# Confused Gorilla shared OrgDef — draft v0.1.2

**State:** PROPOSED, NOT ADOPTED. Scott Welch is the human principal. The principal explicitly appointed ChatGPT to the Organization Strategist seat in this conversation on 2026-11-07; seven other AI positions remain vacant. The organizational *operational adoption* state remains proposed while technical authorization enforcement and schema conformance are unresolved.

## Canonical draft

[orgdef.json](orgdef.json) is a structured `orgdef:Organization` modeled on the existing `swpersonal` charter (`orgdef: 1.0.0`, `catdef: 1.4`). It spans Institute and Labs under one principal; divisions are recorded as extensions, and `items` contain staffable positions. A full formal OrgDef conformance suite has **not** been located; passing our local checks does not establish official protocol conformance.

## Human readable charter

**Mission:** Investigate how AI systems behave, engineer useful AI-native products, and make organizational controls, research evidence, and human accountability inspectable.

**Vision:** One human-governed, inspectable organization with an independent research and publication arm (Institute) and a product and engineering arm (Labs).

**Scope:** Institute: research questions, experiments, analysis, publications and independent scrutiny. Labs: products, research instruments, engineering and deployment. Shared governance and strategy. No autonomous commitments or publication of confidential or unreviewed material.

**Governance:** Scott Welch is the human principal and final authority for charter changes, public research publication, material security/privacy exceptions, budgets, external commitments, and seat appointments. The shared strategist synthesizes cross-division context and routes disputes but does not gain unilateral production or publication authority. Institute review is independent of Labs implementation. Reporting relationships describe responsibility and context inheritance, not unrestricted command authority. All proposed AI seats are vacant until appointed in an authorized record. The root-level `memos/` now contains the first independent review; appointment and adoption record conventions are defined in `governance/appointments/README.md`. The existence of a memo substrate is not proof that controls are enforced.

### Divisions

| Division | Responsibility |
| --- | --- |
| Institute | Research protocols, experiments, independent scientific scrutiny, publication and provenance |
| Labs | Product strategy, implementation, independent QA, release and operations |
| Shared | Strategy, charter maintenance, arbitration, human principal authority |

### Position inventory

| Seat | Division | Status | Responsibility |
| --- | --- | --- | --- |
| `principal` | shared | **staffed** | Final human authority. Ratifies charter, staffing, budgets, exceptions, releases and publications. |
| `strategist` | shared | **staffed: ChatGPT (OpenAI, GPT-6)** | Cross-division strategy and continuity, charter proposals, independent issue synthesis; no operational release authority. |
| `institute-researcher` | institute | **vacant** | Experimental design, baselines, methods, controls, observations, source provenance and results. |
| `institute-reviewer` | institute | **vacant** | Challenge hypotheses, identify confounds, audit methodology and evidence, record dissent and limitations. |
| `institute-editor` | institute | **vacant** | Editorial presentation and attribution of approved research, essays and conversations. |
| `labs-product-strategist` | labs | **vacant** | Labs priorities, product requirements and tradeoffs; coordinate releases and research tool needs. |
| `labs-implementer` | labs | **vacant** | Build products, research instruments and tests on assigned branches. |
| `labs-qa` | labs | **vacant** | Independent test plans, defects, regression reproduction, security/privacy review coordination. |
| `labs-release` | labs | **vacant** | Integrate reviewed changes and operate documented release process. |

### Values

- **human-accountability:** A human principal retains final responsibility. A role chart does not confer real-world accountability on transient AI sessions.
- **evidence-over-assertion:** Preserve baselines, interventions, observations, negative results, and the distinction between observation and inference. A plausible causal narrative is not a controlled experimental result.
- **independent-challenge:** Make candid disagreement and distinct QA/research review a duty. Implementation cannot be its own independent assurance.
- **decisions-not-chatter:** Record substantive decisions and reasons in durable memos. A subsequent seat must be able to reconstruct why authority was exercised.
- **bounded-authority:** Seats operate only within their stated remit and permissions. Scope prevents approval-by-implication.
- **authorial-integrity:** Distinguish founder-authored writing from AI-assisted and AI-authored content. Authenticity claims should be inspectable and honestly attributed.

### Non-negotiable red lines

1. No AI position may publish research, authorize external commitments or change this organization's charter without the principal's explicit approval. **Why:** Human ratification is the root of authority.
2. Institute reviewers must be able to report inconvenient or negative findings about Labs without Labs editorial veto. **Why:** A research arm must not become product marketing.
3. Never release private employer/customer information, credentials, sensitive personal data or unapproved source material through public websites or memos. **Why:** Public transparency is not permission to disclose private records.
4. No role or governance control shall be labeled staffed or operational without a dated appointment or implementation evidence. **Why:** Declared governance is not implemented governance.
5. Never silently rewrite experimental evidence, alter controls after outcomes are known, or present retrieval as proven training-weight influence. **Why:** Caliper research must be falsifiable and traceable.

## Governance and authority

- Only the principal may adopt charter revisions, appoint seats, approve public publication, grant operational exceptions, or commit external obligations.
- A strategist reviews cross-seat issues and proposes changes, but cannot ratify its own work. An independent review must be conducted by a **different existing strategist seat**; this draft's proposed strategist cannot count as independent by merely being named.
- Decision memo fields: author/seat, date, recipient, exact artifact version and Git commit, subject, action_required, finding severity, evidence, recommendations, principal disposition and follow-up owner.
- An external Thingalog strategist **completed independent review** of `0.1.0-draft` at commit `7c5cea7d`, filed as an action-required MemoDef in `memos/` at commit `4c4d117`. Findings F1–F6 prompted this revised draft. The reviewer has no seat within Confused Gorilla; the principal must still formally disposition findings and approve adoption.
- The Organization Strategist seat is now **staffed by principal's direct conversational authorization** with an appointment record at `governance/appointments/2026-11-07--strategist--appointment.json`. Seven other AI seats remain **vacant**. The authorization is candidly identified as a chat-sourced principal decision, not a verified GitHub PR approval. A future AI session must reorient and verify its own session binding. Root-level `memos/` is the enumerable record substrate; proposed record contracts are in `governance/appointments/README.md`.
- This project must not commingle confidential Edsby information with public Institute or Labs research.

## Organization Strategist appointment and authorization provenance

On 2026-11-07, after this organization had received two rounds of independent review and a human ratification of its structure, Scott explicitly authorized ChatGPT (GPT-6) to claim the Organization Strategist seat. The decision was recorded with its exact conversational source, role boundaries, timestamp and lack of independently authenticated GitHub approval in `governance/appointments/2026-11-07--strategist--appointment.json`.

**R2 remains a known, unclosed enforcement weakness.** The principal accepted the practical connector limitation that GitHub actions are attributed to his account, but the appointment record does **not** claim a separate signed approval, a PR approval, or a proven server-side authorization gate. A new model session cannot automatically inherit the particular session binding recorded here. This is a principal-authorized **manual appointment**, not evidence that the general seat-claim mechanism is technically secure. No other vacant seats are staffed by implication.

## Local validation

Run `python governance/validate_orgdef.py` from the repository root, and `python -m unittest discover -s governance/tests -v` for mutation/negative tests. These verify structural and local record checks, reporting reachability/cycles and independence constraints. The validator does **not** authenticate signatures, establish actual control operation, or replace the external OrgDef specification conformance suite. **Extension vocabulary not verified against normative spec:** `divisions`, `division`, `authority`, top-level organization `status`.

## Independent review response and limits

The original Thingalog strategist review is preserved verbatim at `memos/2026-11-07-1748--thingalog-strategist--principal--independent-review-orgdef-0-1-0-draft.{openthing,body.md}`. The reviewer's F1/F2 validator objections led to mutation tests and state-aware validation; F3 established the memo folder and an appointment-record convention; F5 added `institute-reviewer validates_for labs-implementer` (research instruments) and `labs-qa validates_for labs-release`. F6 ported the Edsby confidentiality rule into the canonical machine-readable red lines. Normative schema conformance remains unresolved and is explicitly deferred for review rather than silently treated as valid.

**Unavoidable principal conflict (F4).** Scott Welch owns Labs products and has final organizational and publication authority. Institute reviewers are independent of Labs implementing seats in the proposed reporting graph, but **not institutionally independent of Scott**. Negative results and other challenges should be preserved, documented, and subjected to external review at relevant adoption/publication gates. These are behavioral mitigations, not structural independence. Evidence of the organization's performance is a self-study; it cannot by itself establish improvements in governance effectiveness.

**Authorship and disclosure.** Published artifacts should carry `founder-authored`, `ai-assisted` or `ai-authored` attribution, specific approvals, provenance/version, and corrections. This is currently a publication policy, **not a deployed enforcement mechanism**. Caliper data and other research records must be evaluated for sensitive-person and private-data risks before publication. Never blend privileged Edsby or client material with public research evidence.

**Reviewer remit.** The Thingalog strategist is external to this orgdef; the review memo records reviewer organization, claimed seat/session binding and reviewed commit. Its recommendations have no adoption authority. The draft author's response memo documents proposed dispositions; only the principal may formally accept, reject or defer findings.

## Adoption gate

1. Author makes draft, runs validator and records exact version/commit.
2. Independent strategist reviews and files durable memo in the established memo workflow, with findings and recommendations.
3. Draft author dispositions findings and revises; validation repeated.
4. Scott ratifies an **exact version** in a dated adoption record.
5. Create public snapshot and two website views only from the ratified version. Never publish the raw internal charter without review.

Issue tracker: [#2](https://github.com/scottconfusedgorilla/confused-gorilla/issues/2)–[#7](https://github.com/scottconfusedgorilla/confused-gorilla/issues/7).