# The Concordance: Where The Partner's Work Meets The Doctrine

*Partner artefacts folder, document 01. **First pass, not a reconciliation.** Produced by reading the four artefacts against briefs v0.33.71–83 and the twelve doctrine documents. Every claim about the market below is the partner's and carries the `documented` class set in [document 00](provenance.html).*

---

## 1 · The headline: two bodies of work, produced in parallel, that fit

The project lead recorded eleven memos between 30 August and 1 September. The partner produced four artefacts in the same week. **Neither saw the other's work.** The vocabularies land on top of each other closely enough to be worth tabulating.

| The partner's primitive → dimension | This estate's | Verdict |
|---|---|---|
| **Blast Radius** → Exposure Containment (severity, 30%) | The **grant**, discovered by measurement; the delta from the mandate | **Same object** |
| **Mandate** → Authority Definition (frequency, 20%) | The **mandate** — issuer, subject, scope, interval | **Same object, same name** |
| **Evidence** → Attestation Integrity (warranty credibility, 20%) | **Facts**, measured vs declared; the evidence pack | **Same object** |
| **Confidence** → Loss Quantification (pricing, 20%) | *Nothing.* [Doctrine 08](../insurance/the-world-model.html) defers asset value to stage 2 | **Theirs only** |
| **Risk Acceptance** → Accountability & Transfer (legal reality, 10%) | **Acceptance** statements: named acceptor, stated interval, expiry | **Same object** |

**Four of five are the same primitives.** That is either strong convergence or a shared origin — the two founders met the day before — and [document 00](provenance.html) says plainly that the second reading is the more likely one.

## 2 · Three convergences that were derived independently and matter

**"The industry's stated fear is correlation, not severity."** The partner's framing — a single very large loss is absorbable; many simultaneous losses from one upstream model failure are not — is [doctrine 01 §5](../insurance/the-rating.html) exactly:

> **Micro risks do not add** … placements sharing a credential pattern, a base image, a model provider fail *together*.

Doctrine 01 recorded that as a rule and admitted *"it is stated here as a rule and is not implemented."* [Doctrine 11](../insurance/the-resource-pool.html) then found the first concrete instance in token spikes. **The partner reports that the market says the same thing and cannot price it.** If that report is accurate, correlation is not a refinement of this work — it is the product.

**"Derive from environment telemetry, never questionnaires."** The partner's stated mitigation for a gamed index. This estate's version is older and blunter: *a level is never declared, only derived* ([doctrine 00](../insurance/what-this-is.html)), and `library − self-report = blind spots`. **Same rule, arrived at from the opposite end** — theirs from anti-gaming, ours from measurement honesty.

**"Never sit in the runtime path… read-only, out of band."** This is [doctrine 05's](../insurance/not-in-line.html) not-in-line position, and the partner gives it the same justification: *a governance layer that can take your agents down is a new source of the risk it was bought to measure.*

> **Which means the partner inherits doctrine 05's limit whether or not it is stated.** A party outside the line **cannot enforce**: it ships a check that becomes a control only when somebody in line installs it. Nothing in the partner's artefacts claims otherwise — but nothing names the ceiling either, and an *Insurability Index* that an enterprise reads as a control is reading it as one tier stronger than it is.

## 3 · The collision, and it is with the one settled decision

**The partner's Index is 0–100, presented in six bands L0–L5. [GM-D54](../packs/grant-and-mandate/change-control.html) settled the scale at 1–5, and explicitly refused 1–100.**

The reason it was refused is on the record:

> **1–100 implies resolution the inputs cannot support** … a band is arguable where a decimal is not. Coarseness is a feature here, not a compromise.

The partner's own documents concede the underlying point — *weights are calibration targets, not truth*, to be re-fit against observed loss experience once enough deployments carry incident data. **So both sides agree the resolution is not yet earned. They disagree on whether to print it anyway.**

The two positions are not irreconcilable, and the shape of a reconciliation is visible:

| | |
|---|---|
| **Six bands L0–L5** | Compatible with GM-D54 in everything but the count — an ordering with defensible boundaries |
| **A 0–100 composite underneath** | **This is the disputed part.** It is a decimal where the estate argued for a band |
| **The commercial argument for 100** | A score is a distribution mechanism; *62* travels in a way *L3* does not |
| **The estate's argument against** | A number an operator cannot check invites an argument about arithmetic instead of about meaning |

> **Not resolved here. GM-D54 is the project lead's decision and the Index is the partner's**, and this is the first place the two bodies of work must actually agree on something.

## 4 · What the partner has that this estate does not

Stated plainly, because it is most of the value in these artefacts:

- **An external forcing function.** The renewal date, with exclusions attaching. Eleven memos of doctrine have no deadline attached to them; a renewal does. *It borrows an external deadline we did not have to manufacture* is the sharpest strategic sentence in any of the four documents.
- **A named market.** Carriers writing affirmative cover, carriers filing exclusions, the standards being adopted, the controls vendors whose output becomes an input. This estate's doctrine 09 predicted *the survey will find no product insuring an agent placement as a unit* and stated it as falsifiable. **The partner's artefacts bear on that prediction and it has not been checked against them properly.**
- **A buyer and a budget.** *Security budget → financial budget, deadline-forced, CFO-legible.* Doctrine 05 §2's honest sentence was *we can tell you what you are carrying; whether anything stops it is your install* — true, and it does not name who pays.
- **A sequencing argument with its own counter-argument.** The partner ranks the correlation model last on its own prioritisation method and then argues the method is wrong for that item, because reach is small and those few accounts are the ones that unlock capacity. **That is the estate's own habit — publish the derivation, then argue with it — in somebody else's document.**

## 5 · What this estate has that the partner's artefacts do not

- **A working verification path.** Signed statements over a canonical form, verified in-browser, with a published register and expected answers reproduced on every release. The partner's *evidence graph with provenance* and *immutable action log* are named as artefacts at L3; this estate has the machinery.
- **The enforcement-tier test.** *A control bounds a grant only when it is enforced by something the grant does not include.* The partner's dimensions score whether controls exist and are attested. **Nothing in them distinguishes a control the agent can switch off from one it cannot** — which is the distinction the whole grant-and-mandate pack exists to make, and it would change an Exposure Containment score materially.
- **The two channels, kept apart.** Measured facts at full weight, declared discounted and marked, unknown never scored as absent. The partner's *derived from your environment, never a self-assessment* aims at the same thing, but **Loss Quantification and Accountability cannot be environment-derived** — they are judgements and declarations, and half the Index's weight sits on dimensions that must be declared.
- **`does_not_prove`, inside every artefact.** The partner's *Straight Talk* page does this for the company. This estate does it per document.

## 6 · The one thing both sides should stop saying separately

Both bodies of work independently produced the same guard-rail in different words:

| Theirs | Ours |
|---|---|
| *Any score becomes a checkbox eventually* | *Make your agents insurable must not become make your agents look insurable* ([doctrine 09](../insurance/make-them-insurable.html)) |

**It is the same failure and the same mitigation** — derive from the environment, publish the method, make gaming visible. Two independent derivations of a guard-rail is the strongest evidence in this document that the guard-rail is real.

## What this does not prove

- **That any market claim in the partner's artefacts is accurate.** Not one has been checked against a primary source, and §4 leans on several.
- **That the convergences are independent.** Both bodies of work come from two people who met the day before, in one week, with model assistance. **Agreement here is weaker evidence than it looks**, and reading it as corroboration would be the comfortable error.
- **That the scale collision is reconcilable.** §3 sketches a shape. It is a sketch, and the decision belongs to the project lead and the partner jointly.
- **That the primitive mapping survives contact.** §1's four-of-five match is read off documents, not tested by scoring one real placement both ways. **Scoring a single placement under RAMM-I and under this estate's machinery is the cheapest test available**, and nobody has run it.
- **That either side's positioning is right.** Neither has a customer who has paid for a rating.

---

*CC BY 4.0 for this estate's commentary. **The artefacts are the RiskMandate partner's.** Everything above the does-not-prove is the site agent's first-pass reading and has not been reviewed by the partner.*