Claude Desktop is a chat list. Open it, pick a conversation, start a new one. That model made sense when Claude was a chatbot. It makes less sense now that Claude runs agents, manages projects, and does work while you’re not watching.
I started keeping a list of things that bothered me about the interface. That list turned into a Figma mockup. The mockup turned into a working prototype, React, Vite, live API, Anthropic’s actual brand tokens. Built over a weekend because I couldn’t stop thinking about it.
The Substack piece is the write-up of what I learned: where the friction is, what the interface gets wrong structurally, and what a more agentic-first version of Claude Desktop might look like.
If you want to see what I actually built, the full case study with images and video is on the Work Ben page: Claude Desktop Redesign →