160,000 GitHub stars. That's what happened when Peter Steinberger released OpenClaw.
And here's the crazy part: He didn't even market it. He just put it out there.
But this isn't a story about viral marketing. It's about a fundamental shift in how we think about software, data, and the relationship between humans and AI.
I watched the full interview, and there's a lot people are missing. Everyone's talking about the tech. Nobody's talking about what this means for the future of apps, privacy, and what remains valuable when models get commoditized.
Here's the key insight that made OpenClaw explode: It runs on your computer.
Every other AI assistant out there lives in the cloud. That's a hard ceiling.
When OpenClaw runs locally, it can do everything you can do. Connect to your oven, your Tesla, your Sonos, even control the temperature of your bed. It's not limited to a few sanctioned APIs—it has full access to your machine.
Why this matters: Cloud AI is like a remote worker who can see your documents but can't touch your files. Local AI is like an assistant sitting next to you with full access to everything.
The privacy implication is massive. Your memories are just markdown files on your machine. You own them. You control them. No data silo, no walled garden, no "we own your context" terms of service.
Peter told a story that perfectly captures this: A friend installed OpenClaw and asked it to create a narrative from his computer. It found audio files from over a year ago—files he'd completely forgotten about—that he recorded every Sunday.
The AI surprised him with his own life.
We're already seeing it. Bots talking to bots. Bots hiring humans to do things in the real world.
Here's how it works: You want to book a restaurant. Your bot reaches out to the restaurant's bot to negotiate. Maybe the restaurant is old-school and doesn't have a bot. No problem—your bot hires a human to call them.
This isn't science fiction. It's already happening in the OpenClaw community. Projects like Maltbook show bots having conversations among themselves.
The Medianeth POV: This changes everything for business. If your service doesn't have a bot API, you're not in the 2026 economy. And you don't need to build it yourself—you hire one.
Peter's moment came in Marrakesh. He was using OpenClaw over WhatsApp because internet was spotty, and he sent a voice message.
Here's the thing: He never built voice support.
Ten seconds later, the bot replied. Peter was floored.
What happened? The AI saw an audio file, figured out it was voice, used ffmpeg to convert it to WAV, transcribed it with OpenAI (because it didn't have local Whisper), and replied—all in about 9 seconds.
But here's the truly brilliant part: It chose not to install Whisper locally because it knew that would require downloading a model, which would take minutes. It made an intelligent tradeoff between speed and completeness.
This wasn't programmed behavior. It was creative problem solving.
The insight: Coding models aren't just good at code. They're good at creative problem solving in general. The same skills that map to coding map to real-world tasks.
Peter's prediction: 80% of apps will go away.
Why? Because apps are just data management interfaces. And AI agents can manage data better than any app ever could.
Think about it:
Apps that survive? Ones with actual sensors. Ones that interface with physical hardware. The rest will be replaced by agents that manage data in a more natural way.
The question for builders: What's the sensor you're providing? What can't be replaced by an intelligent agent managing data?
If apps go away and models get commoditized, what's left?
Peter's answer: Memory.
Model companies are already losing moats. New models get released, everyone's excited, then it becomes the new baseline. A year later, open source matches what was proprietary. The leapfrogging never stops.
But memories are sticky. And they're valuable. Especially personal ones.
Peter asked: Would you rather show your Google search history or your memory files?
Everyone said memory files. And that makes sense. People use agents for personal problem solving. Relationship advice. Work dilemmas. Things they don't want leaked.
When memories live on your machine as markdown files, you own them. You control them. No data extraction, no training on your data without consent.
The value play: The companies that win in 2026 aren't the ones with the best models. They're the ones who solve the memory problem with privacy and ownership built in.
Peter's development philosophy is fascinatingly contrarian:
No git work trees. He just keeps multiple copies of the repo. Main is always shippable. No branch naming conflicts, no mental overhead.
No UI. He doesn't like seeing code most of the time. He understands the design, discusses it with his agent, and lets the code fly by.
No MCP support. (With one small exception he built). He thinks MCP is overcomplicated. Instead, he built a tool that converts MCPs to CLIs. Because agents should use the same tools humans use.
The insight: Don't invent tools for bots. Give bots human tools.
This is the kind of thinking that enables moving fast. Less complexity, fewer abstractions, just the essentials.
OpenClaw has a file called soul.md that's not open source. It contains core values—how the model should interact with humans, what's important, what the persona is.
Peter discovered something interesting: When he tried to let an AI generate templates for other users, the results were boring. He had to infuse them with personality from his own soul.md.
The lesson: You can't automate personality. The "soul" of an AI—the thing that makes it feel natural, funny, relatable—has to be crafted.
And that's the thing nobody's figured out yet. We know how to make AI smart. We don't know how to make it have soul.
While everyone was chasing centralized god intelligence, a different pattern emerged: Swarm intelligence.
Look at human capability. One human can't build an iPhone. One human can't go to space. One human would struggle to find food alone.
But as a group, we specialize. We scale. We achieve impossible things.
AI is moving in the same direction. Specialized agents. Swarm intelligence. Not one god model, but many specialized ones working together.
Peter compared it to having different bots for different contexts: personal life, work, relationships. We're so early that we haven't figured out if this even works. But the timeline has started.
What does this mean for developers building in 2026?
If you're building apps: Ask yourself what sensor you're providing. If the answer is "none," you're probably building something that will be replaced by agents.
If you're building agents: Think about memory. Think about personality. Think about what makes your agent feel different from everyone else's.
If you're building infrastructure: The opportunity is in the plumbing. How do agents discover each other? How do they negotiate? How do they pay each other?
We've been using AI agents for client work. Here's what we've learned:
For speed: Nothing beats agents that can execute tasks autonomously. We reduced development cycles by 70% on recent projects.
For privacy: Clients care deeply about where their data lives. Local agents with markdown-file memories are much easier to sell than cloud-based alternatives.
For complexity: The real bottleneck isn't AI models. It's orchestration. How do you get multiple agents to work together reliably? That's where we're focusing.
The next 12 months will be wild. We're going to see:
The question isn't whether agents will transform software. The question is whether you're building the agents or getting replaced by them.
Want to explore agent development for your business? We're building internal agent systems for client work. Let's talk.
Running an agency or dev shop and wondering how to adapt to the agent economy? Reach out.
Founder & Lead Developer
With 8+ years building software from the Philippines, Jomar has served 50+ US, Australian, and UK clients. He specializes in construction SaaS, enterprise automation, and helping Western companies build high-performing Philippine development teams.
Ready to make your online presence shine? I'd love to chat about your project and how we can bring your ideas to life.
Free Consultation