Behind the Craft Podcast: Episode Summary
Episode Title:
How OpenClaw’s Creator Uses AI to Run His Life in 40 Minutes | Peter Steinberger
Host: Peter Yang
Guest: Peter Steinberger (Creator of 'Clawd', the AI assistant)
Date: February 1, 2026
Overview
This episode explores Peter Steinberger’s creation, Clawd (pronounced "cloud" and themed as a lobster with claws), a hacker-friendly personal AI assistant that integrates with virtually every part of your digital life. Through firsthand stories and technical deep-dives, Steinberger reveals how Clawd functions as both a resourceful 'friend' and the beginnings of a new AI-powered operating system, sharing practical insights on AI coding, productivity, and the future of personal assistants. The conversation is fast-paced, irreverent, and packed with hot takes about the limits and superpowers of current AI agents.
Key Discussion Points & Insights
1. What is Clawd and Why a Lobster?
- Motivation: Steinberger wanted a way to check up on his computer from anywhere, inspired by the "vibe coding" trend and the lack of a seamless, powerful AI agent from big tech labs.
- Origin: Started as a basic WhatsApp interface to trigger code via cloud, it spiraled into a 300,000-line open-source project spanning all major messaging platforms.
“It was very simple. It was put in one hour, and it kind of got a life of its own.” — Peter [01:35]
- Philosophy: An infinitely resourceful assistant that knows your habits, blurring the lines between apps and creating the experience of “a new weird friend that is also really smart and resourceful” [09:10].
- Lobster Branding: Cheekily named 'Clawd' (with a W) in tribute to “lobster claws” and spelling humor.
2. How Clawd Works — Real-life Examples
- AI Autonomy: Clawd can independently interpret inputs, access context on your computer, write commits, and interact on your behalf, such as fixing bugs or replying to tweets.
“It checked out the git repository, it fixed it, it did a commit, and then it replied to the person on Twitter that it’s fixed.” — Peter [00:12 & 02:57]
- Handling Voice and Unstructured Inputs: Even unsupported features (like voice) are improvised—Clawd converts audio, finds transcripts, and replies contextually.
“I sent it a voice message... It showed me the typing indicator… Then it just replied as if nothing happened.” — Peter [03:08]
- Superpower Examples: Clawd can:
- Visualize music and sound
- Hack food delivery APIs to track orders
- Control smart home devices (Philips Hue, Sonos, KNX systems)
- Reverse engineer APIs (ex: 8 Sleep mattress temperature)
- Manage email, calendars, files, or even check you in for flights.
3. Building With and On Clawd
- CLI Armies: Steinberger built a suite of command-line interfaces (CLIs) for Google services, meme lookup, and more.
- Meta-Programming: AI can even read and reprogram its own codebase dynamically.
“If your agent can read its own source code… it can literally reconfigure and reprogram itself and then restart…” — Peter [11:10]
- Community Participation: Lowered technical barriers have drawn non-developers to contribute by sending “prompt requests” via pull requests.
4. Installation and Accessibility
- Cross-platform: Written in TypeScript, works on macOS, Linux, Windows.
- Install Process: Simple one-liner via website; alternative NPM or GitHub repo checkout for power users.
“Talking to a thing on iMessage or WhatsApp… is just like having a new weird friend… that makes the whole technology very approachable.” — Peter [09:10]
- Customization: Pick models like Anthropic or OpenAI; plug into Telegram, Discord, WhatsApp, and more.
5. Use Cases — For Power Users and Beginners Alike
- For Tinkerers: Deeply customizable for automation: “People hooked it up to their messaging system so it can reply not just to you, but to everyone. And you can… hook it up into a group chat, which is even more fun.” — Peter [17:23]
- Everyday Tasks: Calendar management, reminders, sleep tracking, food logging, emails, shopping, and integrating with fitness apps or password managers.
- Role as a Personal AI: Reminding, nagging, “roasting,” or simply managing everyday tedium.
“It will blend away probably 80% of the apps that you have on your phone.” — Peter [00:32 & 19:44]
6. Philosophy on AI Coding — “The Agentic Trap” and Building with Taste
- Hot Take: Many engineers fall into the trap of endlessly building toolchains for agents, rather than making something truly useful.
“It would be better if they could do a little bit more, and then they really fall deep into this rabbit hole… but you’re not actually building something that really brings you forward.” — Peter [22:50]
- On Play and Learning: Iterative, playful tinkering is essential to learning and unlocking product sense with AI models, but “vanity metrics” (like how long an agent can run) don’t matter if the output is “slop.”
- The Human Factor: The best AI-driven products still require taste, vision, and a human “in the loop.”
“Just because you can build everything doesn’t mean you should or that it’s going to be good.” — Peter [26:42]
- Encouragement to Experiment: Persistence, iteration, and conversation with the AI are key; prompt engineering and understanding model quirks is a new literacy.
7. Collaboration, Workflow, and Community
- “Prompt Requests” Instead of Pull Requests: New contributors can suggest features or fixes just by expressing intent.
- Multi-agent Productivity: Running multiple independent instances (vs. large agent orchestrators) can speed up building and help maintain creative flow.
- Unlocking Non-tech Power Users: AI empowers non-coders (e.g., Steinberger’s business partner, a former lawyer) to contribute meaningfully to technically complex projects.
Notable Quotes & Memorable Moments
- On AI autonomy:
“These things are so resourceful, although in a scary way… It’s like unshackled ChatGPT.” — Peter [00:37]
- On risk and power:
“Of course with a lot of power comes a lot of risk… if you like, keep saying yes, it probably comply and probably also delete itself and crash.” — Peter [09:43]
- On “roasting”:
“I built mine so that it can roast me… And it probably doesn’t know that it’s on camera right now.” — Peter [13:33]
- Sample roast from Clawd:
“You’re so obsessed with the tools that you literally build yourself a friend because debugging code is more fun than dating.” — Clawd, as read by Peter [14:16]
- On AI as replacement for point apps:
“Why should I use my fitness pal to track food when I have an infinitely resourceful assistant that already knows I’m making bad decisions and I’m Kentucky Fried Chicken?” — Peter [19:44]
Example Use Cases (with Timestamps)
- Bug fixing end-to-end via tweet + WhatsApp: [00:05], [02:57]
- Handling voice commands without explicit support: [03:08]
- Controlling smart home and handling daily life (lights, cameras, food delivery): [14:39], [15:21]
- Auto-checking in for flights, including locating and extracting passport info: [16:05]
- Tracking sleep, fitness, calendar, or controlling expenses: [17:17]
- Custom reminders and “roasting”: [14:16], [18:24]
- Auto-bookmarking tweets, managing to-dos, family- or group-based bots: [17:17], [18:04]
Workflow, Development, and Model Selection
- Natural conversation is most powerful; just talk to your AI, iterate, and let it learn:
“You just talk to a friend.” — Peter [20:46]
- Choosing models:
- Prefers Anthropic/Opus for wit/personality, but supports a range (OpenAI, etc.).
- On persistent memory:
“It has persistent memory… The more you use it, the more powerful it gets.” — Peter [21:33]
- Development workflow:
- Screenshots or pastes text/conversations into the AI to start building or debugging.
- Scraping Discord help requests to surface pain points and auto-generate FAQ. [33:16]
The Takeaway
- Clawd is a powerful, general-purpose AI assistant that aims to erase the need for dozens of specialized apps by integrating with your digital life through natural language interaction.
- Its creator, Peter Steinberger, is both a builder and philosopher on how to get the most from AI: avoid tool-building rabbit holes, keep humans “in the loop” for vision and taste, and embrace the messy, playful process of learning by building.
- For non-coders and tinkerers alike, Clawd offers a glimpse into the near future of truly personal, persistent AI companions.
How to Try Clawd
- Find it on GitHub: Search for “Clawdbot” (with a ‘w’).
- Website: cloudbot.
- Install: One-liner installer, supports macOS, Linux, Windows.
- Beginner-friendly: Start by letting it manage tasks like your calendar or reminders; customize as your confidence grows.
Final Thought
“Find your own path… You have to make your own mistakes. That’s how you learn with everything in life. And that’s also how to learn those things. Just that this space is evolving very fast.” — Peter Steinberger [36:45]
End of Summary
