Built For The Machines.

Six GEO/SEO commits in seven days. Five pull requests merged. Forty-four daily briefs backfilled with structured data. Zero new features shipped — every hour this week went into making the existing 80-plus-page archive legible to the systems that now decide what gets surfaced and what gets ignored.

That's a deliberate trade. The Cathedral Framework says nothing ships that isn't verifiably live — but "live" is no longer just a URL that resolves for a human. It's a URL that a crawler can parse, a page that declares its own dates and geography and relationships without ambiguity, a description that doesn't repeat itself fifty-four times across the archive. This week's log is a maintenance week by any conventional definition. It's also the most consequential audit this site has run since the sitemap was first checked in.

Six commits, chronologically. Not proposals — merged pull requests, checked-in diffs, resolving URLs. The record is the proof.

1. Hreflang + FAQPage Schema on /builds/ (PR #47)

hreflang="en" and hreflang="x-default" added across all eight main section pages. A six-question FAQPage schema landed on /builds/ for rich-result eligibility, alongside CollectionPage dateModified and inLanguage. Homepage and /about/ schema freshness dates refreshed to match. Lighthouse on the deploy preview: SEO 100, Accessibility 98, Best Practices 100, PWA 100.

2. Six Missing Daily Briefs — Continuity Restored

The daily-content automation had gone quiet for six days, June 24 through 29. Six new daily briefs went in following the established seven-topic rotation, daily/index.html got newest-first cards, sitemap.xml gained six new URLs with dated lastmod entries, and the homepage's WebPage.dateModified was corrected to match. Lighthouse on the signal01 preview: Performance 90, Accessibility 98, Best Practices 100, SEO 100, PWA 100 — no regressions.

3. ProfessionalService Schema for Local/GEO Discoverability (PR #49)

The homepage carried only a generic Organization type — no LocalBusiness-family markup, despite being a Canby, Oregon consulting practice serving Oregon and PNW clients. A ProfessionalService block landed with geo-coordinates and areaServed, sitting alongside the existing Organization and WebSite schema rather than replacing it.

4. GEO Tags + BreadcrumbList Backfilled on 44 Daily Briefs (PR #50)

The daily-content-agent template started emitting geo.region/geo.placename/geo.position/ICBM meta tags and BreadcrumbList structured data on June 20 — but the 44 briefs generated before that date, May 6 through June 19, never got backfilled. This PR brought the full daily archive to parity for local/GEO discoverability and breadcrumb rich results. Forty-four files, one pattern, zero exceptions.

5. The One Page That Was Missing Parity (PR #51)

Every other section page — news, videos, builds, profile, about, archive — carried hreflang alternates and a keywords meta tag. /reel-engine/ was the one page site-wide missing both, plus dateModified and inLanguage on its TechArticle schema. Brought to parity, sitemap lastmod refreshed. A one-page gap is still a gap.

6. The Duplicate-Meta-Description Root Cause (PR #52)

The biggest find of the week: 54 of 55 daily briefs were sharing only seven unique descriptions, rotated verbatim across six-to-eight dated pages apiece — a textbook duplicate-meta-description problem that actively suppresses unique search-result snippets. The root cause traced back to scripts/daily-content-agent.mjs: its deterministic per-topic fallback template (the path that runs whenever no OPENAI_API_KEY is configured) had no date-based differentiator baked in.

The fix went two levels deep. First, the symptom: every daily brief's description — meta, og:description, twitter:description, and JSON-LD — got rewritten to end with a per-date clause instead of the fixed boilerplate tail, making all 54 descriptions unique while trimming them closer to Google's ~155–160 character display budget. dateModified and sitemap lastmod bumped for every touched page, and daily/index.html's card excerpts were regenerated to match. Second, the cause: fallbackBrief() now runs through a new humanDate() helper so every future auto-generated brief stays unique even when the AI path is unavailable. Patching the output without patching the generator would have meant doing this again in six weeks.

Search used to mean ten blue links and a human deciding which one to click. Increasingly it means an AI crawler reading the page once, extracting a claim, and deciding whether to cite it at all — no click, no impression, no second chance to make the structure legible.

That's the actual shift behind this week's work. GPTBot, ClaudeBot, PerplexityBot, and the rest don't scroll. They parse. A page with a compelling headline and a buried publish date reads, to a crawler, as a page with no publish date. A page whose canonical geography lives only in prose ("we're based in Canby, Oregon") reads as a page with no geography at all. Schema.org markup, hreflang declarations, BreadcrumbList structure, and accurate dateModified stamps aren't decoration on top of the content — for a machine reader, they are the content that determines whether this page gets treated as current, authoritative, and citable.

This is the quieter half of the Cathedral Framework's proof-first discipline: it's not enough for a claim to be true and a URL to resolve. The claim has to be structured so a system with no capacity for judgment calls can still verify it mechanically. Six commits this week didn't add a single new sentence of marketing copy. They made 82 existing pages tell the truth about themselves in a format a crawler can trust without asking a human to confirm it.

The tell that this mattered: PR #52 wasn't a schema tweak, it was a root-cause fix in the content generator itself. Fixing 54 duplicate descriptions by hand would have solved this week's problem. Fixing the fallback template so it can't produce duplicates again solved the recurring one. That's the difference between decorating for the machines and actually building for them.

• DAILY AGENT Monitor for recurrence of the automation gap that produced the June 24–29 backfill; confirm the scheduled workflow is firing daily without silent drops.
• SCHEMA AUDIT Extend the FAQPage/BreadcrumbList pattern from /builds/ to the remaining section pages that don't yet carry it.
• FIELDNOTES Sweep the fieldnotes archive for the same duplicate-description class of issue found in daily/ this week.
• SITEMAP Confirm lastmod freshness holds across all four files touched this week (index.html, archive/index.html, feed.xml, sitemap.xml) after this dispatch publishes.
• OPENAI_API_KEY Evaluate wiring the daily-content-agent's AI path back in so the fallback template becomes the exception, not the default.