The Gap in Enterprise Design System Platforms

The large design systems published by the hyperscalers — Carbon, Spectrum, Material, Polaris — are component libraries, not platforms. They ship CSS variables, web components, and documentation sites built for the publisher's own product surface. None of them governs, versions, or publishes a customer's tokens from inside the customer's own perimeter. The management layer that does that work — Supernova, zeroheight, Knapsack, Specify, Figma Variables — is almost entirely cloud SaaS, US-hosted, and structurally disqualified for regulated-industry buyers in financial services, legal, and government who cannot send brand intellectual property to an external host. The single open-source, air-gappable, standards-native option is Penpot, and Penpot is positioned as a design tool, not a publishing substrate. The concrete gaps for a regulated small-to-mid-size team are therefore: no self-hosted token governance, no on-premises documentation publishing, no way to keep brand IP and AI inference inside the firewall, and no single source that drives more than one output surface. The format is now open; the platform that manages and publishes from it is not.

The Token-as-Content-Schema Thesis

Treat the design-token graph as a content schema, not merely as styling configuration. One token source then drives three outputs that are managed separately everywhere today: UI components, marketing pages, and regulated documents. In current practice these live in silos — a design system handles components, a separate headless CMS handles pages, and document templates are maintained by hand. The "design system as CMS" idea is articulated in the literature (Sanity, Adobe Experience Manager Universal Editor, CMS.gov) but no product ships a single token graph that simultaneously generates components, landing pages, and legal documents. The difference is structural: when typography, color, spacing, and elevation are defined once as portable tokens and every surface consumes only those tokens — never literals — a brand or compliance change propagates everywhere from one edit. The marketing page and the contract template inherit the same heading scale and the same brand color as the product button, because they read the same graph. That is the leapfrog seam: not a better component library, but a unified content schema that collapses three management stacks into one.

The 2030 Landscape

The standard war ended in October 2025 when the W3C Design Tokens Community Group shipped its first stable specification, with reference implementations across Style Dictionary, Tokens Studio, Terrazzo, Penpot, Figma, and Sketch. Tokens are now a portable, vendor-neutral interchange format rather than a proprietary asset, which collapses the historical lock-in moat. Value migrates up the stack toward automation, governance, generation, and publishing — exactly the layers where, per the 2026 zeroheight Design Systems Report, only 40% of teams have any pipeline automation and 60% still sync tokens by hand. Carbon, Spectrum, and Material will continue to lead on component breadth and accessibility maturity, and will remain the safe default for teams already inside those ecosystems. A sovereign alternative does not win on component count; it wins on three claims the incumbents cannot match at once: self-hosting and air-gap capability, customer-held IP and keys, and a standards-native workflow that runs authoring, governance, and multi-target publishing in one open graph. The flank to watch is Penpot extending upward into publishing — the nearest credible threat to this position.

Marketing Page Integration

A marketing surface at design.pointsav.com is the proof that the token-as-schema thesis is real, because a marketing page is the output furthest from the design system yet still reads the same graph. The pipeline is deliberately small: token sets plus a resolver file (theme and brand contexts) feed a build step that flattens, resolves aliases, and emits CSS custom properties per context; page content is authored as front-matter plus a block sequence (hero, feature grid, pricing, call-to-action, logo wall); a closed set of block templates consumes only those custom properties; and the renderer binds content into blocks and injects the compiled variables. The author surface exposes exactly three safe choices — which block, which token-defined variation, and the content — so a non-coder can build a page without ever touching styling, because styling is structurally inaccessible. Built in Rust and Axum with no external CMS, this enables a regulated buyer to publish on-brand marketing pages from the same governed token source that drives their product UI, entirely inside their own perimeter, with a brand change propagating to every page automatically.

Recommended Actions for project-marketing

1. Anchor the marketing platform narrative on the token-as-content-schema thesis — one governed token graph driving UI, marketing pages, and documents — as the differentiator no cloud incumbent can match while staying cloud.
2. Lead positioning with sovereignty for regulated buyers (self-hosted, customer-held keys, on-premises AI) rather than competing on component breadth against Carbon or Material.
3. Reuse the established chassis split (shell library plus serving binary) so the marketing surface and the design-system docs surface converge on one pattern rather than two implementations.
4. Frame the open standard as commodity and the management-and-publishing platform as the product, citing the 40% automation / 60% manual-sync gap as the addressable problem.
5. Watch Penpot's move into publishing as the nearest competitive threat and pre-empt it in messaging.