This video is the full interview from Peter Steinberger's first appearance on the TBPN live tech talk show since launching his AI agent project Clawdbot (now renamed Maltbot). He candidly and humorously shares how he recovered from burnout after selling his 13-year software company, discovered his deep interest in AI, the development process and philosophy behind Clawdbot, and the diverse experiences from the project's unexpected explosive popularity. It offers deep insight into AI agents' potential and future, as well as the importance of the open-source community.

1. Burnout and Return to AI

Peter recalls his career history. After working at his software company for 13 years, he sold it 4 years ago in a state of complete burnout. He needed three years of rest.

"I worked at my software company for 13 years, and I sold it 4 years ago. I was completely burnt out."

In April 2025, he rediscovered his passion for coding and became drawn to the emerging AI space.

"I wanted to jump into AI but didn't want to feel stupid without experience. So I looked at AI and it was 'okay.' I thought 'why is nobody talking about this?'"

He started an AI agent anonymous meeting ('Claude Code Anonymous,' later 'Agents Anonymous') and describes his obsession as a 'positive addiction.'

2. Maltbot's Origin Story and Philosophy

Peter's development philosophy is "having fun."

"The best way to learn these new technologies is to have fun playing with them."

He implemented small useful ideas, tried various languages and approaches, and immersed himself in agentic engineering.

Maltbot started with WhatsApp integration:

"One morning I thought 'I want to chat with my computer via WhatsApp.' Agents were running and I wanted to check on them from the kitchen."

He built the simple integration in one hour — messages come in, invoke Claude Code, return results to WhatsApp. He added image-based prompting to provide more context with less typing.

The pivotal moment came during a trip to Marrakesh, Morocco. He sent a voice message to WhatsApp, which Maltbot didn't support — yet the agent responded perfectly. It had detected the file had no extension, checked the file header, identified it as Opus format, converted it to Wave using ffmpeg, tried Whisper but found it wasn't installed, found an OpenAI key in environment variables, sent it via curl to OpenAI, received the transcript, and replied.

"That moment, 'Wow!' — I realized the agent is 'remarkably smart and resourceful.'"

He defines Maltbot as "technology and art and exploration."

"This project is as much art and exploration as it is technology. It's like glue connecting pieces we already have, but in a completely new way of interacting with them. All the technical elements disappear, and it feels like talking to a friend or a ghost."

Peter saw potential in agents leveraging CLI and Unix systems rather than browser-based experiences.

"Agents know Unix. There can be thousands of small programs on the computer. They just need to know the name. They call the help menu, load what they need, learn how to use it, and use it."

3. Explosive Popularity and the Name Change

When Peter first shared Maltbot on Twitter, the tech community's response was lukewarm. But showing it to non-technical friends made everyone want one. He open-sourced the project, motivated by "having fun and inspiring people" rather than money.

"Last year was the year of coding agents. This year will be the year of personal assistants. I think Maltbot showed people there's a real need for this."

Anthropic contacted him requesting a name change — handled gracefully with an internal contact rather than lawyers.

"They were really kind. They sent an internal person, not a lawyer. But renaming a project with this much traffic was quite difficult."

4. The Future of Hardware and the App Ecosystem

Peter uses a Mac Studio with 512GB RAM for running local models. He predicts significant changes in existing business models and that many apps will disappear.

"I just take a photo of my food and my agent already knows I'm making a bad decision at McDonald's. Combined with this information, it knows exactly what I'll eat, and probably modifies my fitness program to meet my goals. Then the fitness app becomes unnecessary."

Most apps will shrink to APIs, and even APIs may become unnecessary. He believes this change isn't generational but will naturally reach non-technical people too.

"I met a design agency CEO with zero coding experience at a Vienna agent meetup. He started using Maltbot in early December and said 'I'm now running 25 web services. Building internal tools I need.' He doesn't know how coding works but is building software by chatting with the agent via Telegram."

This means people can create hyper-personalized software for free, perfectly matching their needs, without subscribing to random startups. He emphasizes "right now is when models are the worst" — it'll only get easier and faster.

5. Security Issues and the Community's Future

Peter identifies security as the biggest current problem. Built for personal use, Maltbot is now being used in untrusted environments.

"The little web app I made for debugging, once put on the public internet — all the threat models I didn't care about became real. Security emails are flooding in now."

He's building a team and remains optimistic Maltbot will eventually be very secure, while emphasizing it's a 'prototype' to show people what's possible, not a finished enterprise product.

He wants Maltbot to become a community larger than himself.

"I need help. I can't sleep and can't do this anymore. I want this project to outlast me. It's too cool to let it die."

When asked about founding a company, he mentioned considering a nonprofit foundation. On open-source licensing, he'll maintain MIT license accessibility and freedom.

"People could take this and sell it. But ultimately that doesn't matter much. Code isn't that valuable anymore. You could delete it and rebuild it in a few months. Ideas, public attention, and maybe the brand hold the real value."

6. Conclusion

Peter Steinberger has proven the enormous potential of AI agents and the power of open-source community through the Maltbot project. What began as pure passion for AI after burnout is now changing the app ecosystem and opening the door to hyper-personalized software development for even non-technical users. He defines Maltbot as not just a product but "technology and art and exploration" — a milestone pointing the direction for future AI agents. Amid security challenges and explosive demand, he hopes this project will endure through community power beyond himself, exemplifying the true spirit of the "indie developer."