The orchestration platform
According to a Figma blog post on performance improvements, "it's common to see files with tens of thousands of layers." That infrastructure strain—the grinding collision between what the tool was built to handle and what teams now demand—is the signal buried in Config 2026's announcement cascade. The layers panel rebuild represented a significant rearchitecting of load-bearing infrastructure, required because the platform was experiencing sluggish interactions in very large files.
This week documents the platform infrastructure that enables orchestration—designers becoming curators of generative output rather than originators of every pixel. The company appears to be building the orchestration environment itself, and Config 2026 makes that explicit: the canvas now absorbs code editing, motion timelines, shader programming, generative plugin creation, and node-based visual workflows, all unified under Dylan Field's philosophical claim that "code is material, just like images, vectors and design layers."
If code, motion, shaders, and generative workflows are all materials to be manipulated on the canvas, then the canvas is no longer a design environment. It's a universal workspace.
That framing—code as material, not separate discipline—justifies the entire architectural pivot, but it also quietly redefines what Figma is. If code, motion, shaders, and generative workflows are all materials to be manipulated on the canvas, then the canvas is no longer a design environment. It's a universal workspace. The expansion suggests the platform now positions itself against not just design tools but with VSCode, After Effects, Runway, and the entire fragmented toolchain that currently spans from initial concept to shipped product.
The strategic logic is gravitational. Every workflow Figma absorbs creates friction for leaving. Code layers let you "turn any design layer into an interactive code layer with just a single click," then extract it back to editable design layers, then update the code layer with changes—a seamless loop that only works if both ends live in Figma. Motion is "built on Figma's platform," so animated components travel across files "the same way fills and typography do," and motion variables update "every animation referencing that variable" at once when you switch modes. The new Motion timeline is "visible and inspectable" in Dev Mode, with "every timing value, every easing curve, every keyframe" readable, and you can "copy animation code directly in CSS, JSON or framework-ready React." Motion is also "MCP-compatible, so you can pass any animated frame directly to a coding agent for implementation."
Each announcement collapses a previously external step. Shader fills and effects, traditionally requiring specialized programming knowledge, are now described to the Figma agent or generated from reference images, "parameterized by default" with controls "directly on the canvas." Generative plugins let you "describe the tool you need: the behavior, the controls and the parameters. No local dev environment or plugin API knowledge required." Weave tools—node-based generative workflows for image and video—now live in the Figma Design left panel as "pre-built Weave workflow, packaged into a simple UI," while complex workflows can be published to Figma Community as reusable templates. Make connects to your production codebase, allowing "direct, visual edits" that "ship changes from Make," with branching, commits, and pull requests all inside Figma.
This isn't feature expansion. It's discipline absorption. And the bet only pays off if teams adopt the full stack. The risk is that partial adoption—using Motion but not code layers, or shaders but not Weave tools—could create a bloated interface with unused surfaces. The layers panel performance work may suggest challenges in handling the compound load of all these materials coexisting. The new architecture uses "two-pass computation" and "caching derived data" through a "platform-level primitive" called derived properties.
The workflow integration functions as a retention mechanism, though the mechanisms are explicit. Animated components carry motion "across every screen and every collaborator's file." Motion variables propagate changes universally. Code layers sync bidirectionally with design layers. Weave workflows connect to Figma nodes, generating "a localized version for every region" by feeding a single brand layout through a CSV of translated copy. The Figma MCP server "brings your design decisions into the tools where code gets written," allowing agents in other environments to read Figma files and write back to them. Every integration point strengthens platform cohesion.
Maxwell Hathaway, Lead Motion Designer at Atlassian, articulates the deeper claim: "People get to see how things move really quickly, and that access is going to help them develop taste faster." This isn't about productivity. It's the argument that collapsing workflows accelerates taste development itself—that removing friction between disciplines doesn't just save time but fundamentally changes how designers learn to see. If true, the separation between design, motion, and code wasn't just inconvenient but pedagogically damaging. The counterargument would be that the platform risks conflating speed with depth, mistaking rapid iteration for the slow accumulation of craft judgment.
The tension surfaces in how each absorbed discipline gets simplified. Motion designers "make things feel alive in ways that are hard to explain, let alone replicate," but Figma Motion offers "preset animation styles—fade, move, scale—for the quickest way in," then lets you "continue refining on the canvas." Shaders, which "have always existed in Figma, powering our canvas rendering behind the scenes" but were inaccessible to most users, are now generated by describing what you want or using "an image as a reference." Generative plugins eliminate the need for "local dev environment or plugin API knowledge." Each time, the expertise barrier drops, the floor lowers, but the ceiling—the claim Field makes that "AI has lowered the floor, it has not raised the ceiling"—becomes harder to verify.
Because what does raising the ceiling look like when the tool handles the technical implementation? Anna Zhang, a creative technologist, built shaders "to remix her own photography with collage, marbling, light leak, metal emboss, and prism effects," describing the process as "a back-and-forth dialogue between me and the agent, and what it was outputting would shape the parameters I would add down the line. It was like a negotiation." That negotiation—between intent and constraint, between what you imagine and what the system can do—is where craft happens. But when the constraint is an agent's interpretation rather than the material's inherent resistance, the negotiation shifts. You're shaping the agent's understanding, not wrestling with shader code.
Field's framing again: "The material was separated from the process." The implicit argument is that tools "forced a choice" between design and code, creating artificial boundaries. But those boundaries also created specialization depth. Motion designers developed taste through years of keyframe work in dedicated tools. Shader programmers understood WebGPU because they wrote shaders from scratch. The separation wasn't just workflow friction—it was the structure that made expertise possible.
Figma's bet is that unification creates a new kind of expertise: fluency across materials, comfort moving between design layers and code layers and motion timelines and generative workflows, judgment about which tool fits which moment. The orchestration skill set. But orchestration expertise is only valuable if the platform maintains depth in each discipline. If Motion becomes "animation presets plus agent refinement," if shaders become "describe what you want and the agent builds it," if code layers become "prompt for what you need and extract to design when you want to tweak," then what Figma offers isn't multi-disciplinary mastery but platform dependency—the ability to operate Figma's specific implementations of motion, shaders, and code, not transferable craft knowledge.
One reading of these announcements suggests Figma is no longer a design tool that added capabilities but has become an orchestration platform where design is one material among many. That distinction matters because tools and platforms optimize for different things. Tools optimize for craft depth within a domain. Platforms optimize for retention and cross-surface engagement. A tool becomes better by doing one thing exceptionally well. A platform becomes better by making it harder to leave.
Every Config 2026 announcement pushes toward platform. Code layers that sync with design layers. Motion that propagates through component systems. Shaders that expose parameters as canvas controls. Generative plugins that "look and feel native to Figma." Weave workflows that connect to Figma nodes. MCP integration that lets external agents read and write Figma files. The architecture is explicitly designed for lock-in, though the language is "workflow unification" and "removing barriers."
The Atlassian case is illustrative. Alexandra Pereira, Senior Product Designer, says Motion "turns animated illustrations from a specialist handoff into a system capability," while Davy Fung, Lead Product Designer, describes prompting the agent for suggestions "on what was working, and what wasn't." The value proposition is democratization—motion becomes accessible to everyone, not just motion specialists. The tradeoff is that motion becomes a system capability, governed by Figma's motion architecture, rather than a specialized craft with its own tools and standards.
Figma is building the platform where AI-era design happens: where agents generate options, where designers orchestrate materials, where code and design and motion collapse into a unified canvas. That platform is enormously valuable if teams adopt it fully. It creates a single source of truth, eliminates handoff friction, accelerates iteration, and makes previously inaccessible capabilities available to everyone.
But the risk—the architectural tension Field's "code is material" framing doesn't address—is that a tool trying to be everything becomes excellent at nothing. The focused craft excellence that made Figma dominant was interface design: multiplayer editing, component systems, vector precision, design-to-dev handoff. The question is whether expanding into multiple disciplines dilutes that focus. The layers panel performance work is suggestive: infrastructure built for design files may face new challenges as files grow to include code, motion timelines, shader programs, and generative workflows.
The question isn't whether Figma can build these capabilities. Config 2026 demonstrates they can. The question is whether the unified canvas remains coherent when it contains this many materials, or whether it fractures into specialized surfaces that happen to share a file format. Whether teams experience it as seamless orchestration or navigational overload. Whether the platform bet pays off in indispensability or collapses under its own weight. Figma has built the infrastructure for AI-era orchestration. Now they have to prove the canvas can hold it.