From design to production, in two weeks

How to ship the new geev.com landing page with no developer available, on a complex codebase?

Geev2023-26AI & delivery
Role
Design, integration, production release, ownership of the topic
Tools
Figma, Cursor, Figma MCP, Claude
Platform
Web
Measured result: In production in 2 weeks

In short

The problem. The new landing page had to launch alongside the brand platform, but integrating it in Angular would have taken too long, and no developer was available.

What I did. Designing and integrating the page myself with Cursor and the Figma MCP, from the first commit to production, with the CTO vouching for code quality.

Result. A landing page shipped in two weeks, design and integration included, and a method reused afterwards to evolve the Geev web app through pull requests.

Context

The redesign of Geev’s homepage accompanied the launch of a new brand platform, created with the agency 4uatre: more colorful, more symbolic, made of stickers, hand-drawn lines and vibrant colors that evoke both the objects and the human connections at the heart of the app.

The page had to launch with the new brand. But integrating it into the website, built on Angular, was too long a project, and the development team was busy with the app.

The problem

A classic situation: the design is ready, or almost, but nobody is available to implement it. We had two options: wait for a developer to free up, or scale down the page’s ambition so that it would fit in the time available.

Neither was right. The landing page was the first showcase of the new brand, and postponing it meant postponing the brand along with it.

Design

I designed the page by iterating directly with the PM and the CEO. To translate the brand platform into a web page, I drew inspiration from landing pages following the neo-brutalist movement: bold geometric shapes, vivid colors, clickable elements underlined by strong outlines.

The main line running through the page illustrates giving and the circulation of objects between users. The pictograms feature bold, colorful outlines, and the headline, heavy and massive, clearly stands out from the rest.

Shipping it myself

Once the design was approved, we decided that I would integrate it myself, with Cursor. I was given access to the codebase and took charge of the topic end to end: integration, verification, production release.

The Figma MCP as a bridge. Rather than describing the mockup to the AI, Cursor read the Figma file directly through the MCP. Among the models tested, Claude gave the best results: relevant implementation choices, and the most faithful to the mockup.

A shared technical framework. I did not bypass the tech team. The CTO took charge of integrating the new code into the existing architecture, which left me in control of the rendering while keeping the codebase healthy.

Knowing what to change, and where. AI produces fast, but it can’t always judge what is visually right. That’s where my dual skill set mattered: my CSS integration experience let me find my way around the code, and my designer’s eye told me what to fix. Without one or the other, the exercise would have fallen short.

From the Figma mockup to code, via Cursor and the Figma MCP.
2 weeks
From design to production
Design, integration and testing included.
See the result

What came next

We reused this process to make occasional improvements to the Geev web app: interface fixes that probably wouldn’t have been prioritized in the backlog came directly under my responsibility.

What I take away from it

AI moves the boundary of the designer’s role. On well-defined interface topics, the designer can go all the way to production, and free up the tech team for what truly requires its expertise.

More work