Keep a small high-context team alive on minimal burn through fallow years so that when an enabling technology finally matures you can commit everything overnight and ride the wave others cannot.
Bolt
Bolt (bolt.new) is StackBlitz's browser-based AI app builder that turns a text prompt into a running full-stack web app. Its four stories share one through-line: a near-dead developer-tools company survived on low burn just long enough to catch an external technology wave, then rode seven years of unglamorous infrastructure into one of the fastest revenue ramps in software history, going from roughly $700K ARR to about $40M ARR in five months.
Pivots: 90 days from shutdown to a hard pivot
The problem. After roughly seven years, StackBlitz's browser IDE business had stalled at about $0.7M ARR, and the company was weeks away from running out of money. The idea it wanted (text-to-app) was not new; the underlying models had never been good enough to make it real, so the team had to stay alive without a working product-market fit.
The approach. Rather than chase a fresh idea, StackBlitz kept a tiny, high-context team (fewer than 20 people) on extremely low burn, then executed a hard pivot: when the enabling model arrived, founder Eric Simons committed the whole company to building Bolt around July 1, 2024, giving themselves roughly a 90-day window to ship or shut down.
How it solved it. The gamble landed. Bolt launched on October 3, 2024, and by day two was adding about $80K in ARR, already matching what StackBlitz had built in seven years. Surviving to the exact moment the market arrived, rather than dying with a good idea, was the entire pivot.
Disruptive Innovation: timing the model curve
The problem. "Text to working app" had been demoed for years but never worked as a product: AI-generated code was unreliable and not production-ready, so the concept stayed a party trick rather than a business.
The approach. StackBlitz treated the most important variable as the model curve, not the UI. Simons got early access to Anthropic's Claude 3.5 Sonnet in June 2024, judged its agentic coding to be a leap over anything prior, and built Bolt directly on top of it as soon as the model made reliable code generation possible.
How it solved it. Claude 3.5 Sonnet converted text-to-app from a demo into a business overnight: a single prompt could now return a running full-stack app good enough to pay for. Because the surrounding product already existed, StackBlitz shipped within months of the model landing and hit roughly $4M ARR in the first four weeks.
Moats: seven years of WebContainer
The problem. A viral AI wrapper is easy to clone. Competitors could point the same models at the same problem, and most AI coding tools depended on spinning up cloud infrastructure per user session, which is slow, costly, and undifferentiated.
The approach. Bolt runs on WebContainer, a WebAssembly-based operating system that boots a full Node.js dev environment entirely inside the browser, which StackBlitz had spent roughly seven years building. The whole development environment runs client-side rather than in per-session cloud VMs.
How it solved it. That deep-tech foundation gave Bolt a faster, cheaper, more reliable experience that rivals could not quickly replicate, since seven years of low-level runtime work cannot be shortcut. WebContainer turned out to be the ideal companion for LLM-generated code: the model writes an app and it runs instantly in the same browser tab, no infrastructure to provision.
GTM: one tweet, 60,000 signups
The problem. A team of under 20, with fewer than 10 people in go-to-market roles, had no budget or headcount for a paid acquisition machine, yet needed distribution fast to justify the pivot.
The approach. Bolt launched with a single X/Twitter post (a thread the team spent about 12 hours crafting) that showed the product doing something immediately impressive: turning a one-line prompt into a working app. The output itself was the marketing, primed to be shared by a founder-and-developer audience.
How it solved it. Over 60,000 people signed up on day one, and Simons described what followed as "true virality across channels," spreading through Twitter, Reddit, YouTube, and TikTok via creator demos. With near-zero paid acquisition, Bolt scaled to roughly 1M monthly active users and over 3M registered users on the way to about $40M ARR by March 2025.