Updates Log

The Famous Cats encyclopedia reached the ROADMAP C.2 target of 600 active rows. The previous 2026-05-10 entry left the table at 378 active rows after the schema extension and the first long-bio pass. Today's work added 222 new entries written under the same anti-template guardrails (Section 8 of the project roadmap), with three cross-class de-duplications caught along the way.

What shipped

• 222 net new famous-cat rows. Each carries an 80–120 word short bio (extended_bio), a 300–450 word long bio (extended_bio_long) for ≥98% of entries, and a wikipedia_url pointer. New rows are spread across nine source-type classes: movie, real-pet, tv, book, comic, internet, historical, game, plus a small number of reclassifications.
• Three classes reached round milestones. The tv class now holds 101 cats, real-pet 105, and book 99. Comic, movie, internet, historical, and game each grew by 8–20 rows depending on candidate availability.
• Four cross-class duplicates removed. "Unsinkable Sam" (re-added as a sample then found to duplicate the existing "The Unsinkable Sam"), "Catbug" and "Cake" (both added to comic but canonically web-series originated, so the comic rows were dropped in favour of the tv rows), "Yoruichi Shihouin" (canonical manga origin, so the anime row was dropped), and "Mr. Mistoffelees" (canonical T. S. Eliot 1939 poem, so the 2019 Cats film row was dropped). Mistakes captured in the project memory so future parallel-agent batches scan the whole table rather than only their own source_type.
• Eight NULL source_type rows resolved. Six were reclassified to comic / tv / real-pet with verifiable owner or work attribution (Charley to Marmaduke, Cheddar to Stephen Harper, Kitty to Scooby-Doo, Lion to John Fahey, Nothing to Jean-Paul Sartre, Heed as a Guinness short-cat candidate). Two were soft-deleted under the honesty principle: their original bios ("the famous calico cat", "a famous cat") carried zero verifiable specifics.
• Wildcat marker on ten non-domestic-cat rows. The new notable_traits token wildcat tags the Lion King characters, Aladdin's Rajah, Kung Fu Panda's Tigress and Tai Lung, Madagascar's Vitaly, and Ice Age's Diego and Soto. The encyclopedia is primarily domestic-cat oriented, and these characters are kept because the names are widely used by cat owners — but readers and filters can now distinguish them.

Process notes

• Anti-template guardrails enforced per row. Each new bio passes word-count band (80–120 short, 300–450 long), em-dash density ≤ 1.5%, zero banned phrases ("perfect for", "delightful", "rich tapestry", etc.), opening framing axis rotated across consecutive rows, and ≥3–5 verifiable specifics. These are the same guardrails used in the original 2026-05-10 long-bio pass.
• Honest skips. Seven older rows (Peter, Tim, Tristram, Velcro, Yseult, Abigail kitten, Kitty) lacked any short bio to expand from, so the long-bio pass left their extended_bio_long empty. The Section 8 honesty principle is to admit gaps rather than invent material.
• Parallel-agent workflow. Today's writing was distributed across batches of 4–8 parallel research agents per source-type class, with whole-table fuzzy duplicate checks at INSERT time. The cross-class duplicates above were caught after deployment by a separate audit pass and removed via soft-delete.

Why this matters: the famous-cats encyclopedia is a major linking surface across the site — every cat-detail page and many category and tip pages link into it. Hitting 600 entries makes the site's cross-link graph denser without lowering the editorial bar; the same guardrails that protected the original 2026-05-10 batch protect the new material.

A site-wide audit run on 2026-05-10 produced a punch list; the items below were executed the same day. None of these were content additions — they were corrections, deduplications, and infrastructure work.

• Case-insensitive name dedupe (126 records). 126 cat-name records were duplicates of an existing name with different capitalization or trailing whitespace. A new redirects_to_id field was added to deleted_cat_name_ids; 135 rows now return HTTP 410 with a 301 redirect to the canonical row. 4,967 entries in template_cat_names were repointed to the canonical row so internal counts stay accurate.
• Banned-phrase scrub on 87 tip articles (129 replacements). Phrases the editorial guardrails forbid — furry friend, purrfect, when it comes to, dive in, realm of, and similar — were replaced in 87 tip-detail articles. Site-wide count of banned phrases is now zero.
• Sitemap split into five files. The single /sitemap.xml became a sitemap index pointing at sitemap-pages.xml, sitemap-templates.xml, sitemap-content.xml, and sitemap-names.xml. Sub-sitemaps render in 1–4 ms each; only sitemap-names (18,140 URLs) takes about 2.2 seconds. Google can now refetch only the changed sub-sitemap rather than the full set.
• Sitemap expanded by 205 URLs. 86 breed×color cross pages, 32 pair detail pages, 10 trio detail pages, 60 top/breed pages, and 17 static pages that had been launched but not yet listed in the sitemap were added.
• Mojibake repair (42 famous-cat bios). 42 long-form famous-cat bios contained double-encoded UTF-8 byte sequences (€", é, stray Â) inherited from an earlier import. Direct SQL REPLACE chain cleaned all 42; site-wide mojibake count is now zero.
• Thin-name hard-delete (11 records). 8 one-character "names" and 3 multi-line junk submissions (Big Bang Theory lyric copy-paste, a Harry Potter ghost full title, a Norbert-the-dragon novelty entry) were hard-deleted. The other ~4,131 thin rows already carry noindex,follow and were left in place.
• AI-tell scrub on 131 famous-cat long bios. Four review passes reduced em-dash density ≥ 1.5% rows from 61 to 0 (total em-dashes 1,140 → 649, −43%), same-pattern-opening rows from 84 to 1, and tricolon rows from 4 to 2 (the remaining two are regex false positives, not bare-adjective tricolons). Word counts stayed within ±15% of original.
• Quick wins. robots.txt now disallows /scripts/, /docs/, and an admin path. 2 duplicate tip titles culled. 36 template meta_title values exceeding 60 characters were trimmed. /search shows "Showing X of Y" counts. /contact got a working form with honeypot and validation. 4 oversized PNG images (Felix, Mike, Dewey Readmore Books, the Cheshire Cat) were converted to JPG, saving 71% of bytes (4.7 MB → 1.4 MB).

Why: Google's site-wide quality signal is a function of the lowest-quality pages and the messiest infrastructure, not the average. Dedup, banned-phrase scrub, sitemap split, and mojibake repair are unglamorous but cumulative — each one removes a small pattern that the quality classifier can detect.

A second wave of work the same day shipped a dedicated reports section, expanded the famous-cats data with long-form bios, launched the famous-cats hub and detail templates, and reorganised the top navigation.

• Reports section — 7 long-form articles. A new /reports hub plus six reports plus one industry roundup are live. Topics: State of Cat Naming 2026 (flagship, cross of breed × meaning × origin), Names That Crossed: From House Cats to Big Cats, Pop Culture × Cat Naming (analysing the famous-cats table by decade and source type), Religious & Mythological Cat Names Across Cultures, Cat Name Length & Phonetics, Cat Acoustic-Recall Science (citing Heffner 1985, McComb 2009, Saito 2013, Saito 2019, Vitale 2019), and a fair-use industry roundup of Rover, Pumpkin, Healthy Paws, Trupanion, and Wikipedia top-name lists with full attribution. Range: 1,154–2,068 words per article.
• Famous-cats expansion to 360 long-form bios. The famous_cats schema gained source_type, year, parent_work, personality_desc, notable_traits, extended_bio, extended_bio_long, plus six image-attribution fields and a wikipedia_url. Five true duplicates were soft-deleted (Doraemon, Grumpy Cat, Luna SM, Mehitabel, Thomasina). An auto-backfill parser filled 222 source_type values, 141 parent_work values, and 35 year values from existing bios. Three parallel review passes (6 agents, then 5 agents, then 4 agents) produced 300–450 word bios on 360 of 378 active rows (95%) and verified free-license imagery on 89 of 378 (24%). The remaining 18 bio gaps and 271 image gaps are honest-skipped after exhaustive non-Wikipedia source search.
• Famous-cats hub and detail templates. /famous-cats hub with letter filter, source-type chips, and pagination. /famous-cat/{id}/{slug} detail with figure (max-height 360 px), license badge, the long bio, Wikipedia cross-reference, same-name siblings, and same-source recommendations. The famous-cats block on every name detail now links into this new detail layer.
• Header navigation reorganised. The previous 12-link flat header became a 5-dropdown structure (Browse / Most Popular / Tips / Tools / Reports). 14 previously orphaned pages — trios, breed-color crosses, sound-recall, compare, name-match, naming-center, methodology, press, updates, about, contact, and two reports — gained menu entry points.

Why two waves the same day: the reports and famous-cats expansion were the substantive content additions; the quality sprint (see the companion entry) was the infrastructure work that had to happen before the additions were safe to surface in the sitemap. Splitting the two entries keeps the changelog readable.

Major content rewrite across the breed, color, pair, and trio sections, plus a new breed × color cross-page module. Detail of what shipped:

• 60 breed pages rewritten with content synthesized from CFA, TICA, and GCCF official breed standards. Each page now leads with a verifiable specific (registration year, named foundation cat, documented historical event) instead of generic descriptive prose.
• 21 coat-color pages rewritten with cat genetics (B locus, D locus, W locus, O locus, A locus) plus cultural folklore (Bastet, Vox in Rama, Tama the calico, Choupette) and named famous cats. The "Black" page corrects the popular myth that Pope Gregory IX's 1233 bull triggered mass cat killings — modern scholarship has debunked that.
• 86 breed × color cross pages (new feature at /breed-color/<breed>/<color>). Combines breed-specific facts with color-specific facts for combinations like Black Persian, Orange Maine Coon, Calico Siamese, Tortoiseshell Ragdoll. Pages exist only where the database holds at least 75 names linking to both the breed and the color — combinations with thinner overlap are not built.
• Pair-naming themes expanded from 8 to 32. Added 24 new themed pair sets including Bonnie & Clyde, Romeo & Juliet, Yin & Yang, Apollo & Artemis, Espresso & Macchiato, Coco & Chanel, Lily & Rose, Hocus & Pocus, and others. Each set includes cultural origin, pairing rationale, and verifiable historical context.
• Trio-naming mode (new feature at /trios). Ten themed sets for three-cat households, drawn from literature, film, and world mythology: Three Musketeers, Sanderson Sisters, Aristocats Kittens (Marie/Toulouse/Berlioz), Powerpuff Girls, Three Fates, Three Norns, Three Graces, Trimurti, Three Wise Monkeys, Three Stooges.

Why this overhaul: the previous content was visibly AI-template-driven, with formulaic openers ("Bringing home a Maine Coon is a little like adopting a small, furry country gentleman…") that Google's quality classifier flagged as low signal. The rewrite replaces those with research-grounded prose specific to each breed and color, citing verifiable facts. Every page contains at least three citable specifics. Closing CTA shapes are rotated across five categories per page batch so the site does not present a uniform pattern across hundreds of similar URLs.

Methodology and data sources are documented separately on our methodology page (forthcoming).

A four-part data-quality cleanup affecting 20,649 records. None of these were content additions — they were corrections and removals.

• Origin field cleanup (16,138 rows). 16,138 cat-name records had their origin field falsely tagged "Hebrew" because of a default-value import bug — names like Boots, Pancakes, Don Juan, Bjork, and Marmelade all said Hebrew. All 16,138 reset to NULL; the frontend now hides the origin row when NULL. Approximately 100–200 actual Hebrew names lost the tag in this cleanup, which is an acceptable trade-off; the meaning field still preserves their religious context where present.
• Junk-name removal (254 rows, HTTP 410). 254 user-submitted long or garbage names physically deleted. Examples: Steve "I love to eat Papa Johns Pizza", Smoochie Boochie Roochie Toochie, Kaboomkafkakaikajunkalikangarookarakarlo. Rule for deletion: name longer than 25 characters or 4+ words AND no breed-template link AND no famous-cat match. The deleted IDs are tracked in a new deleted_cat_name_ids table; their URLs now return HTTP 410 + X-Robots-Tag: noindex (a stronger deindex signal than 404). 12 long names that ARE knowable cultural references — Sir Nicholas De Mimsey-Porpington, Soft Kitty Warm Kitty Little Ball Of Fur — were retained because they had breed-template links indicating curatorial intent.
• Thin-content noindex (4,146 rows). 4,146 names with all five descriptive fields empty (meaning, short_desc, variants, pronunciation, origin) AND no template link AND no famous-cat match marked noindex,follow. These pages render about 150 words of structural chrome with zero unique content — exactly what Google's quality classifier flags. Sitemap excludes them (18,280 name URLs in sitemap, was 22,680). No data deleted; pages still render normally for direct visitors and internal links — only Google indexing is suppressed.
• Wrong-context short_desc cleanup (111 rows). 111 short_desc rows containing baby-name framing markers — "Hunter is a baby boy name…", "Variant of Greek Alexander", "Saint X", etc. — deleted. These were imported from a baby-name database and had wrong context for cats. The remaining 1,565 short_desc rows are descriptive content that works for cats too (e.g. "Vernon: Alder grove.") and were kept intact.

Why: Google's site-wide quality score is a function of the lowest-quality pages on a site, not just the average. Index pollution from thin pages and obviously-wrong content drags down the entire site. These changes target the bottom of the distribution before adding new content layers.