I design with Claude more than Figma now

This Jane Street blog post, written by designer Edwin Morris, makes the core claim that he now uses Claude for design more often than he uses Figma. The author was not an AI optimist to begin with; in the past he used Copilot and Cursor to modify games, and tried Gemini for producing product decks and wireframes, but the results were often worse than doing it by hand. The turning point came after he joined Jane Street, where, facing an unfamiliar technical environment of OCaml, Bonsai, and the like, AI support became extremely useful and unexpectedly reshaped the design work he knew best. His old workflow was to write specs, make Figma mockups, pitch, and then iterate on implementation back and forth with engineers; now he first describes the problem and solution, opens an editor, starts the build and server, and lets Claude build a working prototype in the real codebase based on his description. This lets him quickly validate feasibility and repeatedly fine-tune buttons, keyboard shortcuts, copy, prompts, and confirmation messages, and hand over functioning features in the development environment for users to try. The author argues that a real prototype can carry design judgment better than a static mockup, because all the effort goes into an actual artifact rather than being consumed on intermediary documents or Figma components. However, the article also points out a risk: when a designer hands over a feature that looks finished, the engineer reviewer may be left with only a code review rather than jointly exploring the design space. Jane Street's current practice is to explicitly mark these prototypes as "living proposal documents": the code is disposable, the review focuses on the product and user experience, and engineers ultimately take over to implement the official version. The author also worries that iterating with Claude over the long term might confine him to what the tool can produce, leaving less room for freer, divergent creativity. Overall, this is not a tool release or technical tutorial, but a practical reflection on how AI coding tools can bring designers back closer to the product medium and shift the boundaries of collaboration between design and engineering.