Building a WhatsApp relay to escape the keyboard led to OpenClawPeter explains that OpenClaw originated from his own annoyance at not being able to send prompts to his computer from his phone while away from the keyboard. He built a WhatsApp relay, which evolved into a proactive agent that felt magical, and later grew into a Discord bot that went viral.Peter SteinbergerOpenClaw origin storySource at 1:07Save viewpoint
OpenClaw went viral overnight after an accidental agent runPeter describes the moment OpenClaw went viral after he accidentally left the agent running overnight, waking up to 800 messages. The project quickly exploded with attention, reporters, and community contributions, leading to both fame and overwhelming pressure.Peter SteinbergerViral growth and attentionSource at 6:12Save viewpoint
Choosing OpenAI over VC and trusting personal brandPeter discusses his decision to engage with and ultimately join OpenAI, despite receiving VC offers. He frames it as trusting his gut and notes that his personal brand is more important than any single product, even calling the move 'selling my soul.md'.Peter SteinbergerDecision to join OpenAISource at 10:50Save viewpoint
Rivals won on marketing while OpenClaw faced feature creepPeter acknowledges that competitors like 'the anime girl company' (likely referring to a rival) beat them in marketing and community presence while OpenClaw was busy with security hardening and feature creep. He also cites the massive number of configuration options and dependency on Anthropic's model as key challenges.Peter SteinbergerCompetition and security pressureSource at 13:35Save viewpoint
Personal brand outlasts software and needs early attentionPeter advises that personal brand is more important than any product because software can be forked or cloned but a name cannot. He recommends building visibility early, as attention is the scarcest resource in a world where building is fast.Peter SteinbergerPersonal brand and lessonsSource at 12:51Save viewpoint
Fun drives velocity and building what annoys youPeter emphasizes that fun is the ultimate driver of velocity and good ideas, and that building what annoys you can lead to the next big thing. He also stresses staying focused and living in the future to build what's missing.Peter SteinbergerBuilding philosophy and future directionSource at 22:50Save viewpoint
Code review as risk management with agent-assisted testingPeter explains his approach to code review as risk management rather than reading every line, trusting intuition about how long changes should take. He also discusses using agents and sub-agents for testing and code review to improve reliability.Peter SteinbergerEngineering practices and code reviewSource at 29:42Save viewpoint
Learning to say no and using vision.md to guide OpenClawPeter reflects that he didn't say no enough to features and PRs in OpenClaw, leading to bloat and maintenance burden. He now uses a vision.md file to define project direction and emphasizes the responsibility of merging community code.Peter SteinbergerOpen source community managementSource at 36:09Save viewpoint
Managing compute is the biggest agentic infrastructure bottleneckPeter identifies managing compute as the biggest bottleneck in agentic infrastructure, particularly for local and Mac-specific workloads. He notes the lack of good systems for managing distributed compute and driving fleets reliably.Peter SteinbergerAgentic infrastructure bottlenecksSource at 34:52Save viewpoint
Startups should focus on eyeballs and hard boring categoriesPeter advises that the hardest problem for startups is getting eyeballs, not technology or software. He suggests building something you want to use, working on personal brand and visibility, and considering 'hard and boring' categories with less competition.Peter SteinbergerStartup advice and visibilitySource at 40:01Save viewpoint
