protocol

Protocol — entry point

A fixed pipeline for taking a research question to a finished quantitative paper, plus the machinery to run it and the work that validated it.

What is here

• RESEARCH_PROTOCOL.md — the six-stage per-question pipeline (review, deep research, five hypotheses, methods, run, paper), the quality gates, and the reusable infrastructure each stage uses. The authoritative spec.
• template/ — the per-question directory template (review.md, hypotheses.md, methods.md, paper.md, FINDINGS.md, probe_template.py). Copy it to questions/q<NN>_<slug>/ to run a question by hand.
• question_pipeline.js — the Workflow script that runs all six stages automatically for one question. Manual execution by the template is the fallback whenever the automated run is brittle.
• methodology_review.md — a Stage-1 review of the protocol itself against established norms.
• literature/deep_research_report.md — a Stage-2 deep-research run on the methodology, with references.bib. Key finding: the per-hypothesis method plan (Stage 4), not the bare pre-commitment (Stage 3), is what the evidence shows buys credibility (Brodeur 2024).

Running a question automatically

Invoke the Workflow tool with the script and the question as args:

Workflow({
  scriptPath: "org_frontier/protocol/question_pipeline.js",
  args: { id: "16", slug: "two_coupled_triads",
          question: "<full question text>", startProbe: <next free probe number> }
})

The script writes the question's directory under org_frontier/questions/, calls the deep-research workflow inline for Stage 2, fixes five hypotheses, designs and runs the probes against the exact-Φ instrument (validating a known control first), appends the result rows to probes/PROBES.md, and assembles the paper with a de-slop pass. It returns the directory path, the hypotheses, and the per-hypothesis verdicts.

The pilot

questions/q43_thompson_interdependence/ is the first question run end to end through the automated pipeline: whether the dyadic/triadic verdict reproduces Thompson's (1967) pooled/sequential/reciprocal interdependence ordering. Result: four hypotheses confirmed, one refuted (probes 135–139). The ordering breaks, two of the three types are verdict-ambiguous on a single edge, and the proposed two-primitive replacement fails. The run validated the orchestration with no manual fallback.

The public write-up

../essays/studying_algorithmacy.md lays out the method this lab runs, grounded in the methodology research and the 134-probe track record, with the Q43 pilot as the worked example.

The queue

../RESEARCH_AGENDA_50_V2.md holds 50 questions. Q43 is done; the rest are pending.