06. Four Million in Four weeks : Bolt AI

wo weeks ago, I came to know the existence of another new tool called Bolt by bolt.new while talking with Niresh Nagarajan and how it's changing the pace of ideas in developing software. We meet every weekend over coffee to discuss what transpired in the world of AI because we truly live in an era where decades happen in days.

In our previous edition of the AI Brains Newsletter, we dissected how Cursor made $100 Million in less than a year, becoming the first ever company in business history to achieve this feat. In Bolt's case the numbers are large.

While Bolt was launched in September 2024, its parent company StackBlitz, had been striving to make the modern browser more powerful since 2019 and introduced the concept of WebContainers in 2021. You can read the launch post from StackBlitz's CEO Eric Simons here, At a high level, WebContainers lets you run full Node.js environments entirely inside your web browser — no backend servers or local setup needed. They work by compiling the Node.js runtime to WebAssembly, enabling things like npm install, file system operations, and even running web servers, all within the browser sandbox. For non-techies, WebContainers are like being able to cook a full meal right at your table -- no kitchen setup needed. Everything you need is ready to go instantly, right in your browser.

Bolt’s software stack marries cutting-edge tech with pragmatic choices. On the front-end, it uses TypeScript and the StackBlitz IDE framework for the in-browser experience. On the back-end, interestingly, Ruby on Rails powers core services (user accounts, permissions, billing, etc.) Powered by Anthropic's Claude 3.5's language-understanding and code generation strength, Bolt can take a high-level request and produce a working full-stack project. The AI is “instrumented and integrated at every layer of WebContainers,” meaning Bolt’s system instruments the runtime to catch errors and feed that back into the model for fixes

In-browser execution dramatically reduces cloud costs – instead of paying for cloud GPU time or VM containers for each user’s session, Bolt offloads execution to the user’s own device via WebContainers. This gives it a cost advantage and high margins, as it doesn’t need to run a server for every user action. It does impose some technical limitations: the browser sandbox restricts certain operations but for the vast majority of web app scenarios, the in-browser OS approach works brilliantly.

Startups can fast-track development of their MVPs – with Bolt as an initial engineering team. Bolt’s creators even set up a “Bolt 100K Open Source Fund” to support open-source projects, further encouraging real-world usage and ecosystem growth. While we hear a lot of noise and feedback in the internet, speed and accessibility are the recurring themes.What used to take weeks of engineering effort can now be achieved in a coffee break by someone with no coding background. This has opened the door to a much more diverse set of creators building software to address their own needs.

In Anthropic's case study on Bolt, they mention how a user built an MVP for $50 when she was originally quoted $5000 by an agency at first. Agencies offering software development will have to take a cut on their "Design and prototyping charges" in the future with the advancement of these high-end tools.

Alternately, the same agencies can also benefit by building UI ideas without waiting on engineering [Given their prospect is ignorant of the trend]. Designers who cannot code before can now create functional prototypes and test hypotheses much faster and cheaper. While agencies can get smaller, smarter, efficient and still benefit from these tools, there are massive categories like Low-Code & No-Code tools worth Billions in TAM that are projected to bear the brunt of this shift, why learn a specialised no-code tool or drag-and-drop interface when you can just tell an AI what you want in plain language?

Bolt has achieved extraordinary early growth since its launch. Within two months of release, Bolt reached an annual recurring revenue (ARR) run-rate of about $20 million, rising from $0 to $4M ARR in just four weeks and continuing to climb explosively. Sacra’s analysis estimates Bolt hitting $40M ARR by March 2025 with over a million users signing up in the first couple of months

[ Note : Going forward, When I write about specific tools, I am also talking to people who use these tools and collect their feedback to present an unbiased coverage ]

While researching, I signed up to build simple Scheduling software as a test case and Bolt spun up a clean scheduler app in less than few minutes. It led the way and asked me to sign up with Supabase and connect for DB management and also to another tool called Netlify, which will host the entire application and help go live. This entire transition of building a product from idea in less than 30 minutes made me feel like a genius level developer. At a closer look, we discover that some flows are still not well defined and needs to be re-prompted well for better results. I eventually had to call a developer friend to get it sorted. By then my credits expired and I had to abandon the feat.

I created a post in LinkedIn last week to speak to other folks who've been using Bolt and got some people to respond. [You can also see the comments in the post]

Based on reviews and shared comments, users see Bolt as a fast tool for building frontend features and shipping MVPs or proof-of-concepts. Users appreciate quick deployment, decent coding standards, and its ability to create functional apps in as little as 2 days, which would typically require large dev teams across multiple environments. The use of Test-Driven Development (TDD), iterative architecture planning, and context-preserving files helps users manage AI output more predictably, even with weaker models. However, the platform faces challenges when scaling to larger, more complex projects, as code reliability and consistency tend to decline. It also requires careful prompt management and ongoing supervision, which can impact efficiency and increase costs as project scope expands. Inference:

Bolt excels in structured, small-to-mid-sized projects where tight feedback loops and good prompting discipline are followed, but is not yet ready for complex, large-scale production systems. Maybe we'll get better over time as users share the best practices and approaches that can accelerate it's usage for the better. Got an experience with you that you'd like to share? Please leave a comment.

On a closing note, these platforms are not just enabling software development—they're democratising it. By creating a shared space where developers, product managers, non-technical professionals, and even casual creators can collaborate, they’re redefining who gets to build. For the better part of the last two decades, software development has been perceived as a form of high-level wizardry—reserved for those with deep technical expertise. But this paradigm is beginning to shift. Over the next five years, that barrier may erode significantly - I am predicting 5 years.

If anyone can build software that actually works, the fundamental question of what software should cost will be up for debate.

While these tools are still in their early days, it’s hard to see a durable moat for many traditional software categories in the long run.