Cactus Gaming

Design straight in code, with AI

Designers left Figma and started building directly in the product repository, with AI and support from the developers, taking the work all the way to production.

Role
Product Design Manager
Period
Apr 2026 – Oct 2026
Team
Up to 5 designers
Stack
Claude Code · Cursor · React · TypeScript · Vite · CSS Modules · Next.js
3+ weeks → 2 daysOne feature estimated at 3+ weeks (design, product and dev) delivered in 2 days

01Context

A regulated iGaming platform: the player platform and the operators' backoffice.

02How the process used to work

Design happened in Figma, outside the product repository. A feature went through design, product and dev, and one of them was estimated at 3+ weeks.

03Design in code, with AI

Designers started building directly in the product repository, with Claude Code and Cursor and support from the developers, taking the work all the way to production.

API contracts defined with the back end. Pull requests on GitHub, reviewed by the front-end tech lead.

AI workflow inside the repository

04My role and the developers' role

I led the design process and the design team's rules. I created Cursor and Claude Code rules and skills, a review agent that runs before the pull request, documentation automations and a checkpoint skill that shares the task context across the designers' agents.

The front-end tech lead set up the quality gates, the CI, the repository rules and the mock safeguards.

05Result

One feature estimated at 3+ weeks (design, product and dev) was delivered in 2 days of work.

Two prototypes also came out of it, not in production: Cato, an assistant that answers based on the platform's context, and a guided onboarding where the AI performs the task on screen alongside the user.

06Learnings

AI writes, machines verify, people decide. And the gain was not immediate: at first, a lot still took a long time.

Next case · Dito CRMEngagement, design time and leadership →