Bolt.new
2025
After months of collaboration, StackBlitz asked me to pair with a developer on an R&D question: what happens if you put AI models inside StackBlitz, their in-browser IDE? The only product somewhat in that space was Vercel's v0, but at the time it was still a fairly basic UI component generator. So everything had to be built from scratch.
There were two of us, and we were the entire product team for the first weeks. Regular check-ins with Eric Simons, otherwise full autonomy. I owned design and art direction, including the logo and the product's visual identity.
Approach
At first, the project was meant to be a StackBlitz extension for developers, a bit like Cursor and Windsurf, which were just starting to emerge. That premise fell apart from the first demos. The AI was already powerful enough that people who couldn't code could create what they wanted, simply and online. The "developers" framing was leaving most of the product's value on the table.
That audience shift mid-course is what shaped the product. It needed an interface accessible to the widest audience, without the technical machinery disappearing. Advanced users still had to be able to go as far as on a code editor, the way products like Framer allow.
Watching AI develop from the inside was a wild experience. So much so that I decided to work on Bolt.new exclusively for almost all of 2025, watching the team grow from 2 to nearly 50 people. I designed a great many features and pages there, joined along the way by my colleague Raphaël Regnier.
Outcome
Bolt shipped three months after the R&D started. StackBlitz went from $50k to $20M ARR in the two months that followed, then raised $100M in Series B. The product continues to thrive today.
