Overview

We conducted an internal evaluation of Figma Make (powered by Claude 4 Sonnet) to determine its viability as a prototyping and code generation tool for the Easywaste platform. The test covered both unstructured mockups and structured design system-based frames.

Objectives

• Import “quick & dirty” Figma frames to test visual output generation.
• Use auto-layout and design system frames to test fidelity and accuracy.
• Assess efficiency of the design iteration workflow.
• Evaluate code export quality and V0 compatibility.
• Determine potential for production-level prototyping in our pipeline.

Key Findings

1. Quick & Dirty Frame Import

• Poor results: Figma Make failed to interpret rough, non-auto layout frames.
• Outputs were structurally incoherent and lacked visual fidelity.
• Design consistency could not be maintained despite repeated prompting.

2. Design System Frame Import

• Even clean, auto-layout frames using our design system yielded misaligned results.
• Layouts, hierarchy, and spatial logic were misread or ignored.

3. Workflow & Efficiency

• Modifying components was slow and required excessive back-and-forth prompting.
• Response time from the AI was inconsistent.
• One standard pop-up component (location filter) took 3–4 hours to get right.

4. Code Export (V0)

• Once visually finalized, code output for V0 was mostly accurate.
• The main issue: incorrect font weights.
• A quick review by our developer (Tuomas) found the structure acceptable.

5. Interaction Design

• Strong performance on interactivity:
  • Hover states
  • Button logic
  • Pop-ups and modal animations
•  
• Easy and intuitive to build interaction flows that matched design intent.

Conclusion

Figma Make is not ready to serve as our primary prototyping tool. While its interaction design capabilities are promising, the inability to consistently translate both rough and structured design input into coherent visual output makes it unfit for production use in our Easywaste workflow. Design fidelity demands too much micromanagement, resulting in inefficiency and friction.

We will continue to monitor progress but cannot recommend adoption at this stage.