I design with Claude more than Figma now

A designer at Jane Street now primarily uses Claude AI instead of Figma for creating prototypes and implementing features, marking a significant change in workflow driven by recent AI advancements.

The designer, who previously relied on Figma for mockups and design documentation, now uses Claude AI to build functional prototypes directly within the codebase. This approach allows for rapid iteration, with the AI generating working features based on brief descriptions. For example, a recent prototype added LLM prompting to a JSQL input, which was tested and refined over several days with unlimited AI iterations. The shift has reduced the time spent on ancillary tasks like creating Figma components and documentation, enabling more focus on actual feature development.

Over the past two months, the designer’s use of AI has expanded from small fixes to larger, more complex prototypes, including features with thousands of lines of code. Some projects now skip Figma entirely, starting with visual design in Claude and iterating directly on the code. This change has empowered the designer to test ideas quickly, such as new data models and user interactions, without waiting for engineering cycles or detailed mockups. The workflow also improves collaboration, as prototypes serve as living proposals that others can interact with and iterate on, although this raises questions about review processes and ownership of final implementations.

Why It Matters

This development signifies a potential shift in design and development workflows, where AI tools like Claude can replace traditional mockup tools like Figma for certain tasks. It could lead to faster product iterations, reduce reliance on detailed documentation, and democratize prototyping by enabling non-engineers to create functional features. However, it also introduces challenges around review, collaboration, and maintaining creative fluidity, especially for novel ideas that may be constrained by current AI capabilities.

Background

Historically, designers have relied on tools like Figma for mockups, while developers build prototypes after detailed specifications. The rise of large language models (LLMs) and AI-powered coding tools has begun to blur these boundaries. At Jane Street, a recent shift has occurred as the designer experimented with AI to generate prototypes directly from descriptions, initially for small tasks, then expanding to larger, more complex features over the past two months. This mirrors broader trends in AI-assisted development, where the line between design and implementation is increasingly fluid.

“Using Claude to make ideas real has made it much easier for others to evaluate them—they can just use it.”

— Jane Street designer

“Prototypes are living proposal docs, the code is disposable, and a reviewer’s job is to give feedback about the design and user experience.”

— Jane Street designer

What Remains Unclear

It is still unclear how widespread this workflow will become across teams or organizations. Questions remain about how review and ownership will adapt when prototypes are generated directly in code, and whether creative exploration might be limited by current AI capabilities. Additionally, the long-term impact on design roles and collaboration practices is still uncertain as the technology and workflows evolve.

What’s Next

Next steps include refining the review process to accommodate AI-generated prototypes, developing best practices for collaboration, and exploring the limits of AI in handling truly novel or complex design tasks. Monitoring how this workflow influences productivity and creative freedom will be key, along with potential adoption by other teams or companies.

Key Questions

Can AI replace Figma entirely for design work?

While AI tools like Claude are increasingly capable of generating prototypes and features directly in code, it is not yet clear if they can fully replace Figma for all design tasks, especially for initial ideation and visual refinement.

How does this change affect collaboration and review?

The workflow treats AI-generated prototypes as living proposals, with reviews focusing on design and user experience rather than static mockups. Reviewers provide feedback on functionality and iteration, with eventual ownership transferred to engineering teams.

What are the limitations of using AI for prototyping?

Current AI capabilities may struggle with highly novel ideas, complex interactions, or creative exploration beyond iterative improvements. There is also concern about constraining creativity and the potential for overlooking innovative concepts outside AI’s current scope.

Source: Hacker News