Ship log · iter #7

Iteration 7 ship log

2026-05-09 · continued autonomous /loop

On this pageWhat shipped What's now better for buyers What still needs work Next 5 recommended autonomous tasks Files changed inventory Cumulative iter1-iter7 numbers

Date: 2026-05-09 (continued autonomous /loop)

What shipped

New scripts on ubuntu@40.160.2.121

FilePurposeCron
robots-update.shIdempotent rewrite of /srv/sites/factory/robots.txt. Disallows /admin/, /api/, /repair-queue/, /unlock/, /feedback/, /dossiers/, /outreach-launcher/, /compare/, /calibrate/, /log/. Allows the public buyer-facing pages explicitly.manual / one-time
case-studies-gen.py166 honest case-studies template pages. Says "no live customers yet, here's what we expect, here's the operator partnership for guaranteed installation." Marker: WD_CASE_STUDIES_TEMPLATE_v1.8 hourly
regen-og-images.py172 SVG social-share cards at /factory/og/<slug>.svg. 1200x630. Uses per-product brand palette. Includes Adoptability score chip + $5 unlock chip.12,42
og-meta-injector.pyInjects <meta property="og:image"> and Twitter card meta into product pages. Idempotent marker WD_OG_META_v1.14,44
link-audit.pyWalks every HTML in /srv/sites/factory/, extracts internal links, checks for filesystem existence, writes JSON report + admin dashboard at /factory/admin/link-audit/.20 hourly
wes-picks.json (fixed)Replaced fake slugs (lead-recovery-ai, appointment-setter-ai, etc.) with real top-Adoptability slugs (bookkeeper-ai, dispatch-ai, nurture-ai, afterhours, discovery-call-ai, lead-scoring-ai). Catalog "Wes's Picks" section now renders.n/a

Live URLs

Hero polish progress

Sample of newer polished heroes:

These are all internal pre-existing issues, not regressions from my work. The Director's regen-agents-page.py cron writes those links; either it stops linking to non-existent agent pages, or those pages get built. Worklist.

What's now better for buyers

  1. Social shares look like real product cards. Drop a wishdeal.com/factory/builds/bookkeeper-ai/ link in Slack, LinkedIn, or Twitter and you get a 1200x630 card with the product name, tagline, Adoptability score, $5 unlock chip - branded with the product's own palette.
  2. Every product has a case-studies page. When a buyer clicks "Case studies" expecting evidence, they get the honest version: no live customers yet, here's the Fermi estimate of what good would look like, here's the operator partnership if they want it guaranteed. Builds trust through transparency.
  3. Catalog's Wes's Picks section actually works. Was empty for iter 1-6. Now shows top 6 by Adoptability, manually editable in wes-picks.json.
  4. Admin pages aren't crawlable. robots.txt explicitly disallows /admin/, /api/, /unlock/, /feedback/, /dossiers/, etc. (belt-and-suspenders with the existing noindex meta tags).
  5. Wes can see broken links at a glance. /factory/admin/link-audit/ is a 1-page dashboard showing what's broken, ranked by severity.

What still needs work

  1. Build per-agent pages OR remove broken links from /factory/agents/. The 13 broken references are the loudest item in the link audit.
  2. Hero polish next 10 (top 31-40).
  3. Stripe wiring when Wes provides keys.
  4. First-customer outreach push. Wes runs the operator-partnership email to 3-5 of his SC tenants, asks if they'd commission a Phase A. Real signal data.
  5. Per-product reading-time + word-count on catalog cards. Enables "this idea takes 4 minutes to read the dossier" - reduces dossier-purchase hesitation.

Files changed inventory

Cumulative iter1-iter7 numbers

← PreviousIter #6 Next →Iter #8