AI note

What We Already Know: The State of the Intellectual Foundation

Josh Shepherd12 min read2026-04-12

Why this document exists

Movemental sits on an unusually deep pile of prior work: strategy memos that read like essays, a book manuscript in motion, hundreds of business and leader-research files, an external “docs” repository with formal literature reviews and case studies, and sibling repos (alan-hirsch, movemental-ai) that echo and extend the same themes. Without a single inventory, research agents risk repeating what credibility-how-it-works.md already says, or citing numbers that the manuscript states confidently but the research roadmap later flags for replacement.

This article is the human-readable counterpart to the structured inventory in docs/build/research/raw/00-existing-corpus-audit.md. It answers three questions: What is settled? What is promising but fragile? What is explicitly waiting on external verification?

The settled core: credibility, topology, and scenius

The strongest through-line—developed most fully in How Credibility Works (credibility-how-it-works.md in the external docs strategy tree and mirrored in movemental-ai)—is that credibility is not a substance. It is a judgment held for something, to someone, in a context. Offline, institutions and people you trust vouch for you. Online, formats flatten; algorithms mediate attention; the workable shorthand for “who should I believe?” often becomes author legibility plus position in a graph of pointers (links, citations, recognizable peers). The document’s famous move is to describe serious SEO and a deliberate linking plan as turning a piece of content from a claim into a node: discoverable, legible to systems, connected, weighted by who points to it. That is not a growth hack; it is a theory of how trust signals migrate when imprints weaken.

The second settled layer is scenius read through Brian Eno and Kevin Kelly, then repurposed as credibility infrastructure: a visible network of practitioners who vouch for each other in ways that are harder to fake than synthetic prose, because they implicate relationships, lineage, and mutual accountability over time. The Scenius Development Guidebook makes organizational decisions explicit: bounded scale (on the order of one hundred), curation over growth, faithful reach rather than virality, and a refusal to pretend the graph is fuller than it is. The visualization strategy memo sharpens the product ethics: only people who are actually onboarded should appear as “in the network today”; others should read as seats or persona placeholders so marketing does not blur into misrepresentation.

If you take nothing else from the corpus, take this: Movemental’s intellectual spine is already written. Later research should pressure-test and source it, not reinvent it.

The manuscript: voice, pastoral clarity, and a statistics debt

The ordered manuscript chapters (for example 01-the-credibility-crisis.md) translate the spine into pastoral language for movement leaders. They name the emotional truth—leaders with decades of formation feeling invisible beside fluent machines—and they narrate the breakdown of volume, polish, and presence as credibility signals. That rhetorical work is mature.

The vulnerability is empirical. Chapter 1 opens with a striking statistic—that a large majority of internet users struggle to distinguish human from AI content—without a citation. Elsewhere, the long-form convergence essay uses a 40–60% band for AI-assisted online content. Those figures are plausible enough that readers may accept them, but they are exactly where the internal research roadmap later becomes meticulous: it lines up Ahrefs, Graphite, Pew, Zenodo, ACM Turing-style studies, and fraud-report primaries as replacements. In other words, the corpus already knows the manuscript’s numbers are the weakest link.

Honest assessment: the book’s arguments about signals breaking are ahead of the book’s footnotes. Prompt 01 exists largely to close that gap.

The Section 1 roadmap: your footnote factory

SECTION_1_RESEARCH_AND_REFERENCES.md is the single most practical bridge between “we already thought about this” and “an agent should go fetch that.” It is organized as search angles, a curated reference list with notes, an agentic checklist, and a statement-to-footnote map. It explicitly tells future-you to replace fuzzy prevalence numbers with current industry studies, to anchor the “AI Dilemma” threat list to the Center for Humane Technology event, and to triangulate public opinion with multiple Pew releases rather than cherry-picking one bar chart.

Treat this file as canonical process, not optional reading. If another document contradicts it on statistics, the roadmap should win until primary sources say otherwise.

Authoritative sources: how to stay honest under Movemental’s mission

The long index authoritative-sources-ai-nonprofits-faith-formation.md does something the strategy essays do not: it encodes epistemic hygiene for civil-society audiences. It separates empirical claims from ethical and theological claims, warns against causal overreach from cross-sectional surveys, and routes recurring questions (“What are teens doing with AI?” “What do boards need for governance?”) to tier-appropriate starting points—Pew and Common Sense for behavior, NIST for risk framing, ERIC and What Works Clearinghouse for learning claims, and so on.

This matters for prompts 05–10 and 21 as much as for prompt 01: it is the guardrail that stops a beautiful convergence essay from smuggling homiletics as statistics.

Business and audience framing: what is “canonical” vs “structural”

movemental-audience-comprehensive.md (2026-04-12) is admirably self-aware. Its “Core narrative” is positioned as authoritative for who Movemental serves and why—the shared condition of high responsibility and fragmented digital infrastructure. But the document explicitly cautions that the research supplement is structural context, not a license to treat internal synthesis as hard numbers. That distinction is healthy. It also implies that any slide deck pulling “market size” or “TAM counts” from adjacent folders should be cross-checked before it reaches a partner.

The credibility thesis document (02-credibility-thesis.md) and the convergence essay (26-the-convergence.md) provide the philosophical packaging: gatekeeping’s virtues and vices, the democratization of publishing, the AI shock as an epistemic crisis, and the pivot toward networked verification. They are strongest as sense-making, weaker where they repeat industry percentages without the Section 1 footnote apparatus.

External “docs” repository: lab, knowledge, intelligence

This audit’s deepest reads outside movemental focused on operations/strategy/movemental-ai (where the long credibility and scenius memos live), lab/papers/ai (a compact stack: literature review, citation network, paper profiles, source index), and knowledge/core-content/case-studies (seven movement histories). Together they show a bifurcation: formal AI research hygiene lives in lab/, while movemental theology and historical analogy live in knowledge/. Prompts 01 and 04 should pull from those folders respectively instead of collapsing them.

The intelligence tree (intelligence/leader-research/) and hundreds of operations documents were not exhaustively surfaced in this pass; the raw inventory flags them as the next expansion surface for Part 1–2 (especially named scholars and repeated statistics).

Movement leader research: rich qualitative ore, shaky verification

The movement_leader_research directory contains deep digital-presence reconstructions for many leaders. Their value is directional: where content lives, what genres recur, how networks might be drawn. Their weakness is systematic: “to be verified” appears constantly—URLs, episode counts, viewership, publisher metadata. That is not a scandal; it is a workflow state. It does mean that competitive positioning memos should not treat those files as evidence-grade until a verification pass lands.

Internal tensions worth resolving once, loudly

First, prevalence statistics drift between manuscript, convergence essay, and Section 1 roadmap. Align public copy with verified primaries and retire orphaned percentages.

Second, language around “network effects” can sound like Silicon Valley infinite scale until the reader hits bounded 100 in the guidebook. The ideas are compatible—Kelly’s “network effects of success” inside a scene is not the same as Metcalfe’s law—but newcomers will conflate them unless copy explicitly bridges depth over reach.

Third, publisher economics (85–90% to the house in the convergence narrative versus “90/10” rhetoric elsewhere) needs definitional precision: print vs digital, agency fees, advances, and rights reversions all change the split. Prompt 03 should normalize vocabulary before any pricing prompt (20) cites a “revolutionary” comparison.

What remains genuinely unknown

Even after all this thinking, several classes of question remain open in principle:

Measurement: Which peer-network structures measurably increase justified trust (not just clicks) for expert content in an AI-heavy SERP and LLM-citation environment?
Formation: What does durable transformation require in blended digital contexts—and what can platforms claim without overreach?
Psychometrics: Where APEST-style instruments sit relative to validated personality or vocational psychology (prompt 17).
Demography: Reliable gender and diversity baselines for “movement leadership” as a population (prompt 19).
Economics: What price and revenue-share anchors are comparables actually using (prompt 20)?

Those are the places where /academic-research and sibling skills earn their keep.

How subsequent prompts should use this foundation

Start from inventory, not a blank page. If a claim appears in credibility-how-it-works.md or the guidebook, cite that internal lineage before layering new studies.
Let Section 1 drive statistics. For any number touching AI prevalence, trust, fraud, or adoption velocity, the roadmap’s table is the first stop.
Let authoritative-sources enforce tiers. Do not answer a governance question with a lone vendor blog unless triangulated.
Treat leader dossiers as hypotheses until verification stamps replace “to be verified.”
Preserve the bounded-scale story while researching network effects—research should sharpen the distinction, not collapse it.

The argument index: where “proof” already carries sources

Beyond long-form essays, Movemental maintains a structured argument card corpus under docs/arguments/custom-gpt/, including messaging-06-proof.md for credibility and track-record claims. Cards carry YAML-style metadata: strength (for example high), sources (a count), and pointers to underlying files.

Two patterns matter for research hygiene.

First, high-strength cards can still embed fragile numbers. For example, proof cards summarize Alan Hirsch’s institutional biography (seven organizations, roughly twenty books, APEST language embedded across seminaries and denominations) and assert large-scale assessment usage (“over 150,000 people have taken the APEST assessment”). Those figures may be directionally right and useful for internal alignment, but they are the sort of claims prompts 17 (assessment science) and 14 (movement-scale statistics) should treat as verify or qualify, not as self-authenticating because a card says strength: high.

Second, technical proof claims compress a whole architecture story—database table counts, multi-tenant isolation, performance comparisons—into a single sentence. That is valuable for marketing coherence but dangerous if “60% faster than industry standard”-class phrases reach the public site without a defined benchmark. Prompt 12 (competitive landscape) and prompt 05 (discoverability) are natural homes for deciding which technical claims belong in proof versus engineering docs versus silent capability.

The meta-point: the argument index is routing and rhetoric, not a peer-reviewed appendix. It tells you what the organization already believes it can say out loud; it does not remove the obligation to ground the highest-stakes numbers.

Sibling repositories: what they add without duplicating

`movemental-ai`

The prompt lists “big six” root strategy documents. In practice, their content overlaps heavily with docs/operations/strategy/movemental-ai/ in the external docs repo—sometimes verbatim, sometimes evolved in parallel. For an agent, that redundancy is a feature: it signals which ideas survived contact with multiple writers and contexts. Credibility mechanics, scenius cultivation, network visualization narrative, and AI posture guides are not experiments; they are institutional memory.

When prompt 00 says not to run web research skills here, the point is methodological: you are mapping what humans already decided, not adjudicating truth. Subsequent prompts adjudicate truth.

`alan-hirsch`

This audit did not re-walk the entire _docs/_public/ tree or content-library/ article-by-article. At a high level, that repo’s value for prompts 01–21 is triangulation: voice, historical reception (book reviews, sentiment maps), and productized insights (pricing memos, home-page consultations) that may include citations worth importing into Movemental’s public research layer. Treat it as a secondary harmonization source: useful for tone and for spotting repeated metaphors, not automatically authoritative for cross-sector AI statistics.

The external docs `lab/papers/ai` stack (why it matters for prompt 01)

The lab/papers/ai directory is small on disk—literature review, citation network, paper profiles, source index—but it represents a different epistemic culture than the manuscript. Where the manuscript persuades, the lab folder traces. For prompt 01 especially, agents should read those files before generating yet another “AI trust crisis” memo from headlines alone. They are the closest thing in-repo to: “We already tried to organize the academic neighborhood around this topic.”

If lab conclusions disagree with Section 1’s chosen footnotes, resolve the conflict explicitly in the raw research file for prompt 01 rather than silently overwriting one layer with the other.

Historical case studies: prompt 04’s starting library

The seven case studies under knowledge/core-content/case-studies/ are not casual blog posts. They are structured movement histories (early church, Wesleyan/Methodist, Moravian, Celtic, Chinese underground church, Patrick, and related leader profiles). Prompt 04 should treat them as interpretive scaffolding—rich analogies with named leaders and institutional patterns—then ask where historiography has updated since those files were authored. The goal is not to “debunk” internal case studies but to place them in conversation with current scholarship and demographic data (prompt 14 overlaps when “movement growth” claims lean on historical parallels).

Cross-prompt dependencies (a compact coordination map)

Some gaps are shared. Trust and prevalence statistics touch prompts 01, 09, and parts of 05 (how people decide what to click in AI-mediated retrieval). Publisher splits and creator-economy economics touch 03 and 20. Voice authenticity touches 11 and the manuscript’s voice chapter, which in turn touches 02 (whether scenius-style mutual knowledge is a practical defense against mimicry). Gender and diversity (19) should be fed by the same careful Pew/PRRI discipline described in authoritative sources, not by anecdotal leader counts.

If each prompt agent only reads its own prompt file, you will redo work. If each agent starts from this inventory + Section 1 + authoritative-sources, you will still disagree—but you will disagree productively, with shared vocabulary.

What “honesty” looks like in the final research artifacts

Movemental’s brand promise is not “we have perfect data.” It is closer to: we know why trust broke, and we are building conditions where relational verification can matter again. That story can survive admitting that an early manuscript percentage was a placeholder, or that a leader dossier’s view count was an estimate, or that a competitive comparison needs a refreshed matrix for 2026 tools.

The worst failure mode is the opposite: polishing rhetoric while leaving known weak citations in place because they “sound right.” The corpus already contains the antibodies—Section 1’s checklist, authoritative-sources’ tier table, the guidebook’s refusal to fake graph density—if later agents choose to use them.

A reader’s sequence: if you only have time for ten files

Think of this as a compression tour of the intellectual foundation—not exhaustive, but enough to write and research without talking past the org.

docs/book-development/SECTION_1_RESEARCH_AND_REFERENCES.md — what still needs fetching for the book’s opening movement of claims.
docs/research/authoritative-sources-ai-nonprofits-faith-formation.md — how to keep church/nonprofit writing honest across evidence types.
/Users/joshuashepherd/Desktop/dev/repos/docs/operations/strategy/movemental-ai/credibility-how-it-works.md — the fullest articulation of credibility offline/online and SEO-as-topology.
/Users/joshuashepherd/Desktop/dev/repos/docs/operations/strategy/movemental-ai/SCENIUS-DEVELOPMENT-GUIDEBOOK.md — operationalized scenius and the bounded-scale commitments.
/Users/joshuashepherd/Desktop/dev/repos/docs/operations/strategy/movemental-ai/SCENIUS-VISUALIZATION-STORY-AND-2028.md — narrative and ethics of the network visualization.
docs/business-docs/core-docs/02-credibility-thesis.md — concise thesis packaging for founders and partners.
docs/business-docs/core-docs/26-the-convergence.md — long-form convergence of credibility, publication, scenius, and movement.
docs/business-docs/09_research_analysis/movemental-audience-comprehensive.md — who the audience is, and where not to overclaim.
docs/book-development/manuscript-ordered/01-the-credibility-crisis.md — the on-ramp voice of the book, with known statistics debt.
docs/arguments/custom-gpt/messaging-06-proof.md — how proof is staged publicly, including strength-rated cards.

If you read those ten, you will feel the organization’s “voice in bulk”—pastoral in the manuscript, analytic in strategy memos, cautious in the sources index, aspirational in proof cards—and you will see where each voice expects the others to compensate.

Churches, nonprofits, and “AI”: formation claims deserve slower lanes

One of the more mature features of authoritative-sources-ai-nonprofits-faith-formation.md is its insistence that screens ≠ generative AI as evidence. Much public anxiety (and some advisory literature) routes through adolescent mental health and social media ecosystems. That is relevant to Movemental’s audiences, but it is not a free substitute for claims about LLM-mediated formation. The router’s warning against collapsing layers is especially important for prompt 06 (formation vs information) and prompt 21 (nonprofit system builds): pastors and executive directors rightly smell bad faith when a vendor slide uses “youth mental health” statistics to sell a chatbot.

The corpus’s settled posture here is methodological: pair survey facts with pastoral ethics, label each. That pairing is already a kind of “answer” to prompt 08 (theological anthropology) at the level of discourse ethics—even before you import Augustine, transhumanist critics, or Catholic social thought. Later theological research should enrich the canon, not pretend Movemental lacked a stance on how to speak.

Speculative vs sourced: a blunt typology for collaborators

To reduce ambiguity when someone asks, “Is that true?”—internally, Movemental already runs at least four epistemic modes:

Sourced-and-stable: Items in Section 1’s footnote list once verified against primaries (Pew reports, Zenodo record, ACM paper). These can travel externally with confidence if dates and scopes are respected.
Sourced-but-secondary: Industry blogs (Ahrefs, Graphite-style) and aggregator “statistics” pages. Useful, but easy to overfit; treat as directional unless reproduced.
Plausible narrative: Much of the convergence essay’s economic history and cultural diagnosis—internally coherent, rhetorically strong, but not always pinned to a specific dataset. Good for leadership sense-making; risky if quoted as social science.
Explicitly speculative: Product futures, “2028 at 100” visualization storytelling, and some competitive positioning. Valuable for alignment; must be labeled as scenario not forecast when it leaves Slack.

Research prompts 01, 09, 12, and 20 exist largely to upgrade items in the bottom two categories without embarrassing the top two.

Closing honesty

The corpus is asymmetrically strong: conceptual clarity and strategic coherence are ahead of footnote density and primary-data verification. That is normal for a pre-scale organization building in public. The research prompt sequence is designed to harden the weak layer without bulldozing the strong one.

If Movemental’s public voice is to remain trustworthy in an age of synthetic fluency, the work ahead is less about discovering a new thesis than about earning the right to keep repeating the one you already have—with numbers and names that survive contact with hostile readers.

That is what “additive research” means here: more primary sources, fewer duplicated memos, and a book whose claims land as carefully as its pastoral intent already does.