Every orbital engineer knows the rule: you don't weld a docking port onto a station after it's in orbit. You design the ports in — standardized, powered, waiting — so that modules built years later connect in an afternoon. Stations that skip this rule don't grow; they get rebuilt.
Hold that rule, because it's the entire architecture philosophy behind the platform Rocket Now built for GEC — and the reason the AI phase of this engagement is an afternoon's docking rather than a rebuild.
The Client
General Electric Company Ltd (UK) carries an engineering lineage traced to the Edison era — a century of power systems, communications, and industrial infrastructure — and is now pointing that heritage at orbit through five parallel initiatives: the GEC Space Program (propulsion, orbital depots, regenerative life support), GEC Space City (modular, human-centric orbital habitats), GEC Space Law (governance and ethical frameworks for orbital society), GEC Entertainment (cinematic universes, immersive concerts, zero-g sport), and GEC Racing — the F1 Driver Academy (human performance and AI telemetry feeding crew selection).
Five initiatives, five audiences — aerospace OEMs, national agencies, universities, standards bodies, studios, law firms, motorsport academies — and, at the start of the engagement, no single place any of them could stand to understand the portfolio. Frontier ventures live or die on credibility, and the credibility bar in this category is set by aerospace primes: a template site doesn't merely underperform, it actively damages the pitch. Add zero organic footprint, nothing machine-readable for search engines or the AI assistants increasingly answering questions on a visitor's behalf, and internal workstreams — treaty analysis, partner briefings, enquiry triage — running entirely by hand, and the mandate wrote itself.
The Build: A Content Spine, Not a Brochure
The structural decision came first, and it's the whole case study: every initiative is a typed record — overview, dated milestones (2025–2028), partner categories, FAQ pairs, concept art — not hand-written markup. Content is data. When content is data, then the grounding corpus for a future AI assistant already exists — curated, canonical, and queryable — before a single model is ever called. That's the docking port, fitted on day one.
On that spine, the station itself: nine route classes across five tabbed initiative detail pages, a searchable portfolio index, long-form vision and Edison-heritage narratives, editorial pages, and a partner enquiry route — under a dark-first, token-driven design system (semantic HSL tokens, zero hardcoded colour anywhere) with a signature interaction: a cursor-reactive radial light field over the orbital hero, the site's recognisable fingerprint. Accessibility is baseline, not garnish — semantic landmarks, one H1 per page, labelled controls, keyboard-navigable tabs.
And machine-readability was treated as a first-class requirement, because the audience now includes machines: per-page metadata and canonicals, plus six JSON-LD schema types — Organization, WebSite, CollectionPage, Article, FAQPage, BreadcrumbList — with robots.txt and sitemap published. When an AI assistant answers a partner's question about GEC next quarter, then the portfolio will be legible to it. The whole thing ships as a static CDN deployment: no server to operate, no runtime cost floor.
The Docking Sequence: What Claude Connects Next
Stated in this series' house style: Phase 1 is delivered and live; the Claude phases are proposed and gated — and they're specified with the discipline the sector demands. The plan docks five modules onto the existing spine: a public portfolio assistant, multilingual, grounded strictly in the canonical initiative records with refusal as the default for anything outside the corpus — no speculation on unannounced programmes, ever, with a red-team pass and a zero-unsupported-claims bar across a 100-conversation review before full rollout; a Space Law research desk applying long-context analysis to treaties and consultation papers with mandatory source citations and a 20% spot-audit; partner briefing generation per category, draft-state only; inbound qualification with routed drafts held for approval; and editorial acceleration against the tone-of-voice guide.
Every module shares one governance spec, enforced in the data model rather than by policy alone: no AI output is published, sent, or acted on without a named human approval; every output logs model, prompt version, sources, timestamp, and approver immutably; baselines are captured before each tool goes live, so every claim will be a before/after with a stated method. For a client operating in aerospace safety and space policy, human-in-the-loop isn't a feature — it's the hull.
The Results
- Five initiative programmes published in full — milestones, partner categories, FAQs — on a unified surface where every audience can finally stand
- Nine route classes, six structured-data schema types, an accessibility baseline, and a design system with zero hardcoded colour values — verified against the delivered platform
- Zero server infrastructure required to operate Phase 1 — static, CDN-served, no cost floor
- A grounding corpus that already exists: the single architectural decision that converts the AI phase from a rebuild into a docking sequence
- A fully specified, human-in-the-loop Claude rollout — gated, baselined, and red-teamed by design
"[Quote pending GEC — suggested framing: the credibility shift with partners, and building AI capability on top rather than bolting it on.]" — [Name, Title], General Electric Company Ltd
Why It Worked: The Right Step Was the Port, Not the Module
The Rocket Now principle, at orbital altitude: a single step in the right direction is worth exponentially more than many steps in the wrong one. The wrong steps were the familiar sequence — ship a brochure now, bolt a chatbot on later, rebuild when the bolt-on strains. The one right step was structuring the content as typed, canonical data before anything else, so that every future capability — the assistant, the research desk, the briefing engine — connects to a spine that was always waiting for it. When the ports are designed in, then the station grows instead of getting rebuilt — which is, fittingly, exactly the philosophy GEC is taking to orbit.
Building something whose future is bigger than its launch? Fit the docking ports first. → RocketNow.com
Editor's note — before publishing:
- Confirm client naming/approval with GEC; replace the placeholder quote
- Phase 1 delivered / Claude phases proposed-and-gated framing is accurate per the registration docs — keep the tenses honest in edits
- Analytics metrics (impressions, visitors, enquiries, delivery duration) are deliberately unfilled in the sources — add only confirmed values
- "General Electric Company Ltd (UK)" is the client's stated legal identity in the docs — verify the naming treatment with the client before publication given potential confusion with other GE-branded entities