Bolt turns prompts into running apps

Bolt mattered because it turned AI coding from a cloud service into an instant browser product. Claude 3.5 Sonnet made one prompt good enough to produce a usable first draft of an app, and WebContainers let that app actually run, install packages, and serve pages inside the same tab. That combination removed the usual handoff between chat output and a real dev environment, which is why Bolt could feel more like making software than copying code.1 2 3 4 5

• WebContainers are the technical half of the story. They run Node.js in the browser, which means Bolt can open a terminal, install dependencies, start a local server, and preview a full stack app without paying for a separate cloud container for every session. That makes fast iteration cheaper and simpler to deliver.2 3 4
• Claude 3.5 Sonnet was the model half. Once zero shot code generation got reliable enough in June 2024, Bolt no longer needed a user to stitch together partial snippets. The product could ask for an app in plain English and get back a coherent enough project to run immediately in the browser.1 5 6
• This also shaped the competitive split. Bolt and Lovable became prompt first app generators for non engineers, while tools like Cursor picked up after the app existed and helped edit codebases. In practice, Bolt wins on instant generation and execution, while IDE agents win when a human wants to keep refining the repo over time.5 6 7

The next step is for browser native app builders to add more of the software lifecycle around this core loop, from design import to mobile support to enterprise deployment. As models improve and browser runtimes mature, the advantage shifts toward products that can turn a prompt into a running app with the fewest handoffs, lowest infrastructure cost, and cleanest path into production.4 5 7