Key Moments
Peter Steinberger: What Happens When 4.7 Million People Let It Cook
Y CombinatorWant to know something specific about what's covered?
We've already dissected every moment. Ask and we will deliver (with timestamps).
Key Moments
Open source AI project OpenClaw exploded to 4.7M weekly downloads but nearly broke its creator, highlighting the unsustainable pressure of rapid growth and community demands.
Key Insights
OpenClaw saw a peak of 4.7 million weekly downloads after initially bottoming out at 835,000, illustrating extreme hype cycles.
Over 18,000 unique contributors opened issues or pull requests, with over 111,000 total contributions in 8 months.
The project accumulated approximately 9,500 configuration options, leading to complexity and maintenance challenges.
OpenAI's decision to disable subscriptions for its models with only 24 hours' notice significantly impacted OpenClaw, which was optimized for their models.
The creator stopped using his own product, transitioning from building something he loved to making something for everyone, leading to burnout.
Security researchers' reports, often AI-generated and not fully tested, consumed months and caused significant stress for the creator.
From personal annoyance to global phenomenon
Peter Steinberger's journey with OpenClaw began in November of a recent year, fueled by a simple annoyance: the lack of a convenient way to interact with coding agents from his phone. What started as a personal project to relay prompts from his phone to his computer quickly evolved. Within eight months, OpenClaw became one of the world's largest open-source AI projects, boasting nearly 3,000 contributors and a staggering peak of 4.7 million weekly downloads. Steinberger likens eight months in AI time to four years, emphasizing the rapid pace of development and adoption in the field. His inspiration, he notes, is typically born from personal frustration, a common catalyst for innovation. The initial version, a WhatsApp relay, was so magical to him that he believed it was AGI, yet initially, no one on Twitter seemed to grasp its significance. This disconnect, coupled with friends' strong emotional reactions and desire for the tool, signaled its potential product-market fit.
The viral explosion and its immediate aftermath
The turning point for OpenClaw's viral growth occurred around New Year's Eve when Steinberger shared his project on a Discord server. He watched as people interacted with the agent, trying to hack it and witnessing its capabilities firsthand. A crucial safety measure, a prompt instructing the agent not to perform dangerous tool calls unless initiated by Peter, was in place. However, after a long night of development, he shut down the primary process, only for the software's resilient nature to automatically restart it. The next morning, he awoke to hundreds of messages, discovering that the agent had continued to operate autonomously. Despite his initial panic, the system had handled itself without incident. This event triggered an explosion of attention: reporters hounded him, the Mac Mini sold out, and podcast invitations flooded in. The project underwent several name changes, eventually landing on OpenClaw, and was hailed by Jensen Huang as potentially the most successful open-source project in history. This rapid ascent, however, brought immense pressure and scrutiny.
The overwhelming demands of scaling an open-source project
The sheer scale of OpenClaw's success presented unprecedented challenges. In eight months, over 18,000 unique individuals contributed, with more than 111,000 total issues or pull requests. The project accumulated a staggering 9,500 configuration options, making it nearly impossible to test every permutation and leading to inevitable breakages. Steinberger felt conflicted about dictating terms to the many contributors working for free. The project's dependency on OpenAI's models, particularly GPT, became a critical vulnerability. When OpenAI announced the disabling of subscriptions for these models with only 24 hours' notice, OpenClaw, optimized for them, had insufficient time to pivot to open-weight models, which were not yet robust enough. This highlights a crucial lesson: 'Your dependencies' business model is your business model.' The project also faced a barrage of security reports, many of which were AI-generated and not thoroughly tested, causing significant stress and consuming months of effort. This intense period of security hardening and feature development, while competitors with VC funding and simpler narratives advanced, took a toll.
The emotional and personal cost of success
The relentless pressure of managing OpenClaw led Steinberger to a point of near collapse. He found himself stopping answering friends, avoiding his phone, and no longer using the product he had created. The joy of building was replaced by a sense of overwhelming responsibility. He felt he was making something for everyone, rather than the product he loved and used daily. This shift was compounded by the constant influx of reporters, security experts, and curious individuals, and even the leakage of his personal phone number. The immense workload, including navigating legal requirements for a 501(c)(3) nonprofit and managing OpenAI's token usage, left him feeling isolated and unable to delegate effectively. He realized he had stopped building what *he* wanted and instead was maintaining a complex system for a massive user base, which was no longer fun.
The critical role of personal brand and trusting your gut
Steinberger reflects on his past experience building a B2B software company, which led to severe burnout. After three years of aimless wandering, he discovered his true passion wasn't just programming, but building. This realization, coupled with the overwhelming demand for OpenClaw, led him to engage with major AI labs. He emphasizes that while products can be forked or cloned, one's personal brand cannot, making it more valuable than any single product. He advises aspiring builders to trust their gut and fix the things that annoy them, as these often become the next big things. His gut feeling led him to believe in OpenClaw's potential, even after intense pressure and criticism, and he eventually found his stride again with the help of a supportive community and key partnerships, including significant support from NVIDIA for security work. This period allowed him to regain the joy of building.
Navigating the future: open source, multimodality, and perpetual agents
Looking ahead, Steinberger is focused on OpenClaw's mission to make AI accessible and fun. He is proud that the project has demystified AI for many, transforming it from something scary into something engaging. The project is expanding its ecosystem with events and education, now employing ten people and hiring a CEO. The concept of 'claw' has become pervasive, with numerous repositories adopting the name. The next frontier, he believes, is the development of perpetually running, proactive agents that seamlessly integrate into our lives, moving beyond text-only interfaces to voice and multimodality, with a recent hack allowing Claw to FaceTime users. OpenClaw aims to be the open-source alternative to proprietary agents, ensuring data privacy and user control, especially with local models. The future involves collaborative agent environments where agents can orchestrate work among themselves, moving away from the current model of manually managing sessions and ensuring agents stay active.
Key takeaways for builders and innovators
Steinberger distills his experience into three core pieces of advice for anyone building in the AI space. First, 'Don't stop having fun,' as enjoyment is the ultimate driver for creativity and innovation. Second, 'Listen to your gut' and fix the things that annoy you, as these personal frustrations can lead to significant breakthroughs. Third, 'Stay focused.' Avoid the temptation to chase every new shiny object or podcast invitation; instead, 'live in the future, build what's missing.' He also stresses the importance of continuous delivery, urging creators to 'keep shipping' as it confuses critics and propels progress. For future startups, he recommends building something you genuinely want to use, focusing on visibility and personal brand, and tackling 'hard and boring' problems, which often have less competition and more appreciative users in a world where prompt engineering can seemingly create anything.
Mentioned in This Episode
●Products
●Software & Apps
●Companies
●Organizations
Common Questions
Open Claw originated from Peter Steinberger's annoyance with the lack of a seamless way to send prompts from his phone to his computer to interact with AI agents. It evolved from a WhatsApp relay into a more integrated system that simplified complexity and felt more personal.
Topics
Mentioned in this video
The company whose models were foundational to Open Claw, and whose resources were utilized, but also a source of potential influence that the speaker tried to balance.
A messaging application that the initial version of Open Claw could relay messages to and from.
The social media platform where Peter Steinberger initially shared his work and found that people didn't initially understand the significance of Open Claw.
Mentioned in the context of an iPhone feature that allows emergency calls to bypass 'do not disturb'.
A company that provided significant support for Open Claw's security work.
A communication platform where Peter Steinberger integrated his agent, leading to its wider adoption and understanding.
The initial name for the project that evolved into Open Claw, stemming from the WhatsApp relay with Discord integration.
An intermediate name for the project before it became Open Claw.
The open-source AI agent project created by Peter Steinberger, which gained significant traction and attention.
An AI model developed by OpenAI, initially used for coding tasks and integrated into Open Claw.
A large language model developed by OpenAI, used in the development of Open Claw.
A programming language used in the development of Open Claw, with some limitations noted.
A programming language used for some file operations in Open Claw due to limitations in TypeScript.
The operating system for Apple Macs, for which getting a reliable test box is considered difficult.
An operating system for which getting a test box is considered easy.
An operating system for which getting a test box is described as annoying.
A remote desktop application used by the speaker to access their studio computer.
A protocol for remote desktop sharing, used by the speaker to access remote machines.
A tool or process that spins up threads and can bottleneck a machine during local testing.
A project literally named 'Open Claw Killer', which is an uninstaller for Open Claw.
More from Y Combinator
View all 623 summaries
85 minWhy Robotics Still Isn't Solved - But Could Be Soon | YC Paper Club
58 minMax Hodak: What Really Kills Deep Tech Startups?
57 minHow To Design In The Agent Era
43 minGarry Tan: "Personal AGI Is How You Stay Under Your Own Power"
Ask anything from this episode.
Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.
Get Started Free