Repository navigation
Researcher harness: fix the forker's path and add question placement - #856
Merged
Merged
Conversation
…iterature/; portable run commands new_question.py left the probe stub as probe_template.py while writing probe_<slug> into the FINDINGS.md reproduce line and the probe docstring, so the CI entry a newcomer copied did not resolve. It now renames the stub, creates results/ and literature/, and prints the real module. RESEARCH_PROTOCOL.md, the template, and question_pipeline.js ran under the owner's laptop venv path and named a branch that no longer exists; they now run `python -m …` from the repo root, name the shared infrastructure with its package prefixes, and state the current house style. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H7Xkmpzs9ewHjVmrjPYwYV
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H7Xkmpzs9ewHjVmrjPYwYV
…ing it tools/build_catalog.py generates catalog.json, one row per question, probe, study, foundations experiment, and open thread, with the one-line finding, the registered reproduce commands, and a runtime class; --find prints the rows nearest to a newcomer's question. OPEN_QUESTIONS.md is the registry of open threads with a status column and what each needs; MAP.md counts its open rows and lists the generator. The update-directory job regenerates catalog.json and catalog-current gates it. ci/precommit_audit.py reports, for a commit range, whether hypotheses.md landed before FINDINGS.md and results/ in each question directory. The manifest registers probes #7 (the first probe START_HERE runs), #57 and #116 (slow). q43 gains a literature/ note explaining the missing Stage 2 artifact. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H7Xkmpzs9ewHjVmrjPYwYV
…pe, one style rule START_HERE runs probe_control (two seconds) instead of the nine-minute conjunctive law, adds Phase 3.5 (place the question against the catalog: answered, adjacent, or open), routes the data-minded to DATA_SOURCES.md and the handoff packets, and sends 'find something new' to OPEN_QUESTIONS.md. GETTING_STARTED gains the placement step and the fork remote recipe. AGENTS confirms the instrument with classifier.validate rather than the full manifest, lists the fourth generator, and adds the fork note. AGENTS, CONTRIBUTING, PUBLISHING carry one style sentence pointing at CLAUDE.md's current rules. OVERVIEW, org_frontier/README, and protocol/README drop the drifted counts and the stale 'not yet run' and 'rest pending' lines. NOW.md is refreshed to 27 October. README's setup block warns about Python 3.10 and validates; llms.txt lists START_HERE and OPEN_QUESTIONS. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H7Xkmpzs9ewHjVmrjPYwYV
This was referenced Oct 27, 2026
…dit, fork recipe - build_catalog: match a probe's checks on the exact module after `-m` instead of a substring, so #67 probe_voting no longer takes probe_voting_robustness's reproduce command; regenerate catalog.json. - build_catalog: guard the SIGPIPE handler with hasattr (absent on Windows). - precommit_audit: list only added files (--diff-filter=A), as the docstring says, and count `results/` directories at any depth. - GETTING_STARTED §8: `gh repo fork --clone` already adds `upstream`; drop the redundant remote add and cd into the clone. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01H7XkmoHkqsYk7uTDpv8hp8
This was referenced Oct 27, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What this fixes
An end-to-end walk of the harness as a forker's AI would run it (fresh Python 3.12 venv, pinned PyPhi, every documented command) found four defects on the critical path and one missing step. This PR fixes them. Nothing in a published number changes.
Critical-path defects
new_question.pynamed the probeprobe_<slug>in the FINDINGS.md reproduce line and the docstring but left the file asprobe_template.py, so the CI entry a newcomer copies did not resolve. It now renames the stub, createsresults/andliterature/, and prints the real module. The template placeholder that collided with the probe number (<N>) is renamed.~/iit-playground/venv-4.0/bin/python, and the protocol said to push toorigin/worktree/org-frontier. They now saypython -m …from the repo root and thecontribland flow. The pipeline'sproxy_audit.*import andorg_frontier/CLAUDE.mdreference are corrected.probe_conjunctive_lawas the first probe: it computes exact Φ to seven nodes and takes about nine minutes. They now runprobe_control(Probe 7, two seconds), with the conjunctive law offered later with its runtime.python ci/reproduce.py(the full manifest, the nightly job). It now saysclassifier.validateorreproduce.py instrument-controls.The missing step: placing a question
tools/build_catalog.pygeneratescatalog.json: one row per question, probe, study, foundations experiment, and open thread, with the one-line finding, registered reproduce commands, and a runtime class.--find "<keywords>"prints the nearest rows. It has a--check,catalog-currentgates it, and theupdate-directoryjob regenerates it.OPEN_QUESTIONS.mdis the registry of open threads with a status column (open/claimed/answered) and what each needs (compute/data/tooling/review), seeded from OVERVIEW, the Wave-10 parked designs, and the ergodicity residual. MAP.md counts its open rows.DATA_SOURCES.md. GETTING_STARTED, OVERVIEW, CONTRIBUTING, PUBLISHING, and the protocol's Stage 1 point at it.Also
ci/precommit_audit.pyreports, for a commit range, whetherhypotheses.mdlanded beforeFINDINGS.mdandresults/in each question directory (PUBLISHING pass 2).literature/note on the missing Stage 2 artifact.Not in this PR: the classifier package's README and docstrings carry the same laptop path; a change under
org_frontier/classifier/triggers the full non-slow reproduce set, so that is a separate small PR.Reviewer commands
All of the above pass locally on the pinned PyPhi (
1.2.1.dev1470+gb78d0e342), and the per-PR selection (--changed-file) picks 11 checks, all green.Generated by Claude Code