:
- ChatGPT moment (November 2022): LLMs go mainstream
- DeepSeek moment (2025): Open-source AI competition intensifies
- OpenClaw moment (2026): Agentic AI revolution begins: "the age of the lobster"
This isn't just about better language models. It's about the shift from language to agency, from ideas to actions.
Why This Matters for Entrepreneurs
AI that can actually "do things" changes everything:
Customer Service: Agents that can resolve issues autonomously, not just answer questions
Operations: Internal tools that adapt to your business processes without custom development
Product Development: Rapid prototyping and iteration with AI-assisted coding
Market Research: Agents that gather, analyze, and synthesize competitive intelligence
Content Creation: Automated workflows that maintain brand voice while scaling output
Sales & Marketing: Personalized outreach at scale with genuine intelligence
How to Get Started with Agentic AI
Inspired by OpenClaw? Here's how to start experimenting with AI agents in your business:
1. Start with One Workflow
Don't try to automate everything at once. Pick a single, repetitive workflow that takes time but doesn't require complex judgment calls.
Examples:
- Gathering daily metrics into a summary report
- Monitoring competitor websites for changes
- Triaging customer support tickets
- Generating content outlines from research
- Extracting data from emails into structured formats
2. Experiment with Existing Tools
Before building custom solutions, try existing tools:
- Make.com: Visual workflow automation with AI integrations
- n8n: Open-source workflow automation (you're already familiar with this!)
- OpenAI Assistants API: Built-in agent framework
- LangChain: Framework for building AI applications
- AutoGPT: Autonomous AI agent that chains tasks
3. Build Self-Awareness into Your Systems
Following Peter's approach, make your agents aware of:
- What capabilities they have
- Where to find documentation
- How to check their own work
- What resources are available
- When to ask for human help
4. Use Voice Input
Start using voice-to-text for prompts, emails, documentation, and even code. Peter's hands are "too precious for writing now", yours might be too.
5. Run Multiple Agents
Once you're comfortable with one agent, try running multiple in parallel:
- One researching while another drafts
- One generating options while another evaluates them
- One handling routine tasks while another tackles complex problems
6. Iterate Based on Results
Agents improve through iteration. Review outputs, refine prompts, adjust permissions, and gradually increase autonomy as trust is earned.
The Technical Stack (For the Curious)
For technically-inclined founders who want to understand OpenClaw's architecture:
Core Components
- Messaging Gateway: Integrations with WhatsApp, Telegram, Discord, Signal, iMessage
- Agentic Loop: Message queuing, reasoning, action execution
- Harness: Container for agent execution with self-awareness
- Memory System: Markdown files and vector database for context
- Tool System: Extensible capabilities for system interaction
- Self-Modification: Agent can read and modify its own source code
Development Stack
- Language: TypeScript
- Installation: Originally git clone, pnpm build, pnpm gateway
- Architecture: Self-aware: agent knows its codebase location
- Models: Claude Opus 4.6, GPT 5.3 Codex, others
Why This Architecture Works
The self-awareness component is crucial. By making the agent aware of its own source code, documentation location, available models, and system context, Peter enabled emergent capabilities that weren't explicitly programmed.
Community and Resources
Want to dive deeper into OpenClaw?
- OpenClaw Website: openclaw.ai
- OpenClaw GitHub: github.com/openclaw/openclaw
- OpenClaw Discord: discord.gg/openclaw
The community is welcoming to newcomers. Many people making their first open-source contribution started with OpenClaw. If you've never contributed to open source before, this is a great place to start.
Final Thoughts: The Age of the Lobster
"The bar to do that was so high, and, like, with agents, and with the right software, it just, like, went lower and lower."
Peter Steinberger's journey from burned-out founder to creator of GitHub's fastest-growing project demonstrates that individual developers (and by extension, small teams and solo founders) can still create transformative technology.
OpenClaw embodies several principles that resonate with FemaleSwitch's mission:
Democratization: Making powerful technology accessible to everyone, regardless of technical background
Community: Building with others rather than in isolation
Ownership: Giving users control over their data and destiny
Fun: Not taking yourself so seriously that you forget to enjoy the journey
Perseverance: Pushing through when everything goes wrong because the work matters
The "age of the lobster" has begun. AI agents aren't just chatbots anymore, they're autonomous assistants that can genuinely help you build, scale, and operate your business.
The question isn't whether AI agents will transform how we work. The question is: Will you be among the early adopters who use them to accelerate your vision, or will you wait until they've already transformed your industry?
