Eric Simons on Bolt, WebContainer, and the Text-to-App Revolution

Eric Simons on Bolt, WebContainer, and the Text-to-App Revolution

transcriptlennyeric-simonsboltstackblitzvibe-codingai-codingproduct-management

Eric Simons on Bolt, WebContainer, and the Text-to-App Revolution

Eric Simons — Lenny’s Podcast · ~2025 · Source

Eric Simons, co-founder and CEO of StackBlitz, tells the story of Bolt.new — a text-to-app builder that went from $0 to $20M ARR in its first two months and $40M in five, with a 20-person team serving roughly one million monthly active users. The conversation covers the seven-year technical bet on WebContainer (a browser-native OS), why Claude Sonnet was the enabling threshold for AI code generation, what 67% non-developer usage reveals about the future of software org charts, and the bootstrapping philosophy that kept the company alive long enough to catch the wave.

Key ideas

  • 7-year overnight success. StackBlitz spent seven years building WebContainer — a WebAssembly-based operating system that runs Node.js and full dev toolchains inside the browser, using the user’s own CPU rather than cloud VMs. The company was nearly shut down when they launched Bolt in October 2024 with a single tweet; $0 to $20M ARR in two months followed. The technical advantage: zero-latency, zero cloud cost compute that scales to one billion devices rather than the ~100M VMs available globally. Inspiration came from observing Figma’s analogous WebGL bet in 2012.

  • Claude Sonnet was the unlock. StackBlitz tried building Bolt a year earlier with the then-frontier models; output was unreliable and apps were broken or ugly. A sneak peek of Sonnet convinced them to restart the project. Simons’s insight on why code is the ideal LLM vertical: software execution is deterministic — code either runs or it doesn’t — which enables massive RL training data permutation in a way that non-deterministic domains (law, medicine) cannot support. Anthropic “went deep” on this; everyone else is now following.

  • 67% non-developer users — the PM opportunity. Most Bolt users are PMs, designers, and entrepreneurs, not developers. The skill set that makes an excellent PM — defining scope clearly, articulating outcomes, having taste to evaluate results — maps directly onto effective AI-assisted development. Simons’s prediction: future org charts will have PM/designer-led pods writing code directly (via Bolt), with 1–2 engineers reviewing rather than executing. “The entire software world order is going to get rewritten.”

  • Low burn as strategic asset. StackBlitz bootstrapped for the first two or three years, then raised but barely spent. During 2020–21 exuberance, they deliberately chose not to triple headcount (“if we had, there would be no Bolt”). The operating philosophy: “Until you see pull, just people pulling the product out of your hands, you don’t want to be spending money.” Core team retention — five to seven people together for five or more years — is named as the primary competitive moat, enabling the rapid execution once Bolt launched.

  • Zero-fidelity-loss operations at hypergrowth. With one million MAU and a 20-person team, StackBlitz held a company-wide Zoom call every single day for at least an hour during the rapid growth phase. Rationale: “When you’re in these times of just extreme growth, you want as close to 0% loss on communications.” Simons is explicit this won’t scale indefinitely, but credits it as the mechanism that allowed correct independent decisions to be made at pace.