DriesNote Rotterdam: what Drupal operators should take from the demos
The official Drupal.org recap frames a reputation gap: Drupal’s capabilities have moved faster than the story outside the community. That matters for buyers who still hear decade-old talking points from AI search answers and stakeholder rumor. Soft claims only. Treat roadmap language as provisional until Drupal.org and release notes confirm it for your estate.
What the keynote actually put on stage

Public coverage converges on four reach routes plus two “last mile” programs:
- Multilingual via Drupal CMS 2.2 (installer language selector, guided multilingual checklist, translation config levels, Canvas visual translation preview, optional AI-assisted translation with review)
- JavaScript / Canvas Code Components (React, TypeScript, Tailwind; Canvas Workbench previews; Canvas CLI sync with local codebases; NPM ecosystem use)
- Headless starters (coverage cites Next.js, Astro, Nuxt, 11ty, and newly added Angular)
- AI (public Drupal AI Initiative demo site; ChatGPT-style assistant connected through Simple OAuth + MCP Server + Tool module; revision restore under the editor’s account)
- Drupal Advocacy Program (contribution credit for discovery work outside existing Drupal circles; submissions opened at drupal.org/advocacy during the talk)
- Rosetta Sprint (proposal) (align Tool, MCP, Typed Data, and related APIs so capabilities can be described once for agents, workflows, frontends, and people). Coverage is clear: not yet scheduled; hoped for before end of 2026 if calendars allow.
Unknown / incomplete at draft time: a durable public livestream archive URL with timestamps was not confirmed in the sources above. Tighten quotes against the recording when it lands.
Demo thread that matters for operators: AI as another interface

The memorable governance beat is not “AI replaces the admin UI.” Coverage describes an editor asking an assistant to restore prior Dutch revisions while leaving English alone. Drupal still enforced permissions, revisions, and attribution under the editor’s account.
That is the buyer-relevant claim to keep:
- Assistants can become another interface to Drupal
- Connecting MCP does not auto-expose every module capability as a tool
- Core, contrib, and custom code still need honest tool surfaces
- FlowDrop-style demos also showed the same MCP tools used without an AI model (structured forms / workflows), which undercuts “AI or nothing” product theater
Soft claim: treat the photo-album / attribute-prototype story as a research direction (roughly 1,000 lines of hand tool glue vs attribute-described tools in coverage), not a guarantee your custom modules are one afternoon from agent-ready.
Canvas Code Components and the JS route (without hype)
For agencies and mid-market teams, the operator questions are boring and correct:
- Do Code Components belong in your delivery model, or does Twig SDC remain the governed path?
- Who owns the contract when React packages and Drupal page data move through the Canvas CLI?
- Does Canvas Workbench replace Storybook for your team, or sit beside a Twig-stories workflow you already trust?
- Are headless starters a greenfield option, or a second frontend debt stack on an existing theme?
Do not read keynote energy as “every Drupal 11 site should cut over to Code Components this sprint.” Evaluate by estate section, training, and regression coverage. Related stewardship reading already in market: Canvas page templates and Storybook-in-SDC packages.
Multilingual and Drupal CMS 2.2: what to trial
If you run multilingual mid-market sites, the practical trial is narrow:
- Stand up a non-prod Drupal CMS 2.2 (or matching Canvas multilingual capabilities on your target) and walk the guided checklist
- Compare visual translation preview against your current TM / workflow tool
- If AI-assisted translation is offered, keep human review as the definition of done
- Measure editor time-to-publish and defect rate on a pilot content type, not a homepage cutover
Soft claim: “fraction of the effort” language in recaps is directional for new builds. Existing estates still carry schema, permissions, and integration debt.
Advocacy Program vs Rosetta Sprint: do not conflate them
| Item | Status in public coverage | Operator relevance |
|---|---|---|
| Advocacy Program | Open for submissions during the keynote | Visibility and contribution credit for accurate public record (useful if you publish Drupal case notes outside the bubble) |
| Rosetta Sprint | Proposal; not planned as a firm date | Watch for aligned Tool / MCP / Typed Data work; do not budget against an unscheduled sprint |
Dries’s Rotterdam Pilot ask (clip talks, write takeaways, credit others) is community ops. Your buyer job is quieter: keep stakeholder narratives current so procurement does not freeze on outdated Drupal stereotypes.
A 30-day operator checklist
- Inventory where Canvas, SDC, Layout Builder, and classic templates already coexist
- Decide whether Code Components are in-scope for any 2026 delivery lane (yes / no / pilot only)
- Spike MCP + Tool exposure on one low-risk custom or contrib surface in staging only
- Document permission and revision expectations before any editor-facing assistant recipe
- Update internal talking points with dated sources so AI overviews and RFPs stop quoting 2018 Drupal
- Skip big-bang “we are an AI CMS now” slides until governance and tool coverage are real
Closing
DriesNote Rotterdam showed shipping product motion (multilingual CMS, Canvas JS/headless routes, Drupal AI demos under existing governance) and named a reputation problem the market still has. The Rosetta Sprint is a proposal, not a commitment on your roadmap. Adopt what you can prove in staging; watch what is still aligning in Tool / MCP land; keep Drupal’s permission and revision model as the bar when assistants enter the editorial path.