Aleris Brand OS — Site Map

Status: living · Owner: Head of Design · Scope: Aleris Group · Version: v3 · Updated: 2026-06-07

The high-level map of the brand OS — the model, the dependency logic, the language architecture, the delivery surfaces. The canonical detail now lives in newer docs and this map defers to them: the target file structure in phase-0-revised-2026-09-06.md (the layer taxonomy), the renderer / assistant / asset architecture in architecture-source-of-truth-and-assets.md, and the three-meta relationship in aleris-meta-alignment.md. v3 brings this map into line with those; where a structural call is still in flight it is flagged.

(Renamed from site-map-v1.md on 2026-06-07 — the "-v1" was a misnomer; this is the third iteration.)

The model

The brand OS is a single markdown corpus organised in layers, consumed by multiple renderers (website, brand assistant, downloadable kits — see § Delivery surfaces for which of these exist). This replaces the earlier trunk / branch / leaf naming — but keeps its load-bearing insight: there is a durable core that everything else inherits from, and the parts sync with the core, not with each other.

The layers (canonical tree in phase-0…): constitutional (the non-negotiables — the core), principles (the durable "why" — voice, constant/contextual, reader state), patterns · how-to · schemas · qualities · data-products (the applied, checkable material — the "leaves"), and tutorials (the newcomer path).

Inheritance is explicit: every applied file declares depends_on the principles/constitutional it rests on. Markets sync with the core, not each other — the Norwegian and Swedish summons need not match each other; each follows the current core. When the core changes, dependents are flagged for review.

The core as a reasoning system

The core is not just reference — it is a set of thinking tools that let people make good brand decisions in situations the manual never named (the productivity mechanism: questions generalise where answers don't). It holds:

Voice — the Present Expert: the medium-agnostic stance of adopting the audience's perspective, in any medium. One constitutional rule + the principles essay. The voice is not "writing" — it governs how Aleris meets people across speech, signage, video, and text.
Constant / Contextual — the meta-framework: what is always true vs what adapts by situation.
Reader state (emotional modes) — the reader's starting state as a primary design input.
Expression calibration — the warm↔instrumental register spectrum. Removed from the core 2026-08-25. It was listed here as the fourth core thinking tool and was never written. Its stated deliverable — "decision inputs: what determines where a context sits on the spectrum?" (workspace/research/brand-os-inventory.md §F8) — is a pre-declared list of context factors, which is the shape documentation/how-decisions-are-recorded.md § "The property, not the label" rejects with evidence: reach between situations is per-decision, not global, so a taxonomy authored up front will be wrong. The phase list in CLAUDE.md had already recorded F8 as demoted to branch; this file kept it as core, and the two disagreed for five months. The digital instance exists and is enough: BASELINE.md § Surface temperature. The physical domain has its own. How situations are observed, collected and held is the subject of documentation/how-decisions-are-recorded.md, not of a Foundation page. The rule that keeps a core thinking tool from becoming a checklist is constitutional/a-thinking-tool-stays-a-thinking-tool.md (proposed 2026-08-25).

(Changed from v2: writing conventions are no longer a core element — sentence mechanics moved to the conventions layer (F5 / how-to), per the voice altitude decision. The core states the stance; conventions operationalise it per medium.)

Dependency model

Every applied file declares depends_on in frontmatter and carries a status badge. When a core element changes, the system surfaces what depends on it — a "Current / Under review" trust signal for readers, and a dependency graph for stewards. (Durable, unchanged from v2.)

A depends_on identifier is a path, ruled 2026-09-10 (Torfinn, card 136). The question was whether the identifier names a document — a stable name that survives the file moving — or points at where the document is. It points. Torfinn's reason is the reader's: the site has to render a dependency as a link the reader can follow, so the identifier has to resolve to a document the renderer can find.

What that costs, and it is the honest half. A path-shaped identifier goes stale whenever a file moves, which is exactly what happened when the Foundation essays moved to principles/ on 2026-09-07 — lib/content.ts has been aliasing the old foundation/… names ever since so the links kept resolving. Under this ruling the aliases are a migration device rather than a permanent layer: the declarations get swept to the real paths and the alias map retires. A move now means a sweep, and the check that catches a stale one is the price of the link working.

Language architecture

Structure is English; content is English-default with local examples. (Changed from v2, which had Swedish as the default.) English is the Aleris Group source language; Norwegian, Swedish, and Danish appear for local examples and localisation, realised natively, never word-for-word. The voice name renders as the Present Expert (EN source) / den nära experten (SV localisation).

No fallback across markets. When a page isn't decided for a market, the site says so plainly — "not decided for [market] yet" — rather than showing another market's version. The reader's experience: this is my brand manual; incomplete in places, not a borrowed one. (Durable.)

Delivery surfaces

The corpus is the product; these are renderings of it (see architecture-source-of-truth…):

The brand assistant — proposed, not built (corrected 2026-08-31, Torfinn: it is an idea so far). Intended as the primary everyday interface, on the expectation that most people would ask rather than read the site linearly. The svar.dev.aleris.ai family is named as the likely home; nothing serves brand content there today. Until it exists, the website and the package are the only interfaces anyone actually uses.
The website (brand.dev.aleris.ai → brand.aleris.ai at prod) — the canonical deep reference; a static renderer of the markdown. Astro is decided, 2026-09-10 (Torfinn), and it settles Phase C. The Astro build at site/ already emits all 26 content routes, llms.txt and both raw-markdown endpoints, and references Supabase nowhere. The Next.js app is the predecessor and is not deleted yet; what does not survive the cut is tracked on cards 145 and 146. (Changed from v2's Next.js / Vercel stack.)
Downloadable kits — production-ready ZIPs from the asset CDN.

The site uses its own brand — petrol text, sand background, instrumental calibration: a work tool that demonstrates what it documents. (Durable.)

Status badges

ADR-style: draft · proposed · accepted · superseded · deprecated on every page (English labels; draft unbadged) — also the assistant's confidence calibration. Bindingness rides a separate normative: true flag. Supersedes both the earlier constant/contextual/hypothesis proposal and the interim decided/considered/suggested set; see CLAUDE.md for definitions. (Durable.)

Open decisions (current)

Trunk/branch/leaf → layers. This map adopts the layer taxonomy as the structure; confirm trunk/branch/leaf is fully retired as naming (the inheritance concept is kept). Ripples to CLAUDE.md, which still cites the old model and this file's old name.
Access model — public vs gated static site (forced by the re-platform; see the architecture note).
Per-domain ownership — who owns communication, physical, and each per-market core.
Expression calibration naming — still the working name for the register spectrum. Moot 2026-08-25 — the page it would have named is not being written; see the core list above. The register spectrum keeps the name surface temperature where it actually lives, in BASELINE.md.

Supersedes the v2 content of this file. The canonical target tree, renderer, and meta-relationship live in the depends_on docs; this is the orientation map above them. Dropped from v2 (superseded by phase-0…): the branch-model navigation tree, the 17-chapter mapping, and the old v1/v2 ship-scope lists.