Hypotheses Index¶

This directory holds committed scientific hypotheses in Falsification Card form. Each file is a self-contained artifact: the claim, the assumptions it rests on, the killshot menu ranked by information weight per dollar, pre-committed thresholds for Alive / Killed / Pending / Retracted, kill switches, execution log, and survival score.

The schema and the rationale live in linter-design.md. This page is the entry point and the index.

What "Committed" Means¶

A hypothesis is committed when it is written to a file in this directory with a first git commit. That commit is the pre-registration: the claim, the initial assumption stack, and the killshot menu are frozen as of that SHA. Any subsequent edit creates a new commit with a visible diff. This is the git-native equivalent of the brief's call for {claim, killshots, thresholds} hashing — see linter-design.md §6.

Convention: the hypothesis file's first commit is the load-bearing one. Subsequent commits carry rationale in the commit message (e.g., H01: tighten lactoferrin threshold from 100 to 500 mg/L — Ward 1995 submerged baseline is 2 g/L, 500 mg/L is the Phase B acceptance floor). The git log is the audit trail.

Committing to a hypothesis is a commitment to attempt to falsify it, not a commitment to defend it. A hypothesis that survives zero killshots after a year isn't defended; it's pruned or retracted.

File Naming¶

Pattern: H0N-slug.md where N is zero-padded two-digit (H01, H02, …, H99). Three digits (H001, H002) when the collection exceeds 99.
slug is short, hyphenated, mnemonic. Prefer 2–4 words. Example: H01-ward-dual-cassette.md.
No dates, no authors, no status in the filename. Those live in frontmatter (committed:, status:) and mutate over time. Filenames are stable identifiers.

Current Hypotheses¶

ID	Title	Status	Committed	Link
H01	Ward 1995 glucoamylase-KEX2 dual-cassette architecture layers uricase + lactoferrin in A. oryzae solid-state rice koji	Pending	2026-04-24	H01-ward-dual-cassette.md
H02	Engineered Faecalibacterium prausnitzii delivering colonic butyrate is a viable peer-track therapeutic chassis to engineered koji for gout, with genotype-agnostic ABCG2 induction (WT + Q141K) as the primary mechanism of action	Stub	2026-05-05	H02-engineered-lbp-thesis.md
H03	Kidney-tropic siRNA against URAT1 mRNA is a viable long-horizon therapeutic modality for under-excreter gout, mechanistically cleaner than small-molecule URAT1 inhibitors but gated on the maturation of kidney-tropic conjugate delivery chemistry	Stub	2026-05-05	H03-sirna-urat1-thesis.md
H04	Modern scientific rigor (chokepoint-mapping + ChEMBL cross-check + bioavailability honesty + falsification card discipline) applied to TCM materia medica produces actionable, gout-relevant findings beyond what reductionist single-compound analysis or holistic "alternative medicine" framings produce in isolation	Stub	2026-05-05	H04-tcm-rigor-intersection.md
H05	An A. oryzae-engineered soluble DAF/CD55 SCR1-4 truncated construct (aa 35-285) is a viable fermentable CP0 modulator that closes the platform's documented complement-priming gap	Stub	2026-05-05	H05-daf-scr14-cp0-thesis.md
H06	Open Enzyme can produce reproducible, bioactivity-validated medicinal-mushroom-complement extracts (GLPP, cordycepin, ergothioneine) at quality sufficient to replicate published in vivo effect sizes within 2× — making the medicinal-mushroom-complement track a viable peer to the koji-engineering track	Stub	2026-05-06	H06-medicinal-mushroom-complement-track.md
H07	Clomid's serum-UA-elevating effect in gout-prone men is mediated primarily by intestinal estrogen-receptor antagonism (blocking the female-positive PI3K/Akt → ABCG2 induction signal), not by direct androgen-receptor effects on intestinal ABCG2	Stub	2026-05-07	H07-clomid-intestinal-er-antagonism.md
H08	The gut-lumen uricase sink produces a clinically meaningful serum urate reduction in a typical (non-CKD) gout cohort, with effect size in the −0.5 to −1.0 mg/dL band predicted by the comp-019 flux model — the platform's #1 riskiest assumption (mechanism)	Stub	2026-05-15	H08-gut-lumen-sink-platform-thesis.md
H09	Home- and community-fermented engineered koji can reliably deliver therapeutic doses of multi-cassette protein with batch-to-batch consistency sufficient for a chronic disease intervention, under the Community-BioLab + Home-Fermentation hybrid model — the platform's #2 riskiest assumption (production/delivery)	Stub	2026-05-15	H09-community-fermentation-reliability.md

(Add new rows as hypotheses are committed. Sort by ID.)

Lifecycle¶

Commit. Write the Falsification Card (claim, assumption stack, killshot menu, thresholds, kill switches, empty log) to H0N-slug.md. Git commit. Status: Pending.
Lint. Run Falsification Lint (manual / agent invocation) against the committed file. The linter reads the hypothesis + cross-referenced wiki pages, may propose additional killshots or assumptions, and may suggest the cheapest-upstream redirect per the brief's hero interaction. The author decides what to adopt; adoption creates a new commit with the diff visible.
Test. Execute killshots in the order the menu dictates — cheapest, highest-info first. Record results inline in the Log table. Each executed killshot updates status and survival_count.
Log outcome. For each killshot: date, killshot identifier, outcome (killed / survived / ambiguous), notes, and any deviations from the pre-committed protocol. Survival-count and survival-score are updated per linter-design.md §6.
Update status. Pending → Alive on first survived killshot. Pending or Alive → Killed on first threshold-crossing kill. Status can move Alive → Killed if a later killshot crosses a threshold; it can move Killed → Alive only via Retraction (§ below).
Retract if needed. If a killshot result is later invalidated (bad control, assay artifact, substrate batch problem), document the retraction in the file's Retraction History section and recompute status based on remaining valid killshots. Status: Retracted while the recount is pending; resolves to Alive / Killed / Pending on re-evaluation.

The lifecycle is sequential per hypothesis but not per directory. Multiple hypotheses can be at different lifecycle stages simultaneously.

Status Meanings¶

Pending. Committed, no killshot yet executed. The default initial state.
Alive. At least one killshot has been run; the claim has not been falsified per its pre-committed thresholds. Survival count > 0.
Killed. At least one killshot has crossed its Killed threshold. The claim is falsified. Documented in the Log with the killshot result.
Retracted. A previously-counted killshot result has been invalidated. Status is being recomputed; intermediate holding state.

A hypothesis with status Killed is not deleted. The file stays; the killed claim is part of the wiki's epistemic history. A Killed hypothesis can motivate a new committed hypothesis (a revised claim at a new file, H0M-*), but the Killed file itself is preserved.

Relationship to `synthesis/` (architecture: synthesis/README.md) and `wiki/open-questions.md`¶

synthesis/ is the generative action queue — new dot-connections, proposed experiments, cross-doc findings. It is promiscuous, fast-moving, and pruned manually by Brian. Items in synthesis/ are candidates for hypothesis promotion; not all candidates graduate.
open-questions.md is a cross-wiki index of unresolved questions. Some of those questions map to committed hypotheses; most don't. An open question is a prompt for investigation; a committed hypothesis is a specific falsifiable claim plus a plan to attempt falsification.
hypotheses/ is where a specific, falsifiable claim gets a Falsification Card and a commit. Not everything in synthesis or open-questions becomes a hypothesis — only the ones the user is ready to commit resources against.

Flow (typical): synthesis/ finding → open-questions.md entry → (if worth testing) → hypotheses/H0N commit → Falsification Lint → execute killshots → update status.

Conventions Summary¶

One file per hypothesis. Filename is the stable identifier.
First commit is the pre-registration. Subsequent edits are visible diffs with rationale in commit message.
Status is frontmatter (status:). Survival count is frontmatter (survival_count:).
Killshot menu is a table, sorted by score descending.
Thresholds are pre-committed — declared before the first killshot runs.
Failure modes are tagged from linter-design.md §5.
Retraction history is in-line, not in a separate directory. Visibility is the point.