contrib → main: Organization Science Stage 1 reviewed package and irreducibility-pipeline

.github/workflows/update-directory.yml

@@ -1,9 +1,9 @@
 name: update-directory
 
-# Keeps the generated surfaces current: the directory in README.md, MAP.md, and the research-watch
-# INDEX.md. On a pull request from this repo, it regenerates all three and commits the result back to
+# Keeps the generated surfaces current: the directory in README.md, MAP.md, catalog.json, and the
+# research-watch INDEX.md. On a pull request from this repo, it regenerates them and commits the result back to
 # the PR branch, so a contributor who adds an experiment does not have to remember to run the
-# generators. The `directory-current`, `map-current`, and `research-index-current` checks in
+# generators. The `directory-current`, `map-current`, `catalog-current`, and `research-index-current` checks in
 # ci/reproduce.json are the safety net: on fork PRs (where this job cannot push) a stale file fails CI,
 # and the contributor runs the generator themselves.
 
@@ -34,21 +34,22 @@ jobs:
           python-version: "3.12"
 
       # The generators use only the standard library, so no dependency install is needed.
-      - name: Regenerate the directory, the map, and the research index
+      - name: Regenerate the directory, the map, the catalog, and the research index
         run: |
           python tools/build_index.py
           python tools/build_map.py
+          python tools/build_catalog.py
           python org_frontier/research/build_research_index.py
           python org_frontier/research/build_cards_index.py
 
       - name: Commit if any generated surface changed
         run: |
-          if git diff --quiet -- README.md MAP.md org_frontier/research/INDEX.md org_frontier/research/*/literature/REFERENCES.md; then
+          if git diff --quiet -- README.md MAP.md catalog.json org_frontier/research/INDEX.md org_frontier/research/*/literature/REFERENCES.md; then
             echo "Directory, map, research index, and reference-card indexes already current."
             exit 0
           fi
           git config user.name "github-actions[bot]"
           git config user.email "41898282+github-actions[bot]@users.noreply.github.com"
-          git add README.md MAP.md org_frontier/research/INDEX.md org_frontier/research/*/literature/REFERENCES.md
-          git commit -m "docs: regenerate README directory, MAP, research index, and reference-card indexes [skip ci]"
+          git add README.md MAP.md catalog.json org_frontier/research/INDEX.md org_frontier/research/*/literature/REFERENCES.md
+          git commit -m "docs: regenerate README directory, MAP, catalog, research index, and reference-card indexes [skip ci]"
           git push origin HEAD:${{ github.head_ref }}

.gitignore

@@ -60,6 +60,7 @@ spaces/
 
 # APA Summit 2027 phygital calls: the editors' PDFs, page copies and extracted text stay local
 submissions/phygital_*/cfp/source/
+submissions/exetasis_sovereignty/cfp/source/
 
 # Calls-for-papers raw pages for every submission arm (public repo; refetchable)
 submissions/*/cfp/source/

AGENTS.md

@@ -23,7 +23,9 @@ points back here for the shared rules.
 2. **[`MAP.md`](MAP.md)** — the one-screen index: instrument, computational lab, empirical arms,
    submissions, machinery, live counts. Generated, always current.
 3. **[`OVERVIEW.md`](OVERVIEW.md)** — what the lab argues and where it actually stands. Then this file
-   for the operating rules.
+   for the operating rules. **[`OPEN_QUESTIONS.md`](OPEN_QUESTIONS.md)** is the registry of open threads,
+   and `python tools/build_catalog.py --find "<keywords>"` places a question against everything already
+   asked.
 
 The repository is large (hundreds of directories and dozens of READMEs). `MAP.md` and the generated directory in
 `README.md` are the cheap ways in. Reach for a README in the specific subtree only once the map has
@@ -56,12 +58,16 @@ pip install -r requirements.txt                # pulls PyPhi + phyid from git: n
 ```
 
 Run everything from the repo root so `org_frontier.*` and `foundations.*` resolve. Confirm the instrument
-before trusting any result — the control probes assert that a known form reproduces its known verdict:
+before trusting any result — the controls assert that a known form reproduces its known verdict:
 
 ```bash
-python ci/reproduce.py            # re-derive every registered number; the instrument control is core
+python -m org_frontier.classifier.validate     # must print "Instrument validated" (about two seconds)
+python ci/reproduce.py instrument-controls     # the same gate as CI runs it
 ```
 
+`python ci/reproduce.py` with no arguments is the full manifest (hundreds of checks, some of them
+minutes each); that is the nightly job, not the confirm step.
+
 For large cheap-feature panels, the lab-default **margin cascade** screens with a Probe-131 RF and
 calls exact Φ only on the top **B=10%** most uncertain forms (OOF `|p−0.5|`, recomputed per panel):
 
@@ -76,14 +82,15 @@ See [`org_frontier/cascade/README.md`](org_frontier/cascade/README.md). Do not f
 
 ## Verify before you land
 
-Three generators keep the navigable surface current, each with a `--check` that CI gates. Run the ones
-your change touched, then confirm all three are clean:
+Four generators keep the navigable surface current, each with a `--check` that CI gates. Run the ones
+your change touched, then confirm all four are clean:
 
 ```bash
 python tools/build_index.py --check                       # the README directory
 python tools/build_map.py --check                         # MAP.md
+python tools/build_catalog.py --check                     # catalog.json, the searchable index
 python org_frontier/research/build_research_index.py --check   # the research watch index
-python ci/reproduce.py                                    # every number reproduces from its script
+python ci/reproduce.py <check-name> …                     # the numbers your change registered
 ```
 
 If a `--check` reports stale, run the same command without `--check` to regenerate, then commit the
@@ -99,7 +106,8 @@ shortest version:
 | Adding… | Lands under | Start from |
 |---|---|---|
 | a probe | `org_frontier/probes/probe_<slug>.py` | the global numbering in `org_frontier/probes/PROBES.md` |
-| a full question | `org_frontier/questions/q<NN>_<slug>/` | `org_frontier/protocol/template/` or the pipeline |
+| a full question | `org_frontier/questions/q<NN>_<slug>/` | `python tools/build_catalog.py --find …`, then `python -m org_frontier.protocol.new_question` |
+| an open thread | wherever its row in `OPEN_QUESTIONS.md` points | set the row to `claimed` in the same PR |
 | a qualitative study | `org_frontier/qualitative/<slug>/` | `org_frontier/qualitative/template/` |
 | a recurrence experiment | `org_frontier/recurrence/` | `crqa.py` and the existing Φ harnesses |
 | a survey study | `org_frontier/survey/<study>/` | the arm and its first study in `org_frontier/survey/README.md` |
@@ -122,16 +130,24 @@ gh pr create --base contrib --title "…" --body "…"
 # CI green → merge into contrib → open the contrib→main promotion PR → merge → delete the branch
 ```
 
+From a fork, `origin` is the fork: add this repository as `upstream`, branch off `upstream/contrib`,
+and open the PR with `--repo rogerSuperBuilderAlpha/algorithmacy-lab` (the recipe is in
+[`GETTING_STARTED.md`](GETTING_STARTED.md#8-open-a-pull-request)). The `update-directory` job cannot push
+to a fork, so regenerate the indexes locally before pushing.
+
 ## Definition of done — the gate before opening a PR
 
 - The instrument was validated on a known control before any comparison.
 - Numbers come from a committed script and are registered in `ci/reproduce.json`; `python ci/reproduce.py`
   passes locally.
-- All three index `--check`s are clean (`build_index.py`, `build_map.py`, `build_research_index.py`).
-- Hypotheses were fixed before computing, and nulls and refutations are reported as results.
+- All four index `--check`s are clean (`build_index.py`, `build_map.py`, `build_catalog.py`,
+  `build_research_index.py`).
+- Hypotheses were fixed before computing (`python ci/precommit_audit.py origin/contrib..HEAD` shows the
+  order), and nulls and refutations are reported as results.
 - Every result states the validation gap: it is evidence about the model, not about a real organization.
-- Prose follows the house style in [`CLAUDE.md`](CLAUDE.md) — no first person, plain declarative
-  sentences, and the de-slop self-check passed.
+- Prose follows the house style in [`CLAUDE.md`](CLAUDE.md), its current rules: claim-first paragraphs,
+  named agents in active voice, first person only for authorial labor, no emphasis-marker openers, and
+  the flow self-check passed.
 - Citations resolve to real DOIs / arXiv IDs / stable URLs; nothing is fabricated to fill a gap.
 - The git target was confirmed with `git rev-parse --show-toplevel` before committing.
 

CONTRIBUTING.md

@@ -57,8 +57,11 @@ input to a verdict.
   strict numeric order, so there is no scaffolder for a plain probe number — take the next free one as the
   current maximum plus one:
   `grep -oE '^\| [0-9]+' org_frontier/probes/PROBES.md | grep -oE '[0-9]+' | sort -n | tail -1`.
-- A full question goes through the pipeline and lands under `org_frontier/questions/q<NN>_<slug>/`. Copy
-  [`org_frontier/protocol/template/`](org_frontier/protocol/template/), or run the orchestration.
+- A full question goes through the pipeline and lands under `org_frontier/questions/q<NN>_<slug>/`. Place
+  it first with `python tools/build_catalog.py --find "<keywords>"` (the nearest prior questions, probes,
+  and studies), then scaffold it with `python -m org_frontier.protocol.new_question --slug <slug>`, which
+  copies [`org_frontier/protocol/template/`](org_frontier/protocol/template/) and names the probe stub.
+  Open threads waiting for someone are in [`OPEN_QUESTIONS.md`](OPEN_QUESTIONS.md).
 - Reuse the shared infrastructure rather than rebuilding it: `org_frontier/classifier/`,
   `org_frontier/probes/lib.py`, `org_frontier/probes/_info.py`, `foundations/proxy_audit/exact_phi.py`.
 
@@ -114,15 +117,16 @@ and leaves it to the nightly job.
 A contribution that registers no number — a qualitative or survey study — still triggers the `core`
 instrument control on the per-PR gate, and that control needs a working PyPhi environment (Python 3.10+).
 CI runs it for you on every PR, so a contributor without a compute environment verifies locally with the
-three index `--check`s (`build_index.py`, `build_map.py`, `build_research_index.py`) and lets CI run the
-control. The study stands on `directory-current`: once `python tools/build_index.py` indexes it, the gate
+four index `--check`s (`build_index.py`, `build_map.py`, `build_catalog.py`, `build_research_index.py`)
+and lets CI run the control. The study stands on `directory-current`: once `python tools/build_index.py` indexes it, the gate
 confirms it is listed.
 
 ## Prose
 
-All prose follows the house style in [`CLAUDE.md`](CLAUDE.md): no first person, plain declarative
-sentences, section titles as short noun phrases, and a de-slop pass (cut the antithesis machine,
-self-narration of rigor, and metronomic openers). Citations resolve to real sources.
+All prose follows the house style in [`CLAUDE.md`](CLAUDE.md), its current rules: claim-first paragraphs,
+named agents in active voice, first person only for authorial labor (I searched, I coded), no sentence
+commenting on the paper's own rigor, no emphasis-marker openers, section titles as short noun phrases, and
+the flow self-check run on each paragraph. Citations resolve to real sources.
 
 ## Git and the two repositories
 

GETTING_STARTED.md

@@ -45,10 +45,13 @@ every probe you write repeats it.
 ## 3. Run an existing experiment
 
 ```bash
-python -m org_frontier.probes.probe_conjunctive_law
+python -m org_frontier.probes.probe_control          # Probe 7, about two seconds
 ```
 
-A "probe" is one script that tests one hypothesis with exact Φ and prints a verdict. The log of
+A "probe" is one script that tests one hypothesis with exact Φ and prints a verdict. Probe 7 tests the
+thesis's own clause: a mediator that follows one party alone reads dyadic, mutual veto and either-forces
+read triadic. Exact Φ grows fast with node count, so check a probe's FINDINGS or docstring for its
+runtime before running it; `probe_conjunctive_law` goes to seven nodes and takes about nine minutes. The log of
 every probe run is `org_frontier/probes/PROBES.md`; the worked end-to-end question is
 `org_frontier/questions/q43_thompson_interdependence/`. Read one of each before writing your own.
 
@@ -75,20 +78,37 @@ The shared API lives in `org_frontier/classifier/classifier.py` (`classify_rules
 `max_phi_float`), and `foundations/proxy_audit/exact_phi.py` (`exact_big_phi`, `reachable_states`).
 Exact Φ is feasible up to roughly 10–12 nodes; keep forms small.
 
-## 5. Scaffold a new question
+## 5. Place your question, then scaffold it
+
+The lab has asked many questions already, and the protocol's first stage is to place a new one against
+them. Search the generated catalog with your question's keywords:
+
+```bash
+python tools/build_catalog.py --find "majority quorum mediator"
+```
+
+It prints the nearest prior questions, probes, studies, and open threads, each with its one-line
+finding and reproduce command. If a row answers your question, cite it and stress it instead of
+re-asking. If a row answers a neighbor, your study is the residual. If nothing is close, scaffold it.
+If you have no question of your own yet, [`OPEN_QUESTIONS.md`](OPEN_QUESTIONS.md) lists what is
+waiting for someone, with what each thread needs and its first command; set a row to `claimed` in the
+pull request that takes it.
 
 ```bash
 python -m org_frontier.protocol.new_question --slug your_slug \
     --question "Your one-line research question"
 ```
 
-This copies `org_frontier/protocol/template/` to `org_frontier/questions/q<NN>_<slug>/`, fills in
-the next free question number and probe number, and prints the next steps. Read
+This copies `org_frontier/protocol/template/` to `org_frontier/questions/q<NN>_<slug>/`, names the
+probe stub `probe_<slug>.py`, creates `results/` and `literature/`, fills in the next free question
+number and probe number, and prints the next steps. Read
 `org_frontier/protocol/RESEARCH_PROTOCOL.md` for what each artifact (`review.md`, `hypotheses.md`,
 `methods.md`, the `probe_*.py` scripts, `FINDINGS.md`, `paper.md`) must contain.
 
 The one rule that makes the record credible: **commit `hypotheses.md` before you compute the
-results.** The git history is the evidence that the claims were fixed first.
+results.** The git history is the evidence that the claims were fixed first;
+`python ci/precommit_audit.py origin/contrib..HEAD` reports the order for every question directory
+your branch touches.
 
 For qualitative research — reading a real coordination setting against the priors through interviews,
 observation, or document work — start instead from
@@ -131,31 +151,40 @@ python ci/reproduce.py your-check-name      # one check
 python ci/reproduce.py                       # the full manifest
 ```
 
-## 7. Refresh the README directory
+## 7. Refresh the generated indexes
 
 The README carries a generated directory of every question, study, essay, and review
-(`tools/build_index.py`). Regenerate it so your work shows up:
+(`tools/build_index.py`), and `catalog.json` is the searchable index the placement step reads.
+Regenerate both so your work shows up:
 
 ```bash
 python tools/build_index.py
+python tools/build_catalog.py
 ```
 
-A pull request from this repo regenerates it automatically; the `directory-current` CI check
-fails if it is stale, so run it before pushing (it commits a one-line change to README.md).
+A pull request from this repo regenerates them automatically; a pull request from a fork cannot, and
+the `directory-current` and `catalog-current` CI checks fail if either is stale, so run both before
+pushing.
 
 ## 8. Open a pull request
 
 Work lands through `contrib`, never `main` directly. Read `REPO_LAYOUT.md` first if you have not —
 this tree nests a second, private git repo under `dissertation/`, and the directory you run git
 from decides which repo you touch.
 
+Working from a fork, add this repository as `upstream` once, then branch off its `contrib`:
+
 ```bash
-git checkout contrib && git pull
-git checkout -b study/your-slug
+gh repo fork rogerSuperBuilderAlpha/algorithmacy-lab --clone   # or fork on GitHub and clone yours
+cd algorithmacy-lab                  # --clone already adds `upstream`; a hand-made clone needs `git remote add upstream https://github.com/rogerSuperBuilderAlpha/algorithmacy-lab.git`
+git fetch upstream contrib
+git checkout -b study/your-slug upstream/contrib
 git add <your files> && git commit -m "..."   # hypotheses committed before results
 git push -u origin study/your-slug
-gh pr create --base contrib
+gh pr create --repo rogerSuperBuilderAlpha/algorithmacy-lab --base contrib
 ```
 
+With write access to this repository, `origin` is `upstream` and the first two lines are not needed.
+
 `PUBLISHING.md` describes the four-pass review your PR will get (reproduction, pre-commitment
 audit, argument, prose) and `CONTRIBUTING.md` covers house style.

MAP.md

@@ -15,6 +15,7 @@ Start here. The full directory lives in [`README.md`](README.md); the operating
 | [`AGENTS.md`](AGENTS.md) | Operating manual for agents: map, run/verify, git rule, land flow, done-checklist |
 | [`README.md`](README.md) | What the lab is, the thesis, setup, and the full generated directory |
 | [`OVERVIEW.md`](OVERVIEW.md) | Five-minute orientation: the argument, where it stands, where to contribute |
+| [`OPEN_QUESTIONS.md`](OPEN_QUESTIONS.md) | The open threads with a status column: what is waiting for someone, and what each needs |
 | [`GETTING_STARTED.md`](GETTING_STARTED.md) | Hands-on: environment → confirm the instrument → scaffold → register → PR |
 | [`CONTRIBUTING.md`](CONTRIBUTING.md) | Code, probe, study, and house-style conventions a submission must meet |
 | [`REPO_LAYOUT.md`](REPO_LAYOUT.md) | The two-repo git trap: this repo vs the nested private dissertation |
@@ -87,6 +88,7 @@ Run the relevant generator after adding content, then verify with its `--check`.
 |---|---|---|
 | [`tools/build_index.py`](tools/build_index.py) | `python tools/build_index.py` | the generated directory in README.md |
 | [`tools/build_map.py`](tools/build_map.py) | `python tools/build_map.py` | this file, MAP.md |
+| [`tools/build_catalog.py`](tools/build_catalog.py) | `python tools/build_catalog.py` | catalog.json — the searchable index of questions, probes, studies, and open threads (`--find`) |
 | [`org_frontier/research/build_research_index.py`](org_frontier/research/build_research_index.py) | `python org_frontier/research/build_research_index.py` | the research watch INDEX.md |
 | [`ci/reproduce.py`](ci/reproduce.py) | `python ci/reproduce.py` | every registered number in ci/reproduce.json |
 
@@ -101,3 +103,4 @@ Run the relevant generator after adding content, then verify with its `--check`.
 | 9 | foundations experiments | `foundations/*/` with a `FINDINGS.md` |
 | 10 | literature reviews | `org_frontier/reviews/*/` with a `FINDINGS.md` |
 | 773 | watch entries (program-level) | `org_frontier/research/*/literature/references.bib` |
+| 15 | open threads | `OPEN_QUESTIONS.md` rows with status `open` |