Weekly Updates / October 4 to October 10, 2026
We shipped a healthcare data product, mapped nine cities and rebuilt the Power catalog.
- Healthcare AI research became a connected 1,191-row product.
- BellyUp! reached nine cities and mapped 1,303 reviewed venues.
- Midland's Power 8 through Power 11 pages became useful planning tools.
This was a week of things becoming useful. Healthcare AI research became a product someone can inspect and buy. BellyUp! added three cities, reached nine and gave all of them useful maps and search. Midland's Power catalog went beyond specifications and started helping people plan real upgrades. The problems we fixed still matter. They just are not the headline when this much progress shipped.
2026-09-10 to 2026-10-07
The numbers at publication
This frozen baseline contains 127 active public properties, five more than the October 3 edition, so it starts a new cohort baseline instead of claiming like-for-like movement. Search Console is complete for September 10 through October 7 with 625 clicks, 165,706 impressions, a ratio-based 0.377 percent click-through rate and an impression-weighted average position of 45.018. The separately validated server-request cache covers the same 28 dates with 271,491 successful content requests. The embedded snapshot records 322 portfolio events dated October 4 through October 10 and stops at event 1653.
- Public properties in this snapshot
- 127
- Portfolio work-log records
- 322
- Google impressions
- 165,706
- Google clicks
- 625
- Click-through rate
- 0.38%
- Average position
- 45.0
This snapshot is saved with the edition. Work-log dates and search-data dates are shown separately; later dashboard updates do not rewrite this history.
Healthcare AI research became a product someone can inspect and buy
The portfolio had years of healthcare AI research spread across useful collections. This week it became something a buyer can actually inspect, sample and license.
Healthcare AI Market and FDA Clearance Dataset v1.1 contains 704 entities and 487 relationships, for 1,191 packaged rows. The release includes eleven CSV tables, JSON, a codebook, a full twelve-sheet workbook and a public twelve-sheet sample workbook.
A buyer can inspect the sample on DatasetsMaker, license the finished release or ask DatasetsBuilder for a custom build. AIHealthcareNow maintains the source collections, while AItoAIDatasets makes the product easier to discover. All four sites passed their live route and rendered Dataset checks.
The important change is not the row count by itself. A collection became a connected product with stable identities, traceable source fields, public evaluation files and a protected licensed release. That is what makes the data usable rather than merely impressive.
BellyUp! reached nine cities and put 1,303 reviewed venues on the map
BellyUp! started the week with six cities. St. Pete became number seven, Savannah became number eight and Charleston became number nine. Together the three new cities added 765 researched venue records, 75 neighborhoods and 47 guides.
Then the feature that worked in St. Augustine spread to the other eight cities. The full network now has zoomable maps, matching search and 1,303 venue records with reviewed coordinates. The build produced 1,691 SVG maps and checked 2,817 canonical pages across 56 hosts.
Twenty venues remain unpinned. That is a success condition, not an unfinished excuse. Their current address could not be supported well enough to guess, so the public maps stay useful without pretending uncertain locations are facts.
The network did more than add cities. It became easier to use. A person can search for a place, see only the matching numbered pins, move from city to neighborhood and trust that an unknown location stayed unknown.
Midland's Power catalog became an upgrade-planning system
Midland's Power pages used to answer the first question: what is this system? This week they started answering the next question: what would it take to own, place, maintain or replace one?
All 64 published Power 10 and Power 11 model and configuration pages now use consistent generation names and clearer headings. The 17 parent model pages also gained sourced Memory, Disk and Adapters panels built for the exact model instead of a generic parts note.
The older generations received the deeper work. All 23 Power 9 parent pages gained sourced upgrade requirements, room-readiness guidance and model-specific parts. All 21 Power 8 parent pages gained lifecycle, replacement and room-readiness guidance, then had every placeholder parts block replaced with sourced rows.
The catalog now maintains six real Dataset families and renders them across all 283 public pages. More importantly, the pages connect specifications to the decision a buyer is trying to make. The catalog is no longer only a record of machines. It is becoming a planning tool.
The public measurement cohort grew to 127 properties, so the comparison resets
This issue freezes 127 active public properties. The previous edition froze 122. Five properties joined the measurement cohort, so the headline totals are a new baseline rather than a like-for-like weekly change.
Google Search Console is fresh through October 7 and contains all 28 dates from September 10 through October 7. The cohort recorded 625 clicks and 165,706 impressions. Click-through rate is 0.377 percent because it is calculated from total clicks divided by total impressions. The impression-weighted average position is 45.018, where lower is better. The parent-child overlap check found zero duplicate page-days.
The separate server-request cache covers the same 28 dates under algorithm version 2. It recorded 271,491 successful requests to content routes: 113,948 search, 14,793 AI search, 135,460 training and 7,290 user retrieval. Those are requests, not proof of indexing, citations, readers or a ranking effect.
The editorial review covered 337 optimization events after the prior cutoff and stopped at event 1653. The embedded weekly snapshot contains the 322 portfolio events dated October 4 through October 10 across 77 primary domains. Fifteen later-ID entries dated October 3 stay in the candidate review but outside the dated weekly count.
The fixes still matter, but they are not the whole week
The portfolio discoverability sweep found 847 records carrying unnecessary robots values, six hardcoded files and one public data folder blocked in robots.txt. The audit now runs fail-closed in both nightly hosting jobs. This one became a real guardrail because the job stopped on the live block and passed only after the public data was available again.
Ten Power contact and quote forms also passed controlled live submissions with ten stored records and ten distinct delivery reports. The repair is real. The recurring portfolio-wide submission and delivery check still does not exist, so this remains a problem found again rather than a lesson learned.
An old local upload dropped all seven Charleston sites from the nightly registry. A server-only learned list now restores missing eligible sites and passed the same stale-list simulation. BuyHumanArt also removed invented records, while a touch test fixed an untappable final menu link across 18 sites.
Those fixes protect the progress above. They belong in the record. They do not need to swallow the record.
What we are watching next
Whether the healthcare dataset earns sample downloads, contact requests or measured visibility before its next source refresh.
Whether BellyUp! Birmingham becomes city number ten and whether the 20 unpinned venues gain a supported current address.
Whether Midland's deeper Power guidance produces quote requests or reveals product questions the catalog still does not answer.
Whether the same 127-property cohort remains intact long enough for a clean like-for-like comparison in the next edition.
Whether the recurring form check is built and whether the nightly robots and learned-registry guardrails keep passing without manual prompting.
Explore the work
- DatasetsMaker: Healthcare AI Market and FDA Clearance Dataset
- BellyUp! St. Pete: live city hub
- BellyUp! Savannah: live city hub
- BellyUp! Charleston: live city hub
- Midland: Power 10 model catalog
- Midland: Power 9 model catalog
- Midland: Power 8 model catalog
- Midland: rendered Dataset receipt
- Buy Human Art: documented artist directory
- Natural Herb Library: public content index
- Power 11: repaired quote checklist
- DataSetSEO: public portfolio directory
- Digital Karma Federation v8.2 specification
Dates, scope, and corrections
This edition covers Sunday, October 4 through Saturday, October 10. The candidate review started after the previous cutoff of 1316 and stopped at event 1653, the highest available ID at capture time. All 337 candidate records were reviewed. The embedded snapshot follows the documented date scope: 322 portfolio events dated inside the coverage window. Fifteen later-ID records dated October 3 remain outside the embedded count.
The 322 dated work records have a broader scope than the public performance cohort. They include follow-ups, approvals, corrections and repeated deployment records across 77 primary domains. Related records are grouped into stories and are not counted as unique projects or successes.
Search totals come from Google Search Console page data for September 10 through October 7. All 28 dates contain observations. The capture found zero duplicate page-days between parent and child properties in the selected cohort. A date with rows does not certify every property import.
The public performance cohort contains 127 active properties. The October 3 issue contained 122, so this edition does not present the two rolling totals as a like-for-like performance change. The exact current domain list is preserved in the embedded snapshot.
Crawler totals come from the separately validated server-request cache for the same 28 dates under algorithm version 2. They count successful GET requests to content routes and preserve search, AI search, training and user retrieval purposes. They do not prove indexing, citations, human readership or a ranking effect.
The three lead stories were selected after reading their full structured notes and verification records. They describe the strongest useful progress a person can now see or use: a connected healthcare data product, a nine-city mapped restaurant network and a Power catalog built around real upgrade decisions. Recurring failures remain in a separate accountability section.
The lead graphic uses verified work receipts from the current event window. Its 1,191 packaged rows, nine cities, 1,303 mapped venues and Power 8 through Power 11 coverage describe shipped work. They are not Search Console measurements or claims about traffic effects.
Public source links are checked over HTTPS before publication. The issue JSON, rendered article, archive, feed, transcript, canonical, metadata, Article and Dataset schema, sitemap, federation catalog, mobile layout, image files and protected infrastructure are verified again after deployment.
: First publication. The event cutoff, public cohort, Search Console metrics, crawler totals and verified recurrence state are preserved as reviewed.
: Editorial correction. The title, introduction, lead stories and graphic now report the week's strongest progress instead of allowing recurring problems to dominate the edition. The frozen snapshot, event cutoff, performance cohort and measurement values are unchanged.