Skip to content

Editorial language registers

Topic

From the PointSav Documentation

The three PointSav wikis address distinct audiences in distinct language registers: Bloomberg and FT institutional tone for corporate and projects readers, Stripe and Cloudflare technical tone for documentation engineers, and RIBA specification prose for specialist deployment sites. Vocabulary retirement rules govern consistent translation across registers.

Updated 2026-05-21 Β· HistoryEspaΓ±ol

The three PointSav wikis address three distinct primary audiences. Each audience has different financial and technical literacy, different reasons for reading, and different vocabulary expectations. Writing in the wrong register produces content that fails to communicate β€” not because it is inaccurate, but because it addresses the wrong reader. This article defines the three registers, explains who each audience is, and provides the vocabulary rules that govern all three wikis. The underlying language protocol substrate enforces these rules structurally at capture time.

[edit]The editorial standard

Five rules, reconciled and operator-ratified on 2026-05-21, are the editorial standard behind every register defined here. The registers below specialize these rules for a specific audience; they never relax them. Where a register rule and a rule below appear to conflict, the rule below governs.

  1. Sentence length is budgeted by sentence role. An expansion sentence β€” one that develops a mechanism or an argument inside a body section β€” runs to about 45 words at most. A disclosure sentence β€” the lede, a compliance claim, a regulatory statement β€” runs to 25 words at most. Vary the rhythm: every paragraph carries at least one short declarative sentence, so the prose reads as an accordion rather than a monotone.
  2. Active verbs describe present-fact mechanism. Use the active voice to describe how something works now. Do not use it to assert a forward-looking claim as accomplished fact β€” capability, timeline, or outcome that is not yet true keeps planned, intended, may, or target. Do not give a system human intent or feeling. There is no ban on is, are, or was: a plain copula is correct when the sentence states a fact.
  3. Analogy is a ceiling, not a quota. An analogy is optional. Where one is used, hold to at most one per 300 words. Prose with no analogy is fully compliant; prose that reaches for an analogy in every paragraph is not.
  4. The lede is the nut graf; the Franklin arc orders the body. The Bloomberg four-paragraph lede carries the news in roughly the first 10% of the article. The Franklin arc β€” Crisis, then Quest, then Breakthrough β€” governs the order of body sections only. It never displaces the lede or delays the news.
  5. The SaaS-marketing register is rejected. Public content does not adopt the promotional voice of a software-product landing page. Internal codenames β€” "Liquid Glass" among them β€” stay internal; they do not appear in public content.

[edit]What a register is

A language register is the combination of sentence structure, vocabulary, depth, and tone appropriate for a specific audience in a specific context. A financial brief uses a different register than a technical specification. Bloomberg and the Financial Times use a different register than a building code. Stripe's developer documentation uses a different register than either. The PointSav wikis are written in three distinct registers β€” one per audience β€” applied consistently across every article.

Register is not style preference. It is the difference between a reader who understands what they came to learn and a reader who gives up and closes the tab.

[edit]Audience map

Wiki Primary audience Secondary audience Register
content-wiki-corporate Bankers, family offices, institutional investors C-suite principals, corporate advisors Bloomberg / FT / Economist
content-wiki-projects Top-400 development firms, commercial architects, construction programme managers Same institutional investors reading project subject matter Bloomberg / FT / Economist
content-wiki-documentation Software engineers, graphic designers, platform developers Emerging institutional readers evaluating technical credibility Stripe / Cloudflare primary + Corporate accessibility layer
bim.woodfinegroup.com Architects, engineers, building code officials β€” RIBA / IFC specification
gis.woodfinegroup.com GIS analysts, co-location programme managers β€” Technical specification
design.pointsav.com Design system contributors, component authors β€” Design specification (DTCG, component API)

The corporate and projects wikis share a register because they share a primary audience. The subject matter shifts β€” corporate governance and capital structure versus real estate development programmes and co-location markets β€” but the reader's financial literacy and capital-allocation frame does not. A family office principal reading about direct-hold tax structures and then reading about co-location development sites is the same person applying the same evaluative frame to different subject matter.

[edit]Register 1 β€” Bloomberg / FT / Economist

Applies to: content-wiki-corporate and content-wiki-projects

The reader is an institutional decision-maker. They have financial literacy and no obligation to understand software architecture or building codes. They are reading to make a capital allocation decision or to evaluate the platform's sponsors and structure.

Rules:

  • Sentence length: 14–18 words target; 25-word hard limit
  • Lead structure: consequence first β€” the most important fact for a capital allocator goes in sentence one
  • Voice: active. Passive voice hides responsibility and reads as evasion.
  • Numbers: always specific. "$7 per month" not "low cost." "Top-400 development markets" not "major markets."
  • Jargon: translate everything on first use. Platform-internal terms do not appear without a plain-language equivalent immediately preceding them.
  • Code blocks: never. The corporate and projects reader does not need to see terminal commands.
  • Citations: financial research, regulatory filings, industry reports, market data
  • Avoid: academic hedging, abstract nouns, dramatic marketing claims, platform-internal metaphors

Consequence-first pattern:

"The Direct-Hold framework removes the pool. Each property is its own legal and financial unit."

"Each co-location development site is underwritten as an independent capital event. Woodfine does not pool site performance β€” convergence at a node is validated independently before capital is committed."

Same reader, same register, different subject matter.

[edit]Register 2 β€” Stripe / Cloudflare + Corporate accessibility layer

Applies to: content-wiki-documentation

The primary reader is a software engineer, designer, or developer. They have technical literacy and are reading to understand how the platform works, how to deploy it, or how to contribute to it. The secondary reader β€” increasingly important β€” is an emerging institutional reader: a bank's technology committee, a family office's incoming principal, a senior developer from a Top-400 firm who is evaluating whether the platform is credible enough to approve or fund. This reader is technically literate β€” they can read code β€” but does not know the platform's internal vocabulary.

Rules:

  • Structure per section: Concept β†’ Why it matters (one sentence, consequence-first, no jargon) β†’ How it works β†’ Code β†’ Edge cases
  • The "why it matters" sentence must stand alone for the institutional reader scanning section headers β€” it is the Corporate accessibility layer
  • Code blocks: real and runnable. If abbreviated, mark with # ...
  • Field jargon (gRPC, systemd, JSON, async): use freely β€” no need to explain
  • Platform-specific terms: define once in plain language on first use, then use the term
  • Tone: peer-level, confident, direct β€” engineer to engineer
  • Avoid: over-explaining basics (HTTP, SSH, git), vague architecture descriptions, incomplete examples

Accessibility layer pattern:

"service-slm routes every AI request to the cheapest compute tier that meets the deadline, without the caller specifying which tier. Local inference first (5 s), cloud burst second (10 s), external API last (30 s). A request that resolves locally never leaves the customer's infrastructure β€” and never appears on a cloud billing statement."

The last sentence is the Corporate accessibility layer. An institutional reader scanning section headers gets: "it can run locally, and you pay nothing when it does."

[edit]Register 3 β€” RIBA / IFC specification

Applies to: specialist deployment sites only β€” bim.woodfinegroup.com, gis.woodfinegroup.com, design.pointsav.com

Prescriptive specification language (shall / shall not for requirements, measurements with units, normative vs informative distinction, standards citations) is the correct register for the specialist sites. It does not appear in the three main wikis because the main wiki audience is the institutional decision-maker, not the compliance engineer. "Shall" language reads as aggressive to a capital allocator. Specification language signals "compliance document," not "credible platform."

Specialist sites as cross-reference destinations:

When a main wiki article references a topic requiring prescriptive specification, link to the relevant specialist site:

  • "For full climate zone token specifications and IFC mappings, see bim.woodfinegroup.com."
  • "For co-location scoring methodology and zoning criteria, see gis.woodfinegroup.com."
  • "For design token schemas and component specifications, see design.pointsav.com."

These sites are canonical reference destinations, in the same way that Wikipedia articles link to technical standards β€” the main article is readable by a general audience; the specialist site is where you go to build or verify.

[edit]Vocabulary retirement

These terms appear in current wiki articles written for an internal engineering audience. They must be translated for any externally-facing article.

Retire Replace with
Substrate the data layer / the platform code / the security foundation
Doctrine architectural principle / design decision / engineering policy
Compounding training signal aggregation / model improvement over time
Leapfrog designed to exceed / targeted to replace / planned for [date]
Doorman access-control gateway / AI request router
Ring 1 / Ring 2 / Ring 3 archive tier / data tier / inference gateway
Totebox property archive / data vault
Yo-Yo pool on-demand GPU instances / ephemeral inference nodes
Sovereign (adjective) independently verifiable / operator-controlled
Compounding Substrate the platform (first use); link to architecture article
Apprenticeship Substrate the learning pipeline / the training system
LadybugDB property graph database / the DataGraph store

Rule: Retired terms may appear in architecture and services articles in the documentation wiki where they are being explicitly defined. They must not appear in corporate wiki, projects wiki, or GUIDE content without a plain-language translation immediately preceding them.

[edit]See also

Edit this page Β· View source