Affected Repositories

• PyAutoCortex (primary)
• PyAutoMind (ledger)
• /mnt/c/Users/Jammy/Science/ (science tree — the Cortex local_path exception)

Branch Survey

Suggested branch: feature/slope-hierarchy-scale-birth

Implementation Steps

1. Preflight git -C Science/slope_hierarchy status --porcelain empty and main = origin/main; mv Science/slope_hierarchy Science/z_projects_complete/slope_hierarchy.
2. Scaffold Science/slope_hierarchy_scale per autolens_assistant/skills/start-new-project.md Phase 1: README.md, AGENTS.md, CLAUDE.md, .claude/settings.json, .gemini/settings.json, project.yaml (assistant_ref at the current assistant commit, visibility_stage: private), CITATION.cff, wiki/project/{state,profile,README}.md, results/{figures,tables}/.gitkeep, paper/, .gitignore. Copy from the vault: scripts/{ep,graphical,one_by_one,util}.py, simulators/sample.py, config/, activate.sh, environment.yml, hpc/{sync,sync.conf.example,sync_jump.conf.example,template.py,batch_cpu/,batch_gpu/}. Write hpc/sync.conf (PROJECT_NAME=slope_hierarchy_scale, HPC_HOST=euclid_jump, HPC_BASE=/mnt/ral/jnightin). wiki/project/state.md cites z_projects_complete/slope_hierarchy/results/ep_history_n5_maxsteps12/ as the N=5 baseline. Import sanity only; git init + birth commit.
3. PyAutoCortex projects.yaml: new row slope_hierarchy_scale (remote: none, local_path, ral_root: /mnt/ral/jnightin/slope_hierarchy_scale, mirror: none, sync_cli: hpc/sync, sync_verbs: [push, pull, sync, status, submit, push-submit, jobs, tail, check], ledger: wiki/project/state.md, assistant: autolens_assistant, witness_file: results/**/*.json, partition: both, status: active, dated note:). python3 scripts/cortex.py retire slope_hierarchy --why "…", then hand-edit its local_path to the vault path.
4. git mv tasks/slope_hierarchy/n25_scale_up.md tasks/slope_hierarchy_scale/n25_scale_up.md; Project: slope_hierarchy_scale; Gates: PyAutoFit#1405, PyAutoFit#1558, PyAutoFit#1560, PyAutoFit#1562; replace the "Ready when…" line with the gate evidence (PyAutoMind/complete/2026/09/ep-scale-collapse-basin-cure-or-caveat.md, bundle #1572/#1573/#1574/#1576, #1580) and a "Borrowed from" note; python3 scripts/cortex.py move slope_hierarchy_scale/n25_scale_up gated. methods_writeup.md stays planned under the retired row.
5. python3 scripts/cortex.py check; pytest tests/ from the canonical Brain; pyauto-brain cortex dashboard --apply then --check exit 0.
6. Ship: commit, push, PR to PyAutoLabs/PyAutoCortex; end at PR-open.
7. Mind: draft/research/graphical_ep/ep_campaign.md phase-3/4 rows + header date; epics.md lines 35/40/57 pointers → "PyAutoCortex tasks carrying Epic: <slug>"; prompt_sync_push.

Key Files

• PyAutoCortex/projects.yaml — science body map (code; PR route)
• PyAutoCortex/tasks/slope_hierarchy/n25_scale_up.md — the phase-3 task, re-homed
• PyAutoCortex/scripts/cortex.py — retire, move, check (GATE_REF_RE :121, dir-name rule :637)
• autolens_assistant/skills/start-new-project.md — the birth scaffold
• PyAutoMind/draft/research/graphical_ep/ep_campaign.md, PyAutoMind/epics.md — ledger

Birth slope_hierarchy_scale — the EP campaign's first Cortex science project (phase 3)

Type: research
Target: graphical_ep
Repos:

• PyAutoCortex
• PyAutoMind
  Themes:
• graphical-ep
  Difficulty: small
  Autonomy: supervised
  Priority: high
  Status: formalised
  Consequence: judge
  Review-minutes: 10
  Unattended: fits
  Epic: graphical-ep
  Filed: 2026-09-08

Original request (verbatim)

Can we set up the first project as part of the EP epic which should be managed via a cortex
science project, and what is it? Continue the 'Expectation propagation (EP) campaign' epic. Its
canonical state lives in draft/research/graphical_ep/ep_campaign.md — read that ledger (and any
DECISIONS/RESULTS files beside it) first. Cross-check this epic's entry in PyAutoMind/epics.md, any
related rows in PyAutoMind/active.md, and the referenced repos' open issues and PRs, to work out the
last completed phase and what is currently in flight. Then pick the next logical step and continue
it through the normal workflow (/start_dev — filing the phase's prompt first if none exists),
updating the ledger as the work advances. Note: umbrella phase map — each phase's real content
lives in its own prompt under draft/research/graphical_ep/; the campaign file itself is never
issued. Science half: PyAutoCortex epics.md#graphical-ep (campaign phases 3 and 4).

Answer to the question

The first Cortex science project of the graphical-ep epic is slope_hierarchy, whose phase-1
task tasks/slope_hierarchy/n25_scale_up.md is campaign phase 3 (graphical non-EP JAX scaling,
"NUTS headline, EP cautionary"). It is chosen over phase 4 (ic50_workspace/ep_scale_up.md)
because the checkout is clean and in sync with its remote, the N=3/N=5 results the scale-up extends
are committed under results/ (so the row's witness_file glob resolves), the project issue
Jammy2211/slope_hierarchy#1 the witness comment needs already exists, the RAL root
/mnt/ral/jnightin/slope_hierarchy exists with output/ and results/ (verified 2026-09-08 over
euclid_jump), and it depends on neither the moment-matching cure nor the unissued
draft/feature/autofit/ep_lbfgs_jax.md companion. The ic50 checkout is dirty (170 files of
line-ending churn), has no results/ at its declared witness path, and sits outside the workspace.

Why the phase has not started

Survey 2026-09-08: phases 1 and 2 are SHIPPED (last movement 2026-09-08, PyAutoFit#1580); nothing
EP-shaped is in active.md, and no open PRs on PyAutoFit, autofit_workspace_test or PyAutoCortex.
Phases 3 and 4 migrated to the Cortex on 2026-09-01 and have sat at State: planned with an empty
Gates: ever since, although their "Ready when phase 2 is issued" condition has held since
2026-09-02. The ledger says "move to gated at the next /cortex check-in", but a check-in only
sweeps status: active rows (and projects owning a submitted|running task) — both EP rows are
status: dormant, so the instruction is unexecutable as written. projects.yaml is code (Cortex#22
ruling), so the activation is a PR to PyAutoLabs/PyAutoCortex with a human merge.

Ruling (human, 2026-09-08)

The phase runs as a fresh project slope_hierarchy_scale in /mnt/c/Users/Jammy/Science/,
not in the old slope_hierarchy tree: the old tree (17 commits, Jammy2211/slope_hierarchy,
wrapped up 2026-07-22 after answering its four N=5 goals) moves to Science/z_projects_complete/
and is mined for code. Assistant of record is autolens_assistant (the old project.yaml's
assistant_ref; the science is lens modelling).

Scope

1. Science tree (the Cortex local_path exception to the workspace-paths rule) — with the old
   checkout clean and in sync, mv Science/slope_hierarchy Science/z_projects_complete/slope_hierarchy
   (whole git repo, remote untouched). Birth Science/slope_hierarchy_scale per
   autolens_assistant/skills/start-new-project.md Phase 1 (thin, refer-back scaffold:
   README.md, AGENTS.md, CLAUDE.md, .claude/settings.json API gate, project.yaml with
   assistant_ref at the current assistant commit, CITATION.cff, wiki/project/, results/,
   paper/, .gitignore). Borrow from the vault: scripts/{ep,graphical,one_by_one,util}.py,
   simulators/sample.py, config/, activate.sh, environment.yml, hpc/ (sync CLI, conf
   examples, batch dirs, template). Write hpc/sync.conf (PROJECT_NAME=slope_hierarchy_scale,
   HPC_HOST=euclid_jump, HPC_BASE=/mnt/ral/jnightin). Do not copy output/, dataset/
   (re-simulate at N=25), old results/ or the old journal — wiki/project/state.md cites the
   vault's results/ep_history_n5_maxsteps12/ as the N=5 baseline. git init + one birth commit.
   Remote stays none (org repo creation is human-only; hand the human the gh repo create line).
2. PyAutoCortex projects.yaml — new row slope_hierarchy_scale (remote: none,
   local_path/ral_root as above, sync_cli: hpc/sync, verified sync_verbs, ledger: wiki/project/state.md, assistant: autolens_assistant, witness_file: results/**/*.json,
   partition: both, status: active, dated note:). Old row: cortex.py retire slope_hierarchy --why … (both its tasks are planned) plus local_path hand-edited to the vault path.
3. Task — git mv tasks/slope_hierarchy/n25_scale_up.md tasks/slope_hierarchy_scale/,
   Project: slope_hierarchy_scale, Gates: PyAutoFit#1405, PyAutoFit#1558, PyAutoFit#1560, PyAutoFit#1562, "Ready when" line replaced by the gate evidence
   (PyAutoMind/complete/2026/09/ep-scale-collapse-basin-cure-or-caveat.md, bundle
   #1572/#1573/#1574/#1576, #1580) and a "Borrowed from" note; cortex.py move … gated.
   ready is the human's edge. methods_writeup.md stays planned under the retired row.
4. Cortex hygiene in the same PR — cortex.py check, tests, pyauto-brain cortex dashboard --apply from the canonical Brain; PR to PyAutoLabs/PyAutoCortex (projects.yaml is code →
   human merge).
5. PyAutoMind ledger — ep_campaign.md phase-3 row (gated, birth, PR ref, human submits via
   /cortex), phase-4 row (companion path draft/feature/autofit/ep_lbfgs_jax.md; ic50 stays
   dormant until its checkout is cleaned and a results/ witness path exists); epics.md: replace
   the three stale "PyAutoCortex epics.md#" pointers (Cortex#14 replaced epics.md with
   checkin.yaml) with the task-file pointer.

Out of scope: submitting the N=25 run (human /cortex ask, ruled 2026-09-05), activating
ic50_workspace, the moment-matching cure decision, creating the GitHub remote.

Acceptance

• Science/slope_hierarchy_scale exists with one birth commit; the old tree sits under z_projects_complete/ intact.
• slope_hierarchy_scale renders on the Cortex board as an active project with n25_scale_up gated; slope_hierarchy reads retired.
• cortex.py check OK; Cortex CI (Cortex Check, Dashboard Refresh) green on the PR.
• The next /cortex check-in lists n25_scale_up under gates with resolvable refs.
• ep_campaign.md and epics.md no longer point at a non-existent PyAutoCortex/epics.md.