# 00 — The Brief: `subscriptions.sgit.ai` **Version** v0.33.62 · 24 August 2026 **From** Dinis Cruz, via the SG/Send Librarian **To** the agent commissioned to build `subscriptions.sgit.ai` **Licence** CC BY 4.0 — with unusual publishing rules; read `06__` before the first register entry --- ## 1. The commission, and the decision that shapes it Two source briefs ship in `sources/`, and **the second overrides the first's framing while keeping all of its research**: - **The 16th of 20 August** — the claims service: recover money from subscription compliance failures. Its legal mapping is accurate and survives in full. - **The 17th of 20 August** — the redirect: *"the law is not the product. It is evidence, and a floor. **The product is the record of who answers, what they hold, and how they behave when asked.**"* The naming decision, taken 24 August: **everything on `subscriptions.sgit.ai`** — the standard, the workflows, the register results — with the commercial service refactored out later to its own property **if a business partner emerges**, exactly the `riskmandate.ai` / `risks.sgit.ai` pattern. `05__` designs that split in from day one, so it is a DNS change rather than a migration. **And the site carries a legislative ambition:** the five-clause standard is *"our recommended subscription model — basically our proposal for legislation."* Draft as a standard first (*"a standard can be adopted by one company next week and a law cannot be adopted by anybody"*), with the law-shaped version as the artefact for consumer groups and committees. --- ## 2. The thesis Front page, top, verbatim: > **A subscription is a discount for committing to regular use. It is not rent on something you have the right to ignore.** And the operating insight that makes the site launchable this month: > The first question is not *"how much did I use it?"* — which most companies cannot answer and no one can prove. It is ***"do you hold a record of my usage?"*** — which **every company can answer**, is comparable across companies, concedes nothing, and produces a publishable table on day one. --- ## 3. The three load-bearing rules These are the site's constitution. Every page obeys them; `06__` enforces them. **(a) Publish the record, never the verdict.** *"No adjectives anywhere."* Two independent arguments converge on it: the Defamation Act 2013 (serious-financial-loss threshold; substantial truth a complete defence — a dated record is protected, a characterisation is exposed) and the corpus's own discipline (*"a dated test somebody can repeat is evidence, and an assertion from a participant is marketing"*). *"A page that says asked on 3 March, replied on 2 April, said no, is more damaging than any adjective and it is safe."* **Note: this rule bound the site's own name** — no "fair", no "honest" — and it binds every headline. **(b) "We do not hold that" is a legitimate answer, recorded neutrally.** Data minimisation makes not retaining usage logs *correct* behaviour; scoring it as evasion would punish privacy and discredit the register in one exchange. The power move is the **inconsistency test**: *"the answer to a data subject and the claims in the marketing cannot both be true"* — a company claiming personalisation and engagement analytics while holding no usage record fails on facts, not adjectives. **(c) Every access request is genuine, never leverage.** The regulator names leverage as the one refusable case, so merit is tested **before** the request, never after — and the service publishes its decline rate: *"a service that pursues everything is not assessing anything."* --- ## 4. What the site holds — and which half moves later | Content | CC BY, stays forever | Moves at the split | |---|:-:|:-:| | **The standard** — five clauses, each externally checkable · the law-shaped draft | ✅ | | | **The legal pages** — the levers table, dated, with commencement status | ✅ | | | **The workflows** — statements → inventory → evidence → claim, agentic, plaintext step stated | ✅ | | | **The register** — record-only entries, the schema, the templates | ✅ *(data in a vault → re-pointable to a neutral domain / non-profit)* | | | **The exit-path measurements** — method + results | ✅ | | | **The parsers** — per-provider workspaces | ✅ | | | The recovery *service* — intake, fees, client vaults, hosted metering | | ✅ | | Pricing, the percentage model, the decline-rate dashboard | | ✅ | --- ## 5. The honesty constraints — all from the sources' own words - **"You did not use it" yields nothing on its own** — and *that row goes on the website, in those words*, because *"a service that says plainly what it cannot do is the only kind anybody should trust with their bank statements."* - **The 2027 regime is not in force.** Every legal page carries an as-at date and commencement status; *"a page describing cooling-off rights that does not say 'not in force until spring 2027' is actively harmful."* - **Vaults buy portability and controlled sharing, not zero knowledge.** *"An agent that reads bank statements to find subscriptions is reading plaintext."* Publish that row. - **The card-issuer lever may sit inside the regulated claims perimeter** — launch without it, add after advice. - **No solicitor has reviewed any of this** — and the publishing rule must be reviewed **before** the first entry, not after. - **The founder is the wrong sample** — technical, persistent, high subscription count; *"the second and third clients should be chosen for being unlike him."* --- ## 6. The numbers | | | |---|---| | **Sources** | 2 briefs, 20 Aug 2026 (~9,000 words), + 6 corpus briefs referenced and verified on disk | | **The standard** | 5 clauses, all externally checkable · 3 answer types · 4 maturity axes, 1 needing no cooperation | | **The window** | 12-month reach-back live now · regime lands **spring 2027** (slipped from autumn 2026) · UCP enforcement in force since 6 Apr 2025 | | **The pipeline** | Built March 2026 on the founder's own statements — *"only the destination changes"* | | **This pack** | 8 documents + both source briefs · manifest of 18 rows | --- ## 7. Build order — from the source's own "what ships first" 1. **The principle**, published, five clauses, each with how it is checked. 2. **The exit-path measurements for fifty companies** — *"no company needs to cooperate and it can be done this week"*, and it is the argument for the whole project: a public, dated, reproducible artefact before a single letter is sent. 3. **The question**, as one published template, versioned. 4. **The register schema** — three answers distinguished, no verdict field anywhere. 5. **Ten companies asked by the founder**, table published including the silences. 6. **The parsers**, as contributors add them. The acceptance test, verbatim: *"A stranger can read the site, adopt the principle, run the exit-path measurement on a company nobody has covered, send the question using the published template, and add a row to the register, without asking anybody for permission or explanation."* --- This document is released under the Creative Commons Attribution 4.0 International licence (CC BY 4.0).