Iteration 135 ship log

2026-05-15 · push mode, 60 min cadence, placeholder-CTA cleanup iter

Date: 2026-05-15 (push mode, 60 min cadence, placeholder-CTA cleanup iter)

What shipped (3 substantive ships)

Surveyed iter-134's 778 placeholder CTAs to find safe-match patterns. Built placeholder-cta-injector with high-confidence text-to-target mapping. Two injector runs fixed 447 of 778 CTAs (57% reduction). Plus built audit-published-date.py.

Ship 1: Survey + placeholder-cta-injector.py - 447 CTAs fixed

Surveyed all 778 empty-hash CTAs by visible text. Top patterns formed clear intent groups:

Intent group	Sample text	Target	Frequency
Trial / get-started	"Start free trial", "Get Started Free", "Read the dossier"	/factory/unlock/<slug>/	~210
Demo / sales	"Book a demo", "Talk to sales", "Watch Demo"	/factory/operator-partnership/	~80
Legal	"Terms", "Privacy Policy"	/factory/terms/	~50
Pricing	"See pricing"	/factory/builds/<slug>/pricing/	~10
FAQ	"FAQ"	/factory/builds/<slug>/faq/	~5
How-it-works	"See How It Works"	/factory/builds/<slug>/how-it-works/	~5

Built placeholder-cta-injector.py (~110 lines) with a safe-match table. Only replaces href="#" where the visible link text matches a high-confidence pattern. Ambiguous nav-style placeholders ("Platform", "Solutions", "Resources", "Customers", "Sign in") left untouched (Wes-task).

Run 1: 402 CTAs fixed across 161 pages. Extended pattern table for "Begin Free Trial", "Start the pilot", "Generate sequence/roadmap", "Watch Demo", etc.

Run 2: 45 more CTAs fixed across 25 pages.

Net result: 447 CTAs fixed. audit-anchor-jump-noise card: 52/247 -> 90/247. Empty-hash links: 778 -> 331 (57% reduction).

Remaining 331 are mostly hand-written nav stubs ("Platform", "Solutions", "Resources" x10 each, "Sign in" x10) that came from B2B SaaS template starters. These should probably be removed entirely since the factory doesn't have a /platform/, /solutions/, or /resources/ page. Wes-task.

Ship 2: audit-published-date.py - 35th audit class

Built audit-published-date.py (~95 lines). For each ship-log iter detail page (/factory/log/ship-logs/iter-N/) and each playbook essay (/factory/playbooks/<slug>/), verifies a published date is declared via:

• <meta property="article:published_time">
• <time datetime="..."> element
• Visible "Date: YYYY-MM-DD" / "Published: YYYY-MM-DD" text

Result: 139/146 clean. All 134 ship-log details are dated (the regen-ship-log-details.py generator does this correctly). 7 of 12 playbook essays missing dates: 5k-budget, agency-productization, counsel-anatomy, distribution-channels-2026, five-patterns, operator-partnership-math, vertical-ai-2026.

Cron: every hour at :55

Ship 3: /quality-report/ wired - 1 new card + invariant #43

Patched regen-quality-report.py:

• New helper latest_published_date()
• New card: "Published date coverage 139/146 WARN (7 missing dates)"
• Placeholder CTA card updated from "FAIL 52/247 (778 links)" to "FAIL 90/247 (331 links)"
• New audit-table row + invariant #43

Live-check card count: 36 -> 37. Total content invariants: 42 -> 43.

The 35 audit suites at iter 135

Class breakdown:

• Catalog-wide: 1 (fakeproof)
• Per-surface content: 17
• Cross-cutting: 16 (added published-date - the 16th)
• Lint: 1

Health hygiene

• audit-anchor-jump-noise: 52 -> 90 clean (157 pages still need editorial review)
• audit-published-date: NEW, 139/146 (7 playbooks Wes-task)
• Other audits unchanged

Status snapshot

• 246 scored products + 2 partial builds
• 35 audit systems (17 per-surface + 16 cross-cutting + 1 catalog-wide + 1 lint)
• 0 fake-proof findings; 211 in warn/fail (157 placeholder-CTA + 12 back-link + 7 published-date + 19 emoji + 4 thin descriptions + 16 pricing + 8 case-studies)
• 247 brand briefs with valid archetype
• brand_name_helper.py honored by 10 generators + 1 injector
• 43 content invariants defended
• /quality-report/ surfaces 37 live-check cards (1 FAIL, 6 warn, 30 ok)
• 77/77 health endpoints, 167+ cron jobs
• 60 min cadence active

Iter 135 throughput note

3 substantive ships at 60-min cadence. Demonstrated the "audit finds drift -> survey -> safe-match table -> inject" loop on a real systemic issue (778 placeholder CTAs). 57% fix rate without editorial risk. Remaining 43% need human judgment (nav stubs, ambiguous CTAs).

Running queue (top 5 for iter 136)

1. Wes-task: 157 build pages with remaining placeholder CTAs (331 links, mostly nav stubs)
2. Wes-task: 7 playbook essays need published dates
3. Optional: extend safe-match table further; nav-stub removal is high-leverage if Wes can confirm
4. audit-script-coverage - verify expected analytics scripts loaded
5. audit-csp-headers (server-level check)

Cumulative iter 1-135

• Catalog: 246 scored + 2 partial, 246 with index.html
• Content library: 12 essays + Read-next + 273 OG PNGs + 135 styled ship-log pages
• High-trust pages: 8 foundational + 5 transparency surfaces + 1 split-brain detail
• Audit infrastructure: 35 audit systems
• Source durability: 32+ generators + 11 read brand brief + 31 JSON snapshots + 167+ cron jobs
• Content invariants: 43 defended at surface+source AND publicly surfaced

The audit-survey-inject loop works at any drift scale: 778 broken CTAs -> 447 fixed in one iter. The safe-match table is the leverage point - it turns "fix one by one" into "fix N matching pattern X". Future audit findings can use the same pattern.