Figma-to-code: the real shift happening

Short summary

Design-to-code tooling is consolidating around Figma, DTCG standards, and AI agents (Claude Code, Cursor, Copilot), but the real bottleneck isn't tools—it's the 'missing middle': models lack semantic understanding of design tokens and intent, so they guess at deterministic decisions. monday.com solved this through proper scaffolding, but the industry hasn't standardized this translation layer. The split between deterministic architectural requirements and genuine judgment calls is where code quality breaks down.

• •Design-to-code pipeline converging on Figma → MCP → AI agents, but semantic token understanding remains weak
• •Models guess at deterministic decisions because Figma files lack typed meaning; the 'missing middle' is unowned across the industry
• •monday.com's pipeline shows scaffolding can close the gap, but broader standardization of deterministic layer is needed

Generated with AI, which can make mistakes.