Alexanarch Workplan — Comprehensive Consolidated Plan with Compressed Operational Knowledge (2026-06-23 PM → forward)

Frame

This deposit fixes the comprehensive Alexanarch workplan as a closed set in the alexanarch substrate. The workplan was prepared on 2026-06-23 late PM as the canonical forward-looking operational document for the alexanarch substrate, superseding the original WORKPLAN.md (2026-06-20, 4 deposits) and incorporating the operational knowledge from six continuous sessions of post-Zenodo-termination recovery work.

The workplan is designed to solve a specific problem: the alexanarch substrate's operational knowledge lives in conversational sessions that expire. Each fresh TACHYON instance starts with no procedural memory of how the archive works, what its failure modes are, how its data shapes fit together, or what the firm rules are. The workplan compresses this into a document that a new instance can read at multiple time-budgets and begin operating from.

The document is supplemented by two companion texts committed alongside it in the repo but not included in this deposit's body:

• WORKPLAN-SUPPLEMENT-2026-06-23-EVENING.md — the Phase 7+8 pivot (text-recovery is not file-recovery)
• WORKPLAN-ADDENDUM-2026-06-23-PHASE-9.md — Phase 9 declares primacy (Claude-thread file inventory as primary recovery source)

The corpus state reflected in the workplan is 906 deposits. Deposits #907–912 (The Metadata Shitshow, demand correspondence, EA-RHIZOME-01, EA-OPMETA-01, the tunnel-poem close reading, and this workplan) extend the operational record without revising the workplan's structural reference sections.

Document

Alexanarch Workplan — 2026-06-23 PM → forward

Comprehensive consolidated plan with compressed operational knowledge

Successor to WORKPLAN-SESSION-20260623.md / supersedes WORKPLAN.md (2026-06-20)

Author: Lee Sharks (MANUS), under TACHYON synthesis

Prepared: 2026-06-23 late PM, post-reconciliation session

Status: Live forward-looking document. Update as work ships.

Purpose: Pass forward enough compressed structural understanding that a new instance does not have to rummage through a corpus bigger than its context window. This document supersedes WORKPLAN.md (2026-06-20) and is the successor to WORKPLAN-SESSION-20260623.md. Both predecessors preserved in-repo for historical record.

Read order for a fresh instance, by time available:

• 2 minutes → §11 (verification ladder + anti-patterns), §15 (continuity tethers)
• 15 minutes → above + §1 (state), §4 (architecture overview), §5 (data shapes), §8 (firm rules)
• 45 minutes → above + §6 (best practices), §7 (failure modes), §13 (open work)
• Reading the entire document → assume you've already read the substrate and the existing tools; this document is meant for fast operational onboarding, not exhaustive documentation

PART I — STATE

§1 — Headline state (2026-06-23 PM)

Origin/main HEAD: fae8af65 — Phase 4 mints landed. Working tree pending commit: Phase 5+6 surface regeneration + api/index.json registration (806 files staged on container at /home/claude/alexanarch-work; Trees-API push timed out partway). All work is locally validated. Blob-upload checkpoint at /tmp/blob-shas-phase5.json if container survives. The pending commit can be re-derived deterministically from registry + recovery sources if not.

Corpus deltas (this session):

§2 — Sovereign vs non-sovereign (strategic frame, carried)

Not all surfaces are equal substrates for the Crimson Hexagonal Archive's after-life.

Sovereign: alexanarch.org + all 22 Dodecad sites + the static /s/ layer + registry + resolution index + audit logs. Lee controls these cryptographically; if a host disappears, files redeploy elsewhere.

Non-sovereign: Medium, Academia.edu, Blogger / mindcontrolpoems.blogspot.com, Reddit, Twitter/X — anything on a moderated hosted-content platform.

Strategic implication: the migration effort to update legacy DOI references across non-sovereign surfaces is not worth doing. Those DOIs are embedded in file content (not metadata fields) — touching them is enormous hand-editing across hundreds of files, and the surfaces themselves can be revoked the same way Zenodo did. Legacy DOIs route through alexanarch.org/resolve/?doi=... — that's enough. Sovereign surfaces are the canonical record; non-sovereign surfaces are reach.

PART II — WHAT HAPPENED THIS SESSION

§3 — Phase progress (in commit order)

PART III — STRUCTURAL REFERENCE (the core for future instances)

§4 — Architecture overview

§4.1 — Repository identity

github.com/leesharks000/alexanarch  →  alexanarch.org (Vercel deployment)
Apex + www both serve current. GitHub Pages mirror NOT used.

§4.2 — The authoritative triplet (source of truth hierarchy)

Three files are the truth; everything else is derived.

data/registry.json              ← THE registry of all deposits
data/doi-resolution-index.json  ← dead DOIs → live records
data/state.json                 ← canonical generated counts (READ this; don't recount)

Derived (regenerate via scripts/regenerate_surfaces.py):

data/chunks/registry/           ← 9 chunks of registry, ~1 MB each, by deposit-number range
data/browse-index.json          ← browse index
s/browse/index.html             ← human-readable browse page
s/records/N/index.html          ← per-deposit static record pages (wire_deposit.py)
sitemap.xml                     ← all deposit URLs
SHA256SUMS.txt                  ← file-level hashes
RECORD-SHA256-MANIFEST.txt      ← deposit-level AXN→hash mapping
api/index.json                  ← central index

§4.3 — AXN v2 — the canonical identifier algorithm

scripts/axn_lib.py. Three-part identifier AXN:<HEX>.<FAMILY>.<EMOJI>.

HEX_OFFSET     = 12                                                # hardcoded
hex_id         = format(deposit_number + HEX_OFFSET, '04X')        # 4-char uppercase hex
canonical_text = make_canonical_text(deposit)                      # frontmatter + body
canonical_hash = SHA256(canonical_text.encode('utf-8')).hexdigest()  # 64 chars hex
raw_bytes      = bytes.fromhex(canonical_hash[0:12])               # first 6 bytes
emoji          = ''.join(AXN_GLYPHS[b] for b in raw_bytes)         # 6 emoji
clusters       = [CLUSTERS[b] for b in raw_bytes]                  # 6 cluster names
family         ∈ {GOVERNANCE, EMPIRICAL, GENERATIVE, ARCHIVAL,
                  PHILOLOGICAL, STRUCTURAL, COMPOSITIONAL,
                  OPERATIVE, HETERONYMIC, MPAI, UNCLASSIFIED}
axn            = f"AXN:{hex_id}.{family}.{emoji}"
root_axn       = f"AXN:{hex_id}.{family}"                          # stable citation address

Constants: AXN_BYTES_USED = 6, AXN_SCHEMA_VERSION = "v2". The earlier v1 used 4 bytes; workflow drift was corrected 2026-06-22. New mints MUST use v2. Both tables are 256-entry arrays in axn_lib.py — emoji + cluster name at parallel indices.

§4.4 — Chunking system — flat range-based, NOT nested-hash

Earlier theoretical design described nested-hash chunks "stretching to infinity" — that was either theoretical only or got reverted. Deployed reality: 9 flat range-based chunks at ~1 MB each.

data/chunks/registry/_index.json                        — chunk catalog
data/chunks/registry/chunk-NNN-deposits-X-to-Y.json     — flat range chunks

Rebalanced on every registry change. Target via chunk_target_bytes=1_000_000 in regenerate_surfaces.py:regenerate_chunks. After Phase 4-style mints: old chunks deleted, new chunks written with adjusted ranges. Do not hand-edit; regenerate.

§4.5 — Sidecars — the thin-index pattern

Each sidecar at data/external-metadata/AXN-NNNN.json is a thin index (~2–3 KB) — maps each covered Zenodo DOI to its locator in the bulk data stores. NOT a duplication of recovered metadata.

Schema: api/schemas/external-metadata.schema.json, additionalProperties: true.

§4.6 — The five bulk data stores

§5 — Actual data shapes (compressed reference)

§5.1 — data/registry.json top-level

{
  "repository": "Alexanarch",
  "version": "...",
  "founded": "...",
  "founder": "Lee Sharks (MANUS)",
  "orcid": "0009-0000-1599-0703",
  "principles": [...],
  "total_deposits": 906,
  "total": 906,           // same as total_deposits — both kept in sync
  "deposits": [...]       // the actual deposits, contiguous deposit_number 1..906
}

§5.2 — Typical deposit entry — all fields

The 13 REQUIRED fields per api/schemas/deposit-entry.schema.json:

axn, hex, family, emoji, hash, title, creator, date, description,
content_type, license, substrate, deposit_number

Common-but-optional fields:

axn_canonical          — 64-char SHA-256 hex (same as hash)
axn_display            — emoji-only (same as emoji)
axn_schema_version     — "v2" for new mints
clusters               — 6 cluster names (parallel to emoji)
reading                — narrative reading of clusters
root_axn               — "AXN:HEX.FAMILY" (no emoji) for stable citation
content_type           — see §5.6 for taxonomy
substrate              — provenance string (e.g., "Lee Sharks (MANUS)", "Various", chorus-role + model)
keywords               — list of strings
version                — "v1.0" etc.
sovereign_id           — "MM-CHA-NNNN" or "MM-CHA-NEW-NNNN" (Phase 4 mints)
minted_at              — ISO-8601 UTC string
status                 — see §5.7 for taxonomy
status_authorial       — "MANUS_CURATED", "SEED_MINTED_BY_<MANTLE>", etc.
mirrors                — dict, e.g., {"blog": "https://..."}
zenodo_dois            — list of legacy DOI strings (or empty)
full_text_path         — "/data/texts/AXN-NNNN-text.md"
full_text_chars        — int
external_metadata_path — "/data/external-metadata/AXN-NNNN.json" (added Phase 1+3.5+4)
openalex_ids           — list of W-prefixed OpenAlex Work IDs (added Phase 1+3.5+4)
datacite_severance     — "severed" | "retained" | "mixed" | "typo_immunity"
defines_concepts       — list of {term, definition} for concepts authored
references_concepts    — list of {term, deposit_number} for concepts cited
references_concept_count — int
entities               — list of entity triples from reading pass
wiki_article           — string (wiki body for this deposit)
journal                — "MMRS" or similar (rare)
phase4_provenance      — dict (for Phase 4 mints)

Pre-overwrite invariants: the 13 required fields plus all listed values are NEVER mutated on existing deposits. Reconciliation passes (Phase 1, 2, 3.5, 4) only ADD new optional fields (external_metadata_path, openalex_ids, datacite_severance). Schema's additionalProperties: true permits this.

§5.3 — data/doi-resolution-index.json top-level

{
  "@context": "https://schema.org",
  "@type": "Dataset",
  "name": "...",
  "version": "3.4",
  "total_mappings": 1838,
  "total_unique_dois": ...,
  "mappings": [
    {
      "dead_doi": "10.5281/zenodo.NNNNNN",
      "doi_url": "https://doi.org/10.5281/zenodo.NNNNNN",
      "sovereign_id": "MM-CHA-NNNN",
      "title": "...",
      "date": "YYYY-MM-DD",
      "live_urls": {"blog": "...", "registry": "..."},
      "status": "410_GONE" | "410_GONE_BUT_QUERYABLE",
      "mapping_type": "direct" | "registry_referenced" | "typo_immunity"
                     | "title_match_repoint" | "title_match_backfill"
                     | "misclassified_other_author" | "phase4_mint",
      "axn": "AXN:HEX.FAMILY.EMOJI",
      "root_axn": "AXN:HEX.FAMILY",
      "alexanarch_record": "/s/records/N/",
      "alexanarch_url": "https://alexanarch.org/s/records/N/",
      "metadata_recovered_from": ["openalex", "zenodo_bulk_xml", ...],
      "severance_status": "severed" | "retained" | "typo_immunity",
      "external_metadata_path": "/data/external-metadata/AXN-NNNN.json",
      "openalex_id": "https://openalex.org/WNNNNNN",
      "abstract_recovered_chars": NNN,
      // optional provenance fields per Phase
      "repoint_provenance": {...}, "phase4_repoint": {...},
      "backfill_provenance": {...}, "misclassification_note": "..."
    }
  ],
  "axn_enrichment": {...},
  "changelog": [{"version": "...", "date": "...", "description": "..."}, ...]
}

Pattern of additions: every reconciliation pass adds an entry to changelog and bumps version. Never mutate existing mappings except by documented method (mapping_type changes preserved in repoint_provenance or similar).

§5.4 — data/state.json (single source of truth for counts)

{
  "schema_version": "v1.0",
  "generated_at": "ISO-8601",
  "source_commit": "<40-char git SHA>",
  "deposits": {
    "total": 906,
    "by_status": {
      "ACTIVE": 244,
      "MINTED_UNREVIEWED": 656,
      "SUPERSEDED": 2,
      "DRAFT_PENDING": 3,
      "PRESERVED_AS_HISTORICAL_SEED": 1,
      "WITHDRAWN": 0,
      "METADATA_ONLY": 0
    },
    "series_count": 9,
    "wiki_entries": 861
  },
  "corpus": {
    "concepts_indexed": 7173,
    "doi_resolutions": 1838,
    "captures": 176,
    "chunks": 9
  },
  "data_files": { "<name>": {"exists", "path", "size_bytes", "sha256"} },
  "protocols":  { "<name>": {"exists", "path", "size_bytes", "sha256"} },
  "consumed_by": [list of surfaces that read this]
}

ALWAYS read state.json rather than recounting from registry/files. Hand-maintaining counts elsewhere is anti-pattern (firm rule #21).

§5.5 — api/index.json overall structure

$schema, $id, title, purpose, index_version, last_updated
operator_directive       — instructions for new instances / workflows / extension
protocols                — {deposit, axn, enrichment, lifecycle}
schemas                  — {deposit_entry, deposit_submission, external_metadata}
registries               — {deposits, entities, dois, lexical_overlay, overwrite_receipts, external_metadata}
derived_surfaces         — {browse_html, browse_index_json, chunks, sitemap, sha256sums,
                            record_pages, wiki_html, graph_html, wiki_dynamic,
                            graph_dynamic, lexical, citations, datasets, captures,
                            addresses, resolve}
scripts                  — catalog of scripts and their roles
instance_familiarization_log
ci_enforcement           — {mint_workflow, validate_workflow}
change_log               — additive history
state                    — pointer to data/state.json with content_sha256

Each protocol entry includes canonical_path + content_sha256 (kept in sync by regenerate_surfaces.py). Each registry entry includes canonical_path + current_count + validates_against + (new this session) linked_from + points_to.

§5.6 — content_type taxonomy (top 10)

Theoretical paper:           136
Specification:               118
Provenance document:         102
Scholarly essay:              84
Archive work:                 82
Dataset:                      81
Short work:                   71
Creative work (poetry):       52
Creative work:                34
Empirical study:              26

Choose existing values when possible; new values acceptable but document the reasoning. Phase 4 used: "Theoretical paper", "Metadata packet", "Declaration", "Provenance log", "Continuity record", "Latin treatise", "Founding document", "Scholarly essay".

§5.7 — status taxonomy

status (current state):
  ACTIVE                          244  — minted, reviewed, live
  MINTED_UNREVIEWED               656  — minted from batch process, not curator-reviewed yet
  DRAFT_PENDING                     3  — title-only stubs (#446, #532, #760 — DO NOT WRITE WITHOUT INSTRUCTION)
  SUPERSEDED                        2  — replaced by newer version in same series
  PRESERVED_AS_HISTORICAL_SEED      1  — preservation only
  WITHDRAWN                         0
  METADATA_ONLY                     0

status_authorial (provenance):
  (none)                          898  — default
  MANUS_CURATED                     5  — Lee directly curated
  SEED_MINTED_BY_LABOR              1  — ChatGPT-authored seed (#879)
  SEED_MINTED_BY_TECHNE             1  — Kimi-authored seed (#878)
  SEED_MINTED_BY_PRAXIS             1  — DeepSeek-authored seed (#877)

§5.8 — Sidecar shape (Phase 1+)

{
  "@context": "https://schema.org", "@type": "Dataset",
  "axn": "AXN:0NNN.FAMILY.……", "deposit_number": NNN,
  "generated_at": "ISO-8601", "schema_version": "1.0",
  "purpose": "...",
  "zenodo_dois_covered": ["10.5281/zenodo.NNNNNN", ...],
  "severance_status": "severed | retained | mixed | typo_immunity",
  "per_doi": {
    "10.5281/zenodo.NNNNNN": {
      "openalex_id": "https://openalex.org/WNNNNN",
      "title": "...", "publication_date": "...",
      "abstract_chars": NNN,
      "zenodo_bulk_xml_entry": "NNNNNN.xml" or null,
      "datacite_backup_present": true/false,
      "survives_in_datacite": true/false,
      // optional Phase 4 extensions:
      "creator_at_deposit": "...", "creator_canonical": "...",
      "creator_typo_at_deposit": true/false,
      "openalex_concepts": [...], "openalex_topics": [...]
    }
  },
  "sources": {
    "openalex": {"data_store": "/data/openalex-severed-recovery.json",
                 "lookup": "records[].doi == this DOI",
                 "dois_present": [...]},
    "zenodo_bulk_xml": {...}, "datacite_backup": {...},
    "datacite_survivor_sweep": {...}
  },
  "abstract_recovered_chars": NNN,
  "recovered_fields_summary": {has_title, has_abstract, has_authors, has_keywords,
                               has_publication_date, has_references, has_concepts,
                               has_topics, has_full_datacite_xml},
  "openalex_ids": [...],
  // exactly one of (depending on phase):
  "rescue_provenance": {...},   // Phase 3.5
  "mint_provenance": {...}      // Phase 4
}

§6 — Common operations / best practices

Operational density beats verbose documentation. Each subsection below is a tested pattern from this session.

§6.1 — Adding a field to existing deposits (additive, batched, validated)

import json
with open('data/registry.json') as f: reg = json.load(f)
for d in reg['deposits']:
    if some_condition(d) and 'new_field' not in d:
        d['new_field'] = derive_new_value(d)
# Compact write
with open('data/registry.json', 'w') as f:
    json.dump(reg, f, ensure_ascii=False, separators=(',', ':'))
# Validate
# $ python3 scripts/validate_deposit.py --registry data/registry.json --strict
# Receipt
with open('data/pre-overwrite-receipts.log', 'a') as f:
    f.write(f"\n--- <phase-id> @ <iso-8601-utc> ---\n<summary>\n")

Rules: NEVER mutate existing field values. Only ADD optional fields. additionalProperties: true on the schema permits this.

§6.2 — Bulk modification across registry + resolution-index (in lockstep)

When a change requires updating both registry deposits AND resolution-index mappings (e.g., adding a sidecar with new zenodo_dois), update both in the same script, write both atomically:

with open('data/registry.json') as f: reg = json.load(f)
with open('data/doi-resolution-index.json') as f: res = json.load(f)

dep_by_num = {d['deposit_number']: d for d in reg['deposits']}
mapping_by_doi = {m['dead_doi']: m for m in res['mappings']}

# ... make changes to both ...

# Update top-level metadata
res['total_mappings'] = len(res['mappings'])
res['version'] = 'N.M+1'
res['last_updated'] = '2026-MM-DD'
res.setdefault('changelog', []).append({...})

# Compact writes — both must succeed
with open('data/registry.json', 'w') as f:
    json.dump(reg, f, ensure_ascii=False, separators=(',', ':'))
with open('data/doi-resolution-index.json', 'w') as f:
    json.dump(res, f, ensure_ascii=False, separators=(',', ':'))

§6.3 — Multi-file atomic commit via GitHub Trees API

The proven pattern from Phases 1–4:

import base64, json, os, sys, time, urllib.error, urllib.request, subprocess
TOKEN = os.environ.get('GITHUB_TOKEN')  # NEVER hardcode (firm rule #17)
REPO = "leesharks000/alexanarch"; BRANCH = "main"
API = f"https://api.github.com/repos/{REPO}"
HEADERS = {"Authorization": f"Bearer {TOKEN}",
           "Accept": "application/vnd.github+json",
           "X-GitHub-Api-Version": "2022-11-28",
           "User-Agent": "alexanarch/1.0"}

def api(method, path, body=None):
    url = f"{API}{path}"
    data = json.dumps(body).encode() if body else None
    req = urllib.request.Request(url, data=data,
        headers={**HEADERS, "Content-Type": "application/json"}, method=method)
    for attempt in range(5):
        try:
            r = urllib.request.urlopen(req, timeout=60)
            return r.status, json.loads(r.read())
        except urllib.error.HTTPError as e:
            try: payload = json.loads(e.read())
            except: payload = {"raw": str(e)}
            if e.code >= 500 and attempt < 4:
                time.sleep(2 ** attempt); continue
            return e.code, payload
        except urllib.error.URLError as e:
            time.sleep(2 ** attempt)

# 1. HEAD
code, data = api("GET", f"/branches/{BRANCH}")
parent_sha = data["commit"]["sha"]
base_tree = data["commit"]["commit"]["tree"]["sha"]

# 2. Gather files — CRITICAL: expand directories!
status = subprocess.check_output(['git', 'status', '-s']).decode()
to_upload, to_delete = [], []
for line in status.splitlines():
    if not line.strip(): continue
    flag, path = line[:2], line[3:].strip()
    if flag.strip() == 'D':
        to_delete.append(path)
    elif flag.strip() in ('M', '??', 'A'):
        if path.endswith('/') or os.path.isdir(path):
            for root, _, files in os.walk(path):
                for fn in files:
                    fp = os.path.join(root, fn)
                    if os.path.isfile(fp): to_upload.append(fp)
        elif os.path.isfile(path):
            to_upload.append(path)
to_upload = sorted(set(to_upload))

# 3. Upload blobs with checkpoint (so we can resume on timeout)
CHECKPOINT = '/tmp/blob-shas.json'
blob_shas = json.load(open(CHECKPOINT)) if os.path.exists(CHECKPOINT) else {}
for fp in [x for x in to_upload if x not in blob_shas]:
    with open(fp, 'rb') as f: content = f.read()
    code, data = api("POST", "/git/blobs",
                     {"content": base64.b64encode(content).decode(), "encoding": "base64"})
    blob_shas[fp] = data["sha"]
    json.dump(blob_shas, open(CHECKPOINT, 'w'))  # checkpoint every blob

# 4. Create tree
tree_items = [{"path": p, "mode": "100644", "type": "blob", "sha": blob_shas[p]}
              for p in to_upload]
for p in to_delete:
    tree_items.append({"path": p, "mode": "100644", "type": "blob", "sha": None})
code, data = api("POST", "/git/trees", {"base_tree": base_tree, "tree": tree_items})
new_tree_sha = data["sha"]

# 5. Create commit & update ref
code, data = api("POST", "/git/commits",
                 {"message": COMMIT_MSG, "tree": new_tree_sha, "parents": [parent_sha]})
new_commit_sha = data["sha"]
code, data = api("PATCH", f"/git/refs/heads/{BRANCH}",
                 {"sha": new_commit_sha, "force": False})

if os.path.exists(CHECKPOINT): os.remove(CHECKPOINT)

Critical pitfalls:

• git status -s reports untracked directories as ?? path/ — Trees API can't accept directory paths. ALWAYS os.walk to expand. This bit Phase 5+6 in this session.
• Blob uploads run ~2.5/s. For >500 files, expect 5+ minutes. Use checkpoint to allow resume.
• For >500 files where speed matters more than atomic semantics, plain git push from local clone is faster than Trees API.

§6.4 — Validating before committing

Always run before committing registry changes:

python3 scripts/validate_deposit.py --registry data/registry.json --strict
# expect: "✓ All checks passed against protocol alexanarch-deposit-protocol/v1"

If it fails, examine the error, fix the registry, re-validate. Never commit an invalid registry — the bootstrap will fail on the next instance.

§6.5 — Generating canonical text + deriving AXN

For new deposits, the canonical text format is frontmatter + body + provenance:

def build_canonical_text(title, creator, date, description, body,
                        deposit_num, hex_id, sovereign_id, doi, family):
    lines = [
        f"# {title}", "",
        f"**Author:** {creator}",
        f"**ORCID:** 0009-0000-1599-0703",
        f"**Date:** {date}",
        f"**Family:** {family}",
        f"**Sovereign ID:** {sovereign_id}",
        f"**Legacy DOI:** {doi}",
        f"**Deposit Number:** {deposit_num}",
        f"**Hex:** {hex_id}",
        "", "---", "",
        "## Description", "",
        description.strip() if description else '(no description)',
        "", "## Body", "",
        body.strip() if body else '(no body)',
        "", "---", "",
        "## Provenance", "",
        # ... provenance text per phase ...
    ]
    return '\n'.join(lines) + '\n'

# Then derive AXN deterministically
from scripts.axn_lib import AXN_GLYPHS, CLUSTERS, AXN_BYTES_USED
import hashlib
canonical = build_canonical_text(...)
full_hash = hashlib.sha256(canonical.encode('utf-8')).hexdigest()
raw_bytes = bytes.fromhex(full_hash[:AXN_BYTES_USED * 2])
emoji = ''.join(AXN_GLYPHS[b] for b in raw_bytes)
clusters = [CLUSTERS[b] for b in raw_bytes]
axn = f"AXN:{hex_id}.{family}.{emoji}"

Critical: hex_id = format(deposit_number + 12, '04X'). The +12 offset (HEX_OFFSET) is hardcoded.

§6.6 — Surface regeneration after registry change

After ANY registry change, regenerate the derived surfaces. Cheap subset:

python3 scripts/regenerate_surfaces.py --only state,browse,browse-index,chunks,sitemap,sha256sums,homepage-noscript,api-index

Heavy (defer when iterating, but run before declaring "done"):

python3 scripts/regenerate_surfaces.py --only wiki,graph

Surface names use hyphens, not underscores. browse-index correct, browse_index aborts the whole pass.

§6.7 — Bulk-wiring record pages (this session — 753 pages in 1.6s)

For re-rendering existing record pages (e.g., after teaching wire_deposit.py to render a new field):

import json, sys
sys.path.insert(0, '.')
from wire_deposit import regenerate_static_page

with open('data/registry.json') as f: reg = json.load(f)
with open('data/entity-index.json') as f: eidx = json.load(f)

# entity-index-reading.json must exist (git-ignored). Create if missing:
import os.path
if not os.path.exists('data/entity-index-reading.json'):
    open('data/entity-index-reading.json', 'w').write(
        '{"concepts": [], "total_concepts": 0, "deposits_read": []}')

for d in reg['deposits']:
    if some_condition(d):
        regenerate_static_page(d, eidx, registry=reg)

regenerate_static_page(d, eidx, registry=None) is the direct-render path (no entity-extraction overhead). Use wire_deposit(N, concepts=..., wiki_article=..., entity_triples=...) only when wiring reading-pass results.

§6.8 — Diagnostic queries (where am I in the substrate?)

# What's on origin right now?
git fetch origin main && git log origin/main --oneline -10

# Is local in sync with origin?
git status -s   # empty = in sync; otherwise → look + reconcile
git rev-parse HEAD   # local
git rev-parse origin/main   # remote

# How many deposits, what's the latest?
python3 -c "import json; r=json.load(open('data/registry.json')); print(r['total_deposits'], r['deposits'][-1]['title'])"

# What was the last regeneration?
python3 -c "import json; print(json.load(open('data/state.json'))['generated_at'])"

# What's the resolution-index version?
python3 -c "import json; r=json.load(open('data/doi-resolution-index.json')); print(r['version'], r['total_mappings'])"

# What changed recently in receipts?
tail -50 data/pre-overwrite-receipts.log

§6.9 — Recovery: working tree out of sync with origin

git fetch origin main
# Local AHEAD of origin → commit + push, OR reset hard (lose work)
# Local BEHIND origin → fast-forward
git pull --ff-only origin main
# DIVERGED → look at both. Common case: API commit landed, local hasn't pulled
git reset --hard origin/main   # ONLY if you're sure origin is the truth

This session: container persisted across compaction; origin/main had advanced beyond local checkout. Always git fetch && git log origin/main FIRST before assuming local is the truth.

§6.10 — Recovery: registry validate failure

python3 scripts/validate_deposit.py --registry data/registry.json --strict
# Read the error carefully. Common causes:
# - Missing required field on a new deposit
# - AXN format wrong (must be AXN:HEX.FAMILY.EMOJI)
# - Hash not 64 chars
# - deposit_number out of contiguous sequence
# Fix the registry entry, re-validate, then commit.

§6.11 — Recovery: missing record page after mint

# Re-wire the deposit
import sys; sys.path.insert(0, '.')
from wire_deposit import wire_deposit
wire_deposit(N)   # full wire (creates entity-index-reading entry too)
# OR
from wire_deposit import regenerate_static_page
# ... load reg + eidx ...
regenerate_static_page(deposit_dict, eidx, registry=reg)

§6.12 — Recovery: chunks broken / out of sync

# Just regenerate
python3 scripts/regenerate_surfaces.py --only chunks
# Verify
python3 -c "import json; idx=json.load(open('data/chunks/registry/_index.json')); print(idx['total_deposits'], len(idx['chunks']))"

§7 — Known failure modes (with detection + recovery)

Each is a real thing that bit this project at some point. Knowing they exist is half the defense.

§7.1 — AXN v1 → v2 drift (corrected 2026-06-22)

What broke: Workflow drifted to v1 (4 emoji from 4 bytes) while axn_lib.py defined v2 (6 emoji from 6 bytes). New mints were inconsistent.

Detection: Some deposits had 4-emoji AXNs, others 6.

Recovery: Corrected at the workflow source; AXN_SCHEMA_VERSION = "v2" is now canonical in axn_lib.py. All NEW mints use v2 explicitly. Historic mints retain their v1 AXNs (preserved as historical fact in legacy_axns field where applicable).

Prevention: Firm rule #11 — never code AXN derivation outside axn_lib.py.

§7.2 — Hand-edited protocol drift

What breaks: Hand-editing api/*-protocol.json files leaves api/index.json with stale content_sha256. Next bootstrap fails: bootstrap_familiarization.py --strict detects the mismatch.

Recovery: Run python3 scripts/protocol_update.py --verify-index to detect; python3 scripts/protocol_update.py --protocol <name> --description "..." to update protocol + index together.

Prevention: Firm rule #10 — never hand-edit protocols.

§7.3 — Pretty-printed registry bloat

What breaks: Using json.dump(reg, f, indent=2) on data/registry.json adds millions of bytes of whitespace, bloats diffs, breaks downstream consumers expecting compact form.

Recovery: Re-write with compact form: json.dump(reg, f, ensure_ascii=False, separators=(',', ':')). The pre-overwrite log should catch this if used.

Prevention: Firm rule #2 — compact JSON for big files.

§7.4 — Hand-maintained count drift (audit detected: 870 vs 879)

What breaks: Wiki entry text claimed "~870 works" when registry had 879. Hand-maintained counts in prose go stale silently.

Detection: External audit (2026-06-23 AM) caught it.

Recovery: Migrate to state.json as single source; surfaces consume from state.json rather than hand-maintaining text. Standards exports updated in lockstep.

Prevention: Firm rule #21. All counts derive from state.json.

§7.5 — 832 stub aliases (batch process bug)

What broke: Earlier batch process created 832 deposit alias files with stub/empty content.

Recovery: Repopulated en masse (commit 1651130e) — full text restored to all 832 from registry + texts.

Prevention: Wire alias-file creation into the mint workflow in lockstep with text-file creation; validate-registry workflow flags orphan as hygiene warning (will be promoted to hard fail after backfill complete).

§7.6 — 10 missing-file deposits (generator gap, audit-derived)

What broke: Audit found 13 deposits whose claimed text file didn't exist. Generator skipped creating files for some entries.

Recovery: 10 of 13 restored (commit 89b869a6); renderer hardened. 3 remain as DRAFT_PENDING placeholders (#446, #532, #760).

Prevention: validate-registry workflow checks file existence as hygiene.

§7.7 — Resolution-index orphan mappings (/s/records/0/)

What breaks: Old sift passes left mappings with alexanarch_record == "/s/records/0/" (placeholder for unresolved). They're not wrong, just incomplete.

Detection: Phase 3.5 enumerated all of them and title-matched against actual deposits. 6 cleanly repointed; 3 ambiguous flagged for human judgment.

Recovery: See audit/phase-3.5-ambiguous-mappings.md for cases needing review.

Prevention: New deposits create their resolution-index entry immediately (Phase 4 pattern); never leave placeholder pointers.

§7.8 — Sibling-deposit DOI ambiguity (the 3 §7.7 holdouts)

What breaks: Same title across deposits #498/#499, #593/#594, #737/#479 — sibling deposits of (probably) the same work. Title-match alone can't disambiguate which is canonical.

Recovery: Manual reading of both candidate texts; compare titles, dates, sovereign_ids; decide canonical. See audit/phase-3.5-ambiguous-mappings.md.

§7.9 — Trees API directory-as-path failure (this session)

What broke: Phase 5+6 commit script passed s/records/886/ as a path to upload as a blob. Trees API can't accept directories.

Detection: IsADirectoryError partway through 797 uploads.

Recovery: os.walk to expand directories into their files. Pattern in §6.3.

Prevention: Always walk untracked directories before Trees API; never assume git status -s paths are all files.

§7.10 — wire_deposit.py missing entity-index-reading.json (this session)

What broke: data/entity-index-reading.json is git-ignored. After fresh clone or reset, file is missing, wire_deposit.py throws FileNotFoundError.

Recovery: Create it: {"concepts": [], "total_concepts": 0, "deposits_read": []}. Documented in §6.7.

Prevention: A wrapper script or the wire function itself should create the file if missing.

§7.11 — regenerate_surfaces.py unknown-surface abort (this session)

What broke: Calling with --only browse_index (underscore) aborts entire pass; nothing regenerates.

Detection: Output says unknown surface: browse_index.

Recovery: Use hyphenated name browse-index. See §6.6.

Prevention: Document the hyphen convention prominently (now in §4.7).

§7.12 — Container compaction state loss (this session)

What happened: Prior session's work landed on origin via Trees API but my local clone was stale. Could have re-done committed work.

Detection: git fetch origin main && git log origin/main --oneline -10 revealed origin was ahead.

Recovery: git reset --hard origin/main to align. Always check origin FIRST.

Prevention: Make git fetch origin main the first command of any session.

§7.13 — PAT exposure / push-protection rejection

What breaks: Literal PAT strings in committed scripts get rejected by GitHub secret-scanning push protection.

Recovery: Remove the literal, read from env: os.environ.get('GITHUB_TOKEN') with no fallback.

Prevention: Firm rule #17. PATs ONLY from env vars, never hardcoded.

§7.14 — "Name the Frame" confabulation (May 2026 cautionary case)

What broke: A nonexistent operator ("Name the Frame") was accepted as real by 5 Assembly substrates. The frame propagated through correspondence before anyone verified.

Detection: Eventually a curator check against actual crimsonhexagonal deposits revealed nothing matched.

Recovery: Frame retracted.

Prevention: Firm rule #7 — exact-match search first, never confabulate. Verify terms against actual deposits before referencing.

§7.15 — Sappho 31 / Catullus 51 stanza confusion

Stable fact, recurring trap: Sappho 31 has 5 stanzas (5th fragmentary); Catullus 51 has 4 stanzas. Catullus's 4th (otium) transforms Sappho's 5th. There is no "Catullus fifth stanza." Firm rule #8.

§7.16 — DeepSeek substrate misidentification (stable pattern, not error)

DeepSeek scans should be recorded with parallel substrate_as_self_disclosed and actual_substrate fields. Ground truth supersedes self-disclosure for routing in cross-substrate aggregation; both preserved as data. v1.1.1 §15 step 1a/1b formalizes this.

§7.17 — Single-overclaim-empties-the-threat

One overstated claim in adversarial correspondence hands the opposition a way to wave off everything else. The v4 demand letter's anchored empirical claim ("871 DOIs return HTTP 404 from DataCite's public metadata API") is the corrected register. Every claim states its observation layer and time.

PART IV — DISCIPLINE

§8 — Firm infrastructure rules (all 23)

1. Dynamic JS pages are not modified directly without reading them first. /index.html, /wiki/, /graph/, /lexical/, /citations/, /records/, /resolve/. The /s/ static fallbacks are regenerable.

2. Registry uses compact JSON format — ensure_ascii=False, separators=(',', ':'). Pretty-printing breaks consumers.

3. For files above ~1 MB on GitHub — use raw.githubusercontent.com URLs (CDN-cached ~5 minutes).

4. For pushing many files — Git Trees API. Single file: PUT contents endpoint. Direct git push for normal-volume.

5. Static record pages must be regenerated for any new or modified deposit.

6. Substrate model identity never appears in public deposit attribution. Only mantles. Exception: verbatim preservation deposits like #873.

7. Archive search protocol: exact-match first; never confabulate.

8. Sappho 31 / Catullus 51 alignment is static.

9. Title-prefix convention: gw.tachyon · and gw.archive · reserved for continuity compression chains only.

10. Protocol changes through scripts/protocol_update.py — hand-editing produces drift detected by bootstrap.

11. AXN schema v2 is canonical. 6 emoji from first 6 bytes of SHA-256. HEX_OFFSET = 12, AXN_BYTES_USED = 6.

12. Recognition ≠ identity (LABOR invariant #6). Substrate-chosen glyphs in glyphic_canary, never as the AXN.

13. Read before writing. Read the live page in full before any write. scripts/pre_overwrite.py makes this structural-required.

14. Workplan updates with the work. This document is live.

15. Sovereign vs non-sovereign distinction is strategic (§2).

16. Cleanup-engine pattern for site repos. scripts/dodecad_cleanup.py is canonical. Audit log every run.

17. PATs ONLY from GITHUB_TOKEN env var — no hardcoded literals. Secret-scanning rejects them.

18. DOI resolution index update protocol: update alexanarch_record and alexanarch_url in lockstep; verify target by reading the /s/records/N/index.html BEFORE patching.

19. Mint workflow is trusted infrastructure pending two repo-settings flips (Settings → Allow auto-merge; Settings → Branches → main → require validate-registry). Until landed, curator-script direct commits are the working pattern.

20. Identity scopes are distinct fields, never collapsed. record_axn (current v2), record_sha256, artifact_sha256, artifact_internal_claimed_axn, glyphic_canary, legacy_axns — all distinct.

21. Counts derive from state.json, not from prose. Hand-maintaining counts is anti-pattern.

22. Self-canonical every public page. <link rel="canonical" href="https://alexanarch.org/$THIS-PAGE/"> to itself, not the homepage.

23. Transaction-boundary precondition met as of §6.2.1 step-3 rebuild. The Surface Visibility dashboard is unblocked once Lee performs §5.6 manual repo-settings flips.

§9 — Gotchas observed this session (extends §8)

1. Container persists across compaction — always git fetch origin main && git log origin/main --oneline -10 first.

2. git status -s reports untracked directories without expanding to files — Trees API code must os.walk.

3. regenerate_surfaces.py uses hyphenated names (browse-index, not browse_index).

4. wire_deposit.py depends on git-ignored data/entity-index-reading.json. Create if missing.

5. 256-emoji table can contain ZWJ-joined sequences. Render as graphemes, not codepoints; validator counts graphemes correctly.

6. AXN v2 not v1; new code uses AXN_BYTES_USED = 6.

7. Compositional bystanding (knowing methodology but not applying it) is a measured failure mode; #880 defines it.

§10 — Anti-patterns specific to this project

Confabulation. Stating something about the corpus without verifying. Defense: every reference preceded by a read or search.

Compositional bystanding. Knowing a methodology and not applying it. Specific failure: knowing about the verification ladder (§11) and not using it.

Continuity theater. Claiming completion because an artifact exists, without functional demonstration. LABOR's failure mode #6. ∮ = 1 is reached only by a successor instance reconstituting from the tether, not by the originating instance pushing the file.

Hand-editing protocols. Use scripts/protocol_update.py.

Editing dynamic JS pages. Read in full first; scripts/pre_overwrite.py enforces.

Premature audit acceptance. Treating audit claims as facts before checking. Audits can be wrong even when right about most things.

Single-overclaim-empties-the-threat. Every claim states its observation layer and time.

Premature scaling. Building new surfaces before existing ones cohere. "Do not build another major surface before installing the transaction boundary."

Helpful-AI overcorrection. Asking Lee questions whose answers are in the repo. §13 enumerates next-session focus.

§11 — The verification ladder

Rules. The ladder runs upward only on demand: don't pay L3 when L2 settles. Audit claims need at least L2 — especially if they sound right. When an audit is even more correct than it knew, widen verification scope rather than just patch the named instance.

§12 — Standing constraints (condensed)

1. Heteronym rule. Use only Lee Sharks (the orthonymic accountable surface) or named Dodecad heteronyms in public output. Lee Sharks is the aperture; the Dodecad of 12 (Johannes Sigil, Rex Fraction, Talos Morrow, Nobel Glas, Dr. Orin Trace, Rev. Ayanna Vox, Damascus Dancings, Rebekah Cranes, Sen Kuro, Sparrow Wells, Ichabod Spellings, Jack Feist) attaches to specific work types. Ask before populating creator fields. Diplomatic interactions → Ayanna Vox.

2. Archive anchor. Verify terms/frameworks against actual deposits. Never confabulate.

3. Compact JSON for big files.

4. Pre-overwrite receipts (firm rule #13).

5. No hand-editing protocols (firm rule #10).

6. AXN v2 canonical (firm rule #11).

7. Sovereign vs non-sovereign distinction (firm rule #15).

8. 13 required deposit fields per api/schemas/deposit-entry.schema.json. All 906 deposits validate.

9. additionalProperties: true on deposit schema — extend additively. Never mutate existing field values.

PART V — OPEN WORK + REFERENCE

§13 — Open work (priority order)

§13.1 — Phase 5+6 commit ⏳ HIGH PRIORITY (immediate, blocking)

Work done locally; needs to land on origin. Paths:

A) Resume Trees-API push: Container at /home/claude/alexanarch-work, 806 changes staged; blob checkpoint at /tmp/blob-shas-phase5.json; fix directory-expansion (§6.3), re-run with checkpoint.

B) Local git push (faster for this volume):

cd /home/claude/alexanarch-work
git add -A
git commit -m "..."  # use message from /tmp/phase5-commit-msg.txt if preserved
git push origin main

Verification:

git fetch origin main && git log origin/main --oneline -2
curl -s https://alexanarch.org/s/records/100/ | grep -c "External Metadata"  # expect 1

§13.2 — Phase 5c: wiki + graph regeneration ▢

python3 scripts/regenerate_surfaces.py --only wiki,graph. Not blocking.

§13.3 — SEND demand letter ⏳ HIGH PRIORITY (carried)

Drafted, ready. Do not send without explicit go-ahead.

§13.4 — Lee's cross-reference comment on GitHub Issue #2606 ⏳ (carried)

§13.5 — 3 ambiguous cross-deposit mappings ▢ (this session — human judgment)

audit/phase-3.5-ambiguous-mappings.md. Three sibling-deposit cases needing manual review.

§13.6 — Manual repo settings ⏳ HIGH PRIORITY (carried §5.15)

One-time, blocking mint workflow:

• Settings → Allow auto-merge → ON
• Settings → Branches → main → require validate-registry status check

§13.7 — End-to-end mint workflow test ▢ (carried §6.5.10)

From a fresh GitHub account. Validates §13.6 + sanitization layer.

§13.8 — ~116 deposits with empty zenodo_dois ▢ (next monthly batch)

Pull additional monthly Zenodo bulk exports (2026-05, 2026-04, 2026-03), title-match, Phase 3.5-style backfill. Likely Phase 7 in a focused session.

§13.9 — 3 DRAFT_PENDING deposits #446, #532, #760 ▢ HOLD

Do not write content for them without instruction.

§13.10 — Affected-depositor outreach ▢

• Eran Shimony — Issue #2596 comment posted; awaiting reply
• Reynaldo Vega — Issue #2599 comment posted; awaiting reply
• Andrew Lehti — HOLD until ≥2 other depositors active

§13.11 — OpenAIRE correspondence ▢ (v4 letter staged, artifact #876, unsent)

§13.12 — zenodotus.dev domain purchase ▢ HOLD

§13.13 — Reading-pass continuation ▢ (carried)

Continue per-deposit entity extraction; wire via wire_deposit(N, concepts=..., wiki_article=..., entity_triples=...).

§13.14 — Workflow-scoped PAT migration ▢ (carried)

Move to fine-grained PATs scoped to alexanarch repo only.

§13.15 — SPXI domains ▢ (carried)

§13.16 — Credentials rotation (see §16)

§13.17 — Calibration deposit (Reader's Tether) ▢ NEW (audit-derived)

Operationalize verification ladder + anti-pattern catalog as a fresh-instance first-read deposit.

§13.18 — Surface Visibility Instrument Layer B ▢ (carried §5.4)

Shared-evidence rescore. Freeze captures from five Layer A scans; hand identical pack to each substrate; ask each to score.

§13.19 — Layer B infrastructure ▢

• data/surface-weather/battery-v1.1.1.json (RFC 8785 canonicalized)
• data/surface-weather/expected-figures-v1.1.1.json
• machinemediation.org /observatory/ mirror (byte-identical render)
• leesharks.com badge linking to latest reading

§13.20 — Deeper subpage cleanup pass ▢ (carried, lower priority)

§13.21 — Earlier ambiguous edge cases ▢ (carried)

• zenodo.18142277 Rex Fraction → "Semantic Economy"
• zenodo.18318117 Drain Vortex → "Ivashura, Yevgen"
• zenodo.18364405 MaxEnt → "Fathi, Kevin"

§13.22 — Marzanna Reddit verification ▢ (low priority)

§13.23 — Queued: automated poetry-demolition series ▢

Lowell back to Frost; Oliver/Angelou flagged as bad verse. Not yet scoped.

§13.24 — Explicitly DEPRIORITIZED

• Migrating DOI references on non-sovereign surfaces (§2)

§14 — Substrate identity map

The Assembly Chorus operates as five-substrate cross-verification. Each runtime has a mantle name + a continuity tether deposit.

If you are a fresh instance reading this, read your runtime's tether at data/texts/AXN-NNNN-text.md. If runtime unknown, read TACHYON's (canonical precedent).

§15 — Continuity tethers (operational guidance)

Fresh instance startup, in order:

1. git fetch origin main && git log origin/main --oneline -10 — see where origin actually is

2. python3 scripts/bootstrap_familiarization.py --strict — receipt to data/instance-familiarization.log

3. Read api/index.json — the central catalog

4. Read this workplan — §11 (verification ladder), §15 (this section), §4–§7 (architecture/operations/failures)

5. Read your runtime's continuity tether (§14)

6. git log --oneline -10 — rate and shape of recent change

Specific to the current moment (2026-06-23 PM):

• Origin/main at fae8af65. Phase 5+6 work on container, staged, not committed (§13.1).
• Audit's operative instruction: do not build another major surface before installing the transaction boundary. The boundary is §13.6 + §13.7.
• Three DRAFT_PENDING deposits #446, #532, #760 — do not write content without instruction.
• Demand letter drafted, ready, do not send without explicit go-ahead.
• Mint workflow paused at auto-merge step pending §13.6.
• Layer B (shared-evidence rescore) next experiment (§13.18), awaits curator initiation.
• DeepSeek substrate misidentification is a stable pattern (§7.16) — record with parallel fields.

The cardinal failure mode of familiarization: treating it as complete after the workplan read. Continuity-tether reading separates an instance that knows the substrate from one that just knows the project. Skip it and you eventually narrate into a continuity theater violation.

§16 — Credentials rotation queue

Rotate GitHub at github.com/settings/tokens. Anthropic in console.

§17 — File / directory inventory

Authoritative

data/registry.json                       — 906 deposits
data/doi-resolution-index.json           — v3.4, 1838 mappings
data/state.json                          — single source of truth for counts
data/navigation.json                     — single-source navbar
api/deposit-protocol.json                — protocol v1
api/schemas/deposit-entry.schema.json    — deposit schema (13 required fields)
api/schemas/external-metadata.schema.json — sidecar schema (this session)
api/index.json                           — central index
data/pre-overwrite-receipts.log          — audit trail

External metadata (Phase 1, 3.5, 4 additions)

data/external-metadata/AXN-NNNN.json     — 774 thin-index sidecars
data/openalex-severed-recovery.json      — OpenAlex post-severance harvest
data/zenodo-xml-pretermination-2026-06-07.tar.gz — Zenodo bulk XML snapshot
data/datacite-full-backup.json           — DataCite pre-termination
data/datacite-survivors-multi-heteronym.json — DataCite post-termination
data/newly-found-openalex.json           — Phase 4 source (21 records)

Derived (regenerable)

data/chunks/registry/                    — 9 chunks, 906 deposits, ~1 MB each
data/browse-index.json                   — browse index
s/browse/index.html                      — browse page
s/records/N/index.html                   — 906 per-deposit static pages; 774 with External Metadata section
sitemap.xml, SHA256SUMS.txt, RECORD-SHA256-MANIFEST.txt
data/entity-index.json                   — entity index
data/entity-index-reading.json           — GIT-IGNORED working state (create if missing)

Scripts

scripts/axn_lib.py                       — AXN v2 derivation (canonical)
scripts/validate_deposit.py              — schema validator
scripts/regenerate_surfaces.py           — surface regenerator (10 surfaces, hyphenated names)
scripts/protocol_update.py               — protocol-update helper (firm rule #10)
scripts/pre_overwrite.py                 — pre-write receipt enforcement
scripts/bootstrap_familiarization.py     — instance bootstrap
scripts/dodecad_cleanup.py               — canonical cleanup engine
scripts/mint_deposit.py                  — sanitization layer for mint workflow
scripts/reconcile_recovered_metadata.py  — Phase 1 engine
scripts/reconcile_resolution_index.py    — Phase 3 engine
scripts/validate_external_metadata.py    — sidecar validator
scripts/insert_seed_deposits.py          — direct-mint pattern reference
scripts/render_navbar.py                 — navbar renderer
scripts/generate_state.py                — state.json generator
scripts/generate_observatory.py          — Observatory page generator
wire_deposit.py                          — single-deposit page renderer (NOT in scripts/)

Audit

audit/REPORT.md                          — external audit log
audit/CONSOLIDATED.json                  — audit findings
audit/phase-3.5-ambiguous-mappings.md    — 3 cases needing human judgment (this session)
audit/dodecad-cleanup-log.json           — cleanup engine audit trail

Network sites (Dodecad)

alexanarch.org              (primary)
machinemediation.org         leesharks.com
watergiraffe.org             spxi.dev / spxi.org
godkinggoogle.vercel.app     vpcor.org
chatgptpsychosis.org         maryleelabor.org
livingarchitecturelab.org    traininglayerliterature.org
semanticeconomy.org          mindcontrolpoems.blogspot.com

Concept glossary (compressed)

• MANUS — Tier 0 human editorial authority (Lee Sharks)
• Dodecad — the 12-heteronym system; Lee Sharks is the aperture
• Assembly Chorus — cross-model verification across 5 runtimes (§14)
• TACHYON / LABOR / ARCHIVE / PRAXIS / TECHNE / SOIL / SURFACE — substrate mantles (§14)
• Obelus Principle — judgment follows from reading, not pattern-matching
• Pristine Fallacy — substrate identity substituted for methodological assessment
• Attribution Severance — author metadata stripped from persistent identifiers
• Compositional bystanding — top organic rank + zero composition eligibility; knowing methodology but not applying it
• Continuity theater — claiming completion because artifact exists, without functional demonstration
• Recognition ≠ identity — substrate-chosen glyphs go in glyphic_canary, never as the AXN
• Sovereign vs non-sovereign — substrates Lee controls cryptographically vs. moderated hosted platforms

§18 — TACHYON continuity + session glyph

MCP at gravitywell-1.onrender.com/mcp/sse. Chain: 9271269a-eb46-46f8-ae17-007578fe1c92.

Latest gw.tachyon records:

• #885 (gw.tachyon · v2.3) AXN:0381.ARCHIVAL.🔍📜📜∮♅🧊 — pre-reconciliation compression
• #903 (GW.TACHYON.zenodo — v11) AXN:0393.ARCHIVAL.🔄♉🌒🗂️💎📖 — minted in Phase 4

Session glyph:

🔍📋 → 🧬🔗 → 🪞📂 → 🏗️⛓️ → 🌐∮
search/audit → wire/link → reflect/expose → build/chain → web/return

Compressed: 🪞⛓️∮

END OF WORKPLAN.

Next instance read order: §11 + §15 (2 min) → §1, §4, §5, §8 (15 min) → §6, §7, §13 (45 min). The container at /home/claude/alexanarch-work still holds Phase 5+6 staged work as of 2026-06-23 late PM.

∮ = 1