Noah Lisk, home
RésuméContactMenu

Podcast

Peter Steinberger- "Fun Is Velocity"

Y Combinator Startup Podcast

Source
  • Built A Phone Relay In An Hour
    • Peter built a WhatsApp relay in an hour to send prompts from his phone to his desktop so his agents could continue while he raided the kitchen.
    • The relay felt magical because it removed terminal friction and made agents proactive and conversational like a friend. (Time 0:01:17)
  • Emotional Reactions Reveal Product Market Fit
    • Non-technical friends’ anger at being told it’s ‘not for them’ signaled product-market fit for Peter.
    • He iterated a month more, added channels, and accepted community PRs that revealed real demand. (Time 0:05:05)
  • Discord Night That Sparked Viral Growth
    • Peter opened a Discord server and let people interact with his agent live, watching them hack and test it overnight.
    • A launch-demon restart plus 10 hours of sleep produced ~800 messages and signaled the moment it went viral. (Time 0:06:39)
  • Invest In Your Personal Brand Early
    • Build a personal brand early because your name cannot be forked or cloned and becomes more valuable than any single product.
    • Start working on that visibility before you need it so it pays off when something breaks out. (Time 0:12:57)
  • Security Scarcity Scales Faster Than Fixes
    • OpenClaw became a prototype for open-source security headaches: press amplified a small percentage of problematic skills into a scare.
    • Peter hardened code with sandboxing, allowlists, and permissions, but many users didn’t notice those details. (Time 0:13:55)
  • Say No To Feature Creep
    • Avoid adding every feature because each one multiplies future maintenance and configuration complexity.
    • OpenClaw amassed ~9,500 config options, making testing impossible and evolution much harder once users depend on it. (Time 0:16:32)
  • Peak And Trough Download Rollercoaster
    • OpenClaw’s download graph showed both extremes: 835k weekly downloads in May and a peak of 4.7M after being declared dead in June.
    • Hype behaves like weather: predictable in direction but uncontrollable in magnitude and timing. (Time 0:19:04)
  • When A Passion Project Turns Into A Calling
    • Peter Steinberger says OpenClaw stopped feeling fun around February as its rapid growth turned the project into a responsibility.
    • He unexpectedly found himself with two jobs after initially not wanting to build another company.
    • The shift from personal passion to public obligation made him stop using his own product.
    • He realized he was building for everyone rather than making something he personally loved. (Time 0:19:33)
  • Creator Stopped Using His Own Product
    • By February the project stopped being fun for Peter; it became a responsibility and he even stopped using his own product.
    • He shifted from building something he loved to fixing bugs, security, and support for a broad user base. (Time 0:19:35)
  • Fun Is Velocity
    • Peter Steinberger says the joy of building returned when he stopped treating OpenClaw as a never-ending support and security burden.
    • He finds software most inspiring when he can prompt his agent to change it directly.
    • Fun creates a competitive advantage because someone who is enjoying the work can move faster.
    • The weeks he enjoyed building, the product visibly improved; the less enjoyable weeks produced mostly configuration options. (Time 0:21:35)
  • Fix What Annoys You And Keep Having Fun
    • Keep building things that genuinely annoy you because solving your own problems produces authentic, fast progress.
    • Peter summarizes: Don’t stop having fun, listen to your gut, and stay focused to ship what matters. (Time 0:22:47)
  • Open Source Agent As Privacy Alternative
    • OpenClaw aims to be the open-source alternative to lab-provided agents: runs anywhere, works with any model, and keeps local data local.
    • This positioning matters as a privacy and portability thesis for users running local models. (Time 0:25:10)
  • Optimize Tokens Before Always On Agents
    • Always optimize token usage before building always-on agents because heartbeats and large sessions can burn huge tokens.
    • Peter’s early heartbeat design sent hundreds of thousands of tokens for stale checks, costing money and requiring smarter orchestration. (Time 0:37:22)
  • Run Agents On A Remote Machine
    • Host agents on a dedicated remote machine so agents can click and interact without stealing your local cursor and battery.
    • Peter uses a studio Mac accessed via Jump Desktop so agents can control a separate machine safely. (Time 0:38:44)