Julien Thibeaut

Thibeaut’s portfolio connects his independent projects with client work through Interface Office. He describes paying attention to interface details until the experience feels right, and names AI startups among the studio’s clients.[1]

Make interface knowledge available to an agent

UI Skills provides a registry of design-engineering instructions. A command-line interface lets a developer browse categories, list skills and fetch a particular entry. Agents can also query the registry through its Model Context Protocol endpoint. A companion playbook distills lessons from the collection. The project organizes access to those instructions rather than replacing design judgment.[2]

Build AI interactions from reusable components

Prompt-kit supplies customizable interface components for AI applications. It builds on shadcn/ui, allowing components such as a prompt input to be installed into a project and imported from the application’s own code. Motion-primitives provides a related set of animated interface components built with Motion and Tailwind CSS.[3][4]

In Anatomy of AI Input, Thibeaut explores how the surface for writing a prompt has grown beyond a text box. Attachments, model selection and other controls share the same interaction space. He recommends keeping typing available during generation, avoiding layout jumps and preserving keyboard access. These details let a user revise or stop work while a model responds.[5]