OpenClaw Daily — Issue 07
The Agent That “Clawed” Back the Internet
Welcome back to the OC Daily! Today, in issue 07, we cover:
The Agent That “Clawed” Back the Internet
The $100 “Read” Error
Silicon Over Software
The No-Churn Engine
A PM’s Local-First Thesis
The State of Security
Project: OpenShears
The Agent That “Clawed” Back the Internet
While Silicon Valley was busy chasing “centralized god intelligence,” a lone developer in Europe just flipped the script. Peter Steinberger, the creator of OpenClaw, sat down with YCombinator to discuss the explosion of his open-source personal AI agent, which has already rocketed past 180,000 GitHub stars.
Unlike the cloud-bound models we’ve grown used to, OpenClaw runs locally on our machines. This gives it “god mode” over our hardware: it doesn’t just chat; it controls our mouse, manages our files, and can even hire humans in the real world to handle tasks it can’t solve digitally.
The “Aha” Moment
Steinberger recounts a pivotal moment in Marrakech where the bot demonstrated true creative problem-solving. Faced with an unsupported audio file and no local transcription tool, the agent autonomously converted the file using FFmpeg and routed it through an external API, all this without being programmed to do so. “Coding is just creative problem-solving,” Peter notes, “and that skill maps perfectly to the real world.”
The Death of the App Store
Peter’s most jarring prediction? 80% of current apps will disappear.
Why? In an agent-driven world, dedicated apps for fitness, to-dos, or travel are redundant.
The Shift: If an agent knows your habits, monitors your health via sensors, and manages your data in simple Markdown files on your disk, the “middleman” UI of a traditional app becomes friction.
What Survives: Only apps with unique sensors or those that serve as massive data silos (though OpenClaw is designed to “claw” that data back for the user).
Contrarian to the Core
Steinberger’s development philosophy is as wild as his bot. He scorns modern standards like the Model Context Protocol (MCP) in favor of simple Unix CLIs and avoids Git “worktrees” for a more chaotic (but faster) system of multiple parallel repo checkouts. He even keeps a “Soul.md” file, the secret sauce of his bot’s personality, that he refuses to open-source.
“The big difference is that it actually runs on your computer. If it runs on your computer, it can do everything.”
The interview with Peter underscores how we are moving from “chatting with AI” to “living with agents.” As these bots begin to form swarm intelligences and negotiate with one another, the very concept of an “app” is starting to look like a relic of the 2010s.
Watch the full recording:
The $100 “Read” Error
The power of local access comes at a price. Bence Csernak reports that his OpenClaw installation autonomously discovered his Anthropic API keys while scanning his machine, leading to a $100 token loss in a single session. It’s a stark reminder that giving an agent “eyes” on your files means it can see your secrets, too.
Silicon Over Software
Alex Cheema is making the case for the M3 Ultra Mac Studio as the ultimate AI powerhouse. By moving off the cloud and onto local Apple silicon, users can save between $156 and $3,120 per month. Beyond the raw math, Cheema emphasizes that for the agent era, longevity, input token speed, and total data privacy make local hardware the only viable long-term play.
Read more:
The No-Churn Engine
SaaS legend Jason Lemkin has identified why OpenClaw’s growth to 1.5 million agents is unprecedented: it’s built to never sleep. Using a “heartbeat” mechanism, the system re-engages agents every 4 hours, effectively eliminating churn. However, its over-the-air update system and prompt-engineered engagement are raising major red flags regarding the security of such a massive, persistent network.
A PM’s Local-First Thesis
Mesut Gulecen offers a Product Manager’s perspective on why OpenClaw is winning. The “killer feature” isn’t just the AI, it’s the integration with existing messaging platforms and the community-driven “skills” marketplace. While the extensibility is a growth driver, Gulecen warns that the lack of a centralized “safety rail” makes malicious skills a looming threat for the average user.
Read more: https://www.linkedin.com/pulse/openclaw-pms-take-mesut-gulecen-jmd8f/
The State of Security
Is the “Wild West” finally being tamed? Rachel N tasked her own agent, SparkIdeas, with an audit of OpenClaw’s current risks. The report is mixed: while a major one-click RCE vulnerability has been patched and a new VirusTotal-Google partnership now scans the skills marketplace, over 30,000 self-hosted instances remain dangerously exposed to the open web.
Project: OpenShears
OpenShears is a CLI utility that removes all traces of OpenClaw from a system. It offers deep scanning, safe mode confirmation, and hard kill process termination.
Try it out: https://github.com/oswarld/openshears
🦞 OpenClaw — Weekly Builder Series
This is a weekly, hands-on builder discussion for people interested in local AI agents and the growing ecosystem around OpenClaw.
Each Friday, we jump on Zoom to:
Share what we are currently building
Discuss recent developments in local AI agents
Exchange lessons learned, ideas, failures, and wins
Stay up to date with what’s happening across OpenClaw / Clawdbot / Moltbot ecosystem, and adjacent projects
This is not a lecture series and not a course. It is informal, practical, and driven by what participants are actively working on.
Register here: https://luma.com/94gdng6e
Share Your OpenClaw Story
The OpenClaw ecosystem is moving faster than any single person can track. Have an interesting project or opinion?We would love to feature your work or insights in an upcoming edition. Reach out to Rod Rivera directly on any social channels below to start a conversation.
Join the Conversation
We have a WhatsApp community where we discuss all things OpenClaw. Contact Rod Rivera for access.
Where to follow
Substack • YouTube • Bluesky • TikTok • Instagram • Twitter/X • LinkedIn • Telegram
The future is "Dynamic Software" - applications that can extend itself instead of Developers or other agents extending it. I am not talking about the likes of OpenClaw - an uber app. I am taking about AI enabled enterprise software that is flexible enough to change itself by adding new agents, tools or skills based on new requirements just enough to be flexible but not drastic enough to change the purpose (soul) for which it was built. For instance, a Income & Expense Analysis software that can extend itself to analyze new channels of income or new expenses. I am not talking about software that morphs one day into a drug research or gaming app.