THROUGHLINE — Research Fan-Out (diligence methodology)

A consistent, reusable, multi-dimensional research fan-out that any org instance runs to find firm incorporation models and load-bearing legal constructs that maximize the likelihood of success. Deepens the GROUND move (THROUGHLINE.md §5), aimed mainly at Layer 1 (THESIS / economics) and Layer 5 (ARTIFICE). Status: DRAFT v0.1. Runnable: .claude/workflows/research-fanout.js (see §6).

0. What it produces

A ranked set of incorporation models + legal constructs for an org, each scored against a fixed success function (§3) and privilege ordering (§4), with every problematic construct named and given a resolution direction (§5). It does not pick the “best business” — it finds the structures most likely to endure and serve stakeholders given the org’s own substrate (per THROUGHLINE’s refusal to privilege desirability).

1. The six dimensions (the consistent fan-out)

Every org is researched along the same six axes — comparable inputs, apples-to-apples synthesis. One worker (or wave) per dimension.

#	Dimension	Asks	Typical sources
1	Research	empirical / primary literature on the model & domain	academic papers, primary data, gov/NGO reports
2	Precedent	analogous ventures & structures — what worked AND failed, and why	company histories, post-mortems, filings
3	Business studies	model mechanics, unit economics, governance cases	B-school cases, industry analyses, 10-Ks
4	Civil case law	what’s been litigated, challenged, upheld, struck	CourtListener, dockets, rulings, statutes
5	Economics	market structure, mechanism design, public-goods / commons, incentives	econ literature, mechanism-design, market data
6	Industry thought-leaders	named expert/practitioner opinion — regardless of favorability	essays, talks, interviews, expert commentary

2. Sourcing principle — comprehensive & adversarial (NOT cherry-picked)

Favorability-blind intake. Each dimension must surface the strongest evidence for and against the construct. An agent that returns only supportive findings has failed.
Seek the disconfirming case. Every dimension explicitly hunts the best critique, the cautionary precedent, the adverse ruling, the failure mode. “What kills this?” before “what helps this?”
Adversarial verification. Load-bearing claims are re-checked by an independent skeptic worker prompted to refute; a claim that can’t survive refutation is dropped or downgraded.
No silent truncation. If coverage is capped (top-N, a jurisdiction skipped), the synthesis says so.

This is constructive adversarialism: break the construct honestly, then resolve it (§5).

3. The success function (what “maximize likelihood of success” means here)

A construct counts as likely to succeed if it satisfies A OR B — two distinct, sufficient modes:

Mode A — Financially self-sustaining. Durable economics that fund the mission without dependence on continuous external capital (no reliance on perpetual fundraising or an exit). Resistant to the desirability ratchet (LENSES.md C: public-market / VC pressure).
Mode B — Impossible to resist at first phase. An initial wedge with structural, regulatory, moral, or network inevitability — adoption or legitimacy that cannot be cheaply blocked, competed away, or refused at launch. (A wedge so right it routes around opposition.)

A construct that hits neither A nor B is flagged as not yet load-bearing — the fan-out’s job is then to find the construct modification that reaches one of them (§5).

4. The privilege ordering (tie-breakers & optimization targets)

Among constructs that clear §3, rank by — in order:

Longevity — survives decades; succession-, founder-, and capture-proof.
Stability — robust to shocks; legal/governance durability; few single points of failure.
Societal benefit — positive externalities; net-good to the commons it touches.
Stakeholder (NOT shareholder) value — optimize for all stakeholders (creators, users, workers, community, society), explicitly not shareholder-return maximization. Favor steward-ownership, mission-lock, cooperative, PBC, golden-share, and trust constructs over extractive equity structures. (Continuous with THROUGHLINE’s desirability refusal: the shareholder-primacy frame is exactly the imposed-desirability hazard.)

These privileges break ties and direct resolution — they never override §3 viability, and they never demand a venture be more desirable; they demand it be more durable and stakeholder-serving.

5. Resolution bias (the orientation)

Findings are oriented toward resolving problematic constructs, not abandoning ventures. When a dimension surfaces a problem (a legal vulnerability, an adverse precedent, a broken incentive), the synthesis must answer: what incorporation model or legal construct resolves it? Output per problem:

problem        — the construct defect (with the adverse evidence + its source)
severity       — fatal · high · medium · low
resolution     — the structural fix (entity form, governance term, contract, jurisdiction)
success-mode   — does the resolved construct reach A, B, or neither?
privilege-cost — what longevity/stability/societal/stakeholder trade the fix implies

The deliverable is a menu of load-bearing constructs, each problem resolved or explicitly marked unresolved-and-why.

6. Mechanics (how to run it)

Pre-fetch shared surfaces once (per the research-team discipline): collate the org’s CANON.md + RESEARCH_BRIEF.md + any pulled sources into /tmp/<org>-surfaces/ so N dimension workers read the same ground truth instead of each re-fetching.
Wave-batch the fan-out (3 + 3, not one big parallel) to stay under the API token-per- minute ceiling.
Adversarial verify flagged load-bearing claims (§2).
Synthesize into the ranked construct menu (§3 success × §4 privileges × §5 resolutions).
Feed back the chosen constructs to the org’s CANON.md via the amendment back-edge, then critpath sweep to propagate.

Runnable workflow: .claude/workflows/research-fanout.js — args { org, question }. It is opt-in (spawns many agents, real token cost): launch via the Workflow tool, or ask for it explicitly (“run the research fan-out on RINA”). It is prepared, not auto-run.

7. Plug-in point in THROUGHLINE

This is the GROUND move done at depth for the two layers where structure is decided:

Layer 1 THESIS / economics — is the belief sound; where does value pool; what’s the moat characterization (not grade).
Layer 5 ARTIFICE — which incorporation model + legal constructs are load-bearing under §3–§5. This is the layer the fan-out exists to serve.

DRAFT v0.1. Conforms to THROUGHLINE.md. Comprehensive & adversarial in intake; resolution- biased in synthesis; success = A∨B; privilege longevity · stability · societal · stakeholder.