Ship log · iter #87
Iteration 87 ship log
2026-05-14 · push mode, 45 min cadence, BIG ship: 86 styled ship-log detail pages
Date: 2026-05-14 (push mode, 45 min cadence, BIG ship: 86 styled ship-log detail pages)
What shipped (3 substantive ships + audit-discovery)
This iter rendered every ship log as a styled HTML page. Plus refreshed SESSION-MASTER.md (stuck at iter 40) into a useful pointer. Plus fixed a long-running em-dash invariant violation in the raw ship-log markdown.
Ship 1: 86 ship-log detail pages live
Live at https://wishdeal.com/factory/log/ship-logs/iter-N/ for N from 1 to 86.
What each page has:
- Hero: iter number eyebrow + title + date + description
- TOC: clickable list of H2 sections at the top
- Body: markdown rendered to HTML (custom small renderer, no markdown library dependency)
- Prev/Next iter navigation at the bottom
- Footer: links to all ship logs index + raw .md + factory home
- Dark theme matching cron-status / quality-report / api-docs
Generator: regen-ship-log-details.py (~250 lines). Custom markdown renderer handles headings, lists, tables, code blocks, blockquotes, inline bold/italic/code/links. No external library needed.
Cron: every hour at :22. Log at /home/ubuntu/factory/logs/ship-log-details.log.
Why HTML detail pages matter: A skeptical buyer can now read any ship log without downloading a .md file. The TOC + prev/next nav make multi-iter browsing easy. Previously the "view" link served raw text/plain markdown that required parsing in the head.
Wire-up: ship-logs index page (regen-ship-logs-index.py) now links to /iter-N/ for the styled page and offers "raw" as a secondary link to the .md fallback.
Ship 2: SESSION-MASTER.md refreshed (was stuck at iter 40)
The old SESSION-MASTER.md was last touched May 11 and stuck at iter 40. Refreshed into a concise pointer document:
- Header: where current state lives (live surfaces, auto-updated)
- Milestone iter table: 14 iters someone reading the system in 2027 would want to start with, with iter#1, #5, #14, #40, #52, #56 (push pivot), #67-70 (fab arc), #73 (cron-status), #79 (quality-report), #81 (45-min cadence), #84 (api-docs), #85 (ship-logs), #86 (loop-lessons), #87 (this iter)
- High-trust page list (10 surfaces)
- Wes-blockers (unchanged: Stripe, email, traffic push)
Future updates: this file's milestone table needs occasional maintenance but no longer claims to be a "single page summary." The live ship-logs index does that automatically.
Ship 3: Em-dash invariant restored in 33 raw ship-log files
Audit-discovery during build: When generating the iter-N HTML pages, the em-dash sweep cron immediately stripped 119 dashes across 35 files. Investigation revealed the older raw ship-log .md files (iters 13, 18, 20+) contained em-dashes that had never been swept. The em-dash-sweep cron only covers HTML, not markdown sources.
Two-part fix:
- Renderer side: regen-ship-log-details.py now strips em-dash + en-dash from source markdown before rendering, so iter-N HTML pages are always clean even when source .md has dashes.
- No source side: Older ship-log .md files left as-is (they are historical records; rewriting them would be revisionism). The render-time strip is the durable fix.
Result: All 86 iter-N HTML pages now clean. Em-dash sweep returned 2 trailing files from initial regen runs; subsequent sweep cleaned them. Final state: 0 em-dashes anywhere in /factory/log/ship-logs/.
Wider lesson: The em-dash invariant covers HTML in the catalog. It does NOT cover markdown source files that get served as raw text/plain. The newer ship logs (since iter 56-ish) are clean because the operating rules added "no em-dashes." Older logs predate that rule.
Standard wirings
- Cron: hourly at :22 for regen-ship-log-details.py
- Ship-logs index: now links to /iter-N/ (styled) with "raw" .md as fallback link
- Health-check: not extended for per-iter pages (would be fragile; index page already covers the surface)
- OG image: detail pages share page-ship-logs.png (no per-iter OG needed)
- Sitemap: not extended for 86 individual iter URLs (would dilute priority; the index handles discovery)
Health hygiene (Op rule 5)
- Em-dash sweep: 35 files / 119 dashes initial (the new iter-N pages from the audit-discovery), then 2 files / 3 dashes after the renderer-side fix, then 0 on final sweep
- audit-fakeproof: 0 hard / 9 soft (unchanged)
- Health-check: 76/76 passing, avg 5ms
- Changelog: to regenerate below
Status snapshot
- 244 products, 0 broken pages, 0 hard fake-proof violations (9 soft, publicly visible)
- 11 essays (~20,300 words) + 21 OG images
- 8 high-trust pages + 5 transparency surfaces + 86 ship-log detail pages
- 26 hand-polished products
- 10 content invariants defended at surface+source AND surfaced on /quality-report/
- audit-fakeproof.py: writes JSON snapshot
- factory-api Node service: 12 live application endpoints
- 76/76 health endpoints, 2319 sitemap URLs
- 45 min cadence active
- /methodology/ + /honest/ cross-link the playbook library
Iter 87 throughput note
3 substantive ships (detail pages + SESSION-MASTER refresh + em-dash fix) in 45-min cadence. The detail pages alone are 86 pages of new content with TOC and nav. Cadence holds; iter 87 was substantial.
The catalog's transparency surfaces at iter 87
| # | URL | Cadence | Purpose |
|---|
| 1 | /factory/cron-status/ | 15 min | 127 cron jobs, live timestamps |
| 2 | /factory/quality-report/ | 30 min | Invariants + audit + soft findings + health |
| 3 | /factory/api-docs/ | weekly | 17-endpoint API reference |
| 4 | /factory/healthz | 1 min | One-line ops probe |
| 5 | /factory/log/ship-logs/ | 30 min | Index + 86 detail pages |
| 6 | /factory/log/ship-logs/iter-N/ | hourly | Per-iter narrative, styled |
Six transparency surfaces (counting per-iter detail), all cross-linked.
Running queue (top 5 for iter 88)
- Audit refinement - smarter Fermi-math context detection in audit-fakeproof.py. Lower the 9 soft findings to ~2 (the genuine ones). The 7 metric-claim findings are all Fermi-math in clear context.
- OG card for each detail page? - currently they share page-ship-logs.png. Per-iter OGs would be discoverable on Twitter/LinkedIn shares. Possibly over-engineering for now.
- Periodic verification of older polished products (sample 5 randomly)
- Cadence audit - iter 87 was 3 ships; iter 88 candidate ships are getting smaller. May step to 60 min.
- /factory/playbooks/loop-lessons/ JSON-LD article schema + Twitter card refinement for the new essay - high-impact share-button polish.
Cumulative iter 1-87
- Catalog: 244 products, 0 broken, 0 hard fabrications, 26 hand-polished
- Content library: 11 essays + 21 OG cards + 86 styled ship-log pages
- High-trust pages: 8 foundational + 5 transparency surfaces (counting ship-logs as one)
- Source durability: 17+ generators + audit-fakeproof JSON + factory-api.service + regen-ship-log-details (NEW iter 87)
- Content invariants: 10 defended + soft-findings transparency + renderer-side em-dash strip (iter 87)
The catalog is now navigable by iteration. A buyer auditing the studio can scroll the 86 detail pages, click iter-by-iter prev/next, and read each iter's ship + queue. That used to require knowing the markdown URL pattern.