Figma Config 2026: Bridging the Gap Between AI Generation and Human Control

At the Config 2026 conference, Figma unveiled a massive suite of updates designed to transform the design canvas into a multidimensional workspace for code, motion, and 3D effects. Rather than competing to build foundational LLMs, Figma is focusing on deep integration, ensuring that AI serves as a granular tool for professional designers rather than a black-box replacement.

Beyond Prompting: Integrating Code, Motion, and Shaders

Figma’s core strategy is to bring AI generation directly onto the canvas, preventing users from jumping to external "vibecoding" tools. The most significant technical leap is the introduction of Code Layers, which allows design and production code to live side-by-side. Users can now pull in production code via GitHub, manipulate it through an agent, and convert design tweaks back into code seamlessly.

The platform is also breaking the boundaries of static 2D design with three key technical additions:

  • Motion & Timelines: Designers can now manage animations and transitions within a shared timeline, which can be pushed to production via Dev Mode and MCP.
  • 3D Transformations: A new depth layer allows for true spatial perspective, moving beyond the traditional method of using stacked shadows to fake depth.
  • WebGPU-Powered Shaders: Using WebGPU, Figma has introduced a shader feature that allows users to generate complex visual effects—such as dithering, frosted glass, or polished chrome—simply by describing them to an AI agent.

Weave and the Shift Toward "AI Materials"

Building on its acquisition of Weave, Figma is integrating what it calls "AI materials" into the design workflow. This system allows teams to combine multiple models and image sources into a unified design direction. Currently, over 20 Weave tools, such as the Texturize plugin, are available on the canvas, enabling designers to apply procedural textures and patterns to elements through natural language prompts.

To solve the problem of fragmented AI workflows, Figma is introducing Agent Skills. This allows teams to save successful prompting techniques—such as a /contrast-improvements command—and deploy them across the entire organization. This turns individual AI "wins" into reusable, searchable team assets.

The Economics of Rented Intelligence

While Figma’s feature set is expanding, the company faces a significant structural challenge: its reliance on "rented" intelligence. Figma’s AI capabilities are powered by third-party models from OpenAI, Google, and Anthropic. This dependency has a tangible impact on the bottom line; Figma’s gross margins fell from approximately 92% to 86% during 2025 due to rising AI inference costs.

Furthermore, this creates a strategic paradox. One of Figma’s primary model providers, Anthropic, has launched Claude Design, a tool capable of generating polished, clickable interfaces from a single prompt. This directly threatens Figma’s position as the starting point for digital product development. By focusing on granular, editable layers (Code, Motion, and Shaders), Figma is betting that professional designers will always prefer a tool they can control over a model that merely "spits out" a finished result.

Key Takeaways

  • Deep Integration over Pure Generation: Figma is countering "one-shot" AI tools by adding code, motion, and 3D shaders directly to the canvas for iterative control.
  • Collaborative AI Workflows: Through Agent Skills and Weave, Figma is transforming AI from a solo productivity booster into a shared, searchable team asset.
  • The Margin Challenge: High inference costs from third-party LLM providers are impacting gross margins, even as the company pushes for deeper AI adoption.