Data · the system

Every number here has a source, an engine and a clock.

54 datasets from 37 publishers feed 33 pages on this site. Scheduled agents refresh the live ones; each publisher’s release calendar holds the rest to account. This page is rebuilt from the data files themselves, so when something falls behind, it says so here before anyone has to ask.

Read from the files at Oct 10, 2026, 7:33 PM MDT. The page rebuilds whenever a feed commits.

12live, refreshed with no one at the keyboard 20on the publisher’s latest release; 22 checked every week 1need a refresh 15 of 54behind a control: 6 gated, 9 tied out to the publisher

Open findings

Anything on the board that isn’t up to date is written up here the way an auditor would: the day it was first flagged, who owns it, the date it is due, and what closes it. It opens and closes with the verdicts; nothing here is typed.

Safe-publish gate receiptOverdue by 21 days
Covers 2 of 4 showcase models.
Opened
Sep 13, 2026
Open for
28 days
Owner
Paul Brown
Fix by
Sep 20, 2026 (7-day window)
Remedy
Rerun dbt build and write-status.mjs.
Holds back The Leak Gate and The Agent-vs-Human Scorecard.

The record

The tiles above are one day. This is every day of the last 90, from the ledger the board’s strips are drawn from: how often the data was current, how long a lapse took to fix, and whether it was fixed on time.

Up to date, last 90 days

79.0%

of 1,430 dataset-days with a clock behind them, Jul 14 – Oct 11. Today’s board reads 32 of 33; this is the average day.

Share up to date, week by week. Low point: 61%, Sep 14–20

97%

Time to recover

11 days

median, across the 17 lapses that closed in the window. The longest: CEO pay ratio filings, 92 days.

Fixed by the deadline

4 of 17

lapses closed inside their fix-by window: 3 days for a feed, 7 for a receipt, 30 for a release. 11 of the 13 misses were feeds.

Traffic on up-to-date data

—

Weighing today’s verdicts by each page’s views…

Days through Sep 27, 2026 were rebuilt from git history with today’s rules, so they judge the past by a standard set after it. Since then a daily job records each day as it closes.

Every lapse that closed in the last 90 days (17)

Dataset	Flagged	Fixed	Days	Fix by	In window
CEO pay ratio disclosures, company by company	Jun 28, 2026refresh due	Sep 28, 2026	92	Jul 28, 2026	No
GDP per capita (PPP, constant)	Jul 31, 2026refresh due	Sep 28, 2026	59	Aug 30, 2026	No
Larder Watch delivery baskets	Sep 19, 2026missed a run	Oct 3, 2026	14	Sep 22, 2026	No
Analytics-leadership job openings	Jul 8, 2026missed a run	Jul 22, 2026	14	Jul 11, 2026	No
CEO-to-typical-worker compensation ratio	Sep 15, 2026refresh due	Sep 28, 2026	13	Oct 15, 2026	Yes
Productivity and typical-worker pay since 1948	Sep 15, 2026refresh due	Sep 28, 2026	13	Oct 15, 2026	Yes
S&P 500, monthly	Sep 15, 2026refresh due	Sep 28, 2026	13	Oct 15, 2026	Yes
Analytics-leadership job openings	Aug 9, 2026missed a run	Aug 22, 2026	13	Aug 12, 2026	No
Ogden-metro rental listings	Aug 2, 2026missed a run	Aug 13, 2026	11	Aug 5, 2026	No
Ogden-metro homes for sale	Aug 2, 2026missed a run	Aug 13, 2026	11	Aug 5, 2026	No
Larder Watch delivery baskets	Sep 7, 2026missed a run	Sep 17, 2026	10	Sep 10, 2026	No
Ogden-metro rental listings	Jul 23, 2026missed a run	Jul 30, 2026	7	Jul 26, 2026	No
Ogden-metro rental listings	Aug 16, 2026missed a run	Aug 22, 2026	6	Aug 19, 2026	No
Ogden-metro homes for sale	Aug 16, 2026missed a run	Aug 22, 2026	6	Aug 19, 2026	No
American Community Survey 1-year detailed tables	Sep 30, 2026refresh due	Oct 5, 2026	5	Oct 30, 2026	Yes
Ogden-metro homes for sale	Jul 23, 2026missed a run	Jul 27, 2026	4	Jul 26, 2026	No
Ogden-metro homes for sale	Sep 29, 2026missed a run	Oct 3, 2026	4	Oct 2, 2026	No

The controls

Current isn’t the same as correct. A clock says the data arrived; a control says it arrived right.

15 of 54

datasets (28%) sit behind a control: 6 pass a gate before they publish and 9 tie out to the publisher’s own figures. Of the 33 on a clock, 15 are covered (45%); the other 21 are snapshots and archives, with no clock to keep.

Tie-outs 11 checks, 7,842 rows, run at every build

A reconcile matches every row of the file a page reads, by key, to the publisher’s own download. A cross-foot recomputes a printed figure from the parts printed beside it. Tolerance is rounding only; one break fails the test suite.

Dataset	What is compared	Result	Last row compared
World Happiness ladder	Cantril ladder score, every country and year, against OWID’s downloadagainst `scripts/sources/happiness-cantril-owid.csv`687 rows have no counterpart: 2005–2010 are reconstructed from the WHR 2023 panel; OWID’s series starts in 2011.	2,089 of 2,089 rows tie	ZWE · 2025: 3.346 vs 3.346
Suicide mortality rates	Suicide rate per 100,000, every country and year, against the World Bank extractagainst `scripts/sources/suicide-who-panel.csv`	4,070 of 4,070 rows tie	ZWE · 2021: 17.34 vs 17.34
Births: final data	Total fertility rate, recomputed as five times the sum of the age-specific rateswithin `public/data/fertility-by-state.csv`	565 of 565 rows recompute	WI 2024: 1,599.5 recomputes to 1,599.5
CEO-to-worker pay ratio	CEO-to-worker ratio (realized), every year, against EPI’s bulk releaseagainst `scripts/sources/epi-ceo-pay-ratio.csv`	61 of 61 rows tie	2025: 324.7 vs 324.677
Productivity vs. pay	Net productivity, cumulative change since 1948, against EPI’s index (1948 = 100)against `scripts/sources/epi-productivity-pay-indexes.csv`	78 of 78 rows tie	2025: 306.8 vs 306.813
Productivity vs. pay	Typical-worker compensation, cumulative change since 1948, against EPI’s indexagainst `scripts/sources/epi-productivity-pay-indexes.csv`	78 of 78 rows tie	2025: 146.9 vs 146.921
S&P 500	December S&P 500 level, every year, against Shiller’s monthly fileagainst `scripts/sources/sp500-shiller.csv`	61 of 61 rows tie	2025: 6,853.03 vs 6,853.03
CEO pay ratio filings	Each disclosed pay ratio, recomputed as the CEO’s pay over the median employee’swithin `public/data/ceo-pay.csv`	16 of 16 rows recompute	GOOGL 2025: 35 recomputes to 35
HCA CEO pay (proxy)	Summary Compensation Table total, re-added from its seven columnswithin `public/data/nurse-ceo-sec.csv`	2 of 2 rows recompute	HCA Healthcare 2025: 26,456,606 recomputes to 26,456,606
HCA CEO pay (proxy)	Disclosed pay ratio, recomputed as that total over the median employee’s paywithin `public/data/nurse-ceo-sec.csv`	2 of 2 rows recompute	HCA Healthcare 2025: 420 recomputes to 420
Nuclear warhead counts	Warheads, every country and year, against the FAS stockpile seriesagainst `scripts/sources/nuclear-stockpiles.csv`	820 of 820 rows tie	USA · 2026: 3,700 vs 3,700

Gates 6 datasets

A gate runs before data publishes and refuses it on a failure, so a bad scrape or a leaky model never reaches a page.

BI seat pricesscripts/bi-pricing-merge.mjs refuses unknown vendors, unknown units and priced rows with no price; history appends, nothing is deleted
Ogden homes for salescripts/housing-validate.mjs refuses a corrupt catalogue before every push
Ogden rentalsscripts/housing-validate.mjs refuses a corrupt catalogue before every push
Strava runsThe dbt leak gate: allow-listed columns only, and no text but the sport enum
Task handling scorecardThe dbt leak gate: allow-list, PII patterns and a contacts oracle
U.S. macro indicatorsscripts/validate-economic-data.mjs checks every series; a failure opens an issue instead of shipping

Not covered yet 18 datasets on a clock

Each of these is kept current, but nothing checks its contents yet. Most were fetched or typed straight into the file a page reads, with no publisher download kept beside it to reconcile against.

ACS state indicators Parent time, by child age Analytics job openings LDS statistical reports Freshness ledger GDP per capita (PPP)Hospital CEO pay (Form 990)Larder Watch baskets LDS membership by country Page view counters Baby names (SSA)Registered nurse wages Release check receipt Gate receipt Fertility, every country Upcoming races Who Americans spend time with World population growth

The map

Publishers on the left, the pages they feed on the right. In between: what moves each dataset onto the site and what it lands as. Several pages share a source, and some sources feed several pages; hover any node to trace it.

9 finished datasets sit in the archive.

Dots move as often as the data does: fast where a job refreshes daily, slow where a publisher releases once a year, still where a dataset is finished. Hover or tab to any node to trace what feeds it. Page dots flash orange when someone opens that page.

The board

Every dataset, with what needs attention first and its last 90 days underneath, one mark a day. Open a row for its publisher, what it holds, where it lands and the check behind its verdict.

Up to dateRefresh dueMissed a runNo clockNo data

Days through Sep 27, 2026 are rebuilt from git history with today’s rules; since then a daily job records each day as it closes. How

Refresh due 1

The publisher has released a newer period than this site holds.

Safe-publish gate receiptCovers 2 of 4 showcase models; rerun dbt build and write-status.mjsdbt gate · 2 pages

Model and test counts from the last full dbt build, which the Leak Gate page draws as its graph. The daily CI run tests the gates but does not rewrite this file, so it can fall behind the models on disk; this board counts them.

Publisher: dbt run results for the showcase models (dbt/scripts/write-status.mjs)
Holds: 2 of 4 models, 27/27 tests
Last 90 days: Jul 14–23: no data · Jul 24 – Sep 12: current · Sep 13 – Oct 9: refresh due · Oct 10: not recorded yet · Oct 11: refresh due
Cadence: Manual — dbt build, then node dbt/scripts/write-status.mjs
Engine: dbt gate: GitHub Actions, dbt-gates.yml, daily 13:17 UTC and on every model change
Who approves: Read live at build or in the browser; never committed
Last written: Jul 22, 2026
Lands in: public/data/dbt_status.json

FeedsThe Leak Gate The Agent-vs-Human Scorecard

Live 12

Written by an unattended job inside its window, or read live at every build.

Analytics-leadership job openingsRefreshed 6 days agoClaude agents · 2 pages

Director-and-up analytics openings by role family, with the posted pay range where the posting carries one. Rows marked as case studies are curated anchors that survive every refresh.

Publisher: Company applicant-tracking boards (Greenhouse, Lever, Ashby, Workday), read by a Claude agent
Holds: Recent openings with their posted pay ranges
Last 90 days: Jul 14–21: missed a run · Jul 22 – Aug 8: live · Aug 9–21: missed a run · Aug 22 – Oct 9: live · Oct 10: not recorded yet · Oct 11: live
Cadence: Weekly, Mondays — mission-control jd-salary-feeder.yml
Engine: Claude agents: mission-control GitHub Actions: housing daily, JD openings Mondays, Larder patrol Wednesdays
Who approves: Commits straight to main, unreviewed; its gate is the only check
Last written: Oct 5, 2026
Lands in: public/data/jd-openings.jsonpublic/data/jd-explorer.jsonpublic/data/jd-openings-digest.md

FeedsJD Salary Tracker What the roles pay

BI vendor list pricesGatedRefreshed 10 days agoClaude agents · 1 page

What each BI vendor prints on its pricing page, in USD per user per month or per month, with the qualifier the page attaches and the date it was read. The agent reads; scripts/bi-pricing-merge.mjs owns the plan keys, units and price history, so a changed price is appended rather than overwritten. Vendors that publish nothing (Sigma, Domo, Mode) and Looker’s per-user seats are carried as quote-only rows. Feeds the vendor picker on the BI line-item page.

Publisher: Ten vendors’ public pricing pages (Tableau, Power BI, Looker, Qlik, ThoughtSpot, Metabase, Hex, Sigma, Domo, Mode), read by a Claude agent
Holds: Cumulative since September 2026, with a price history per plan
Last 90 days: Jul 14 – Sep 29: no data · Sep 30 – Oct 9: live · Oct 10: not recorded yet · Oct 11: live
Cadence: Monthly — mission-control bi-pricing-feeder.yml, the 1st at 08:45 UTC
Engine: Claude agents: mission-control GitHub Actions: housing daily, JD openings Mondays, Larder patrol Wednesdays
Who approves: Commits straight to main, unreviewed; its gate is the only check
Gate: scripts/bi-pricing-merge.mjs refuses unknown vendors, unknown units and priced rows with no price; history appends, nothing is deleted
Last written: Oct 1, 2026
Lands in: public/data/bi-pricing.json

FeedsOff Tableau

Freshness history ledgerRefreshed 2 days agoFreshness check · 1 page

One verdict per dataset per day: what the files on main implied when each UTC day closed, judged by this board’s rules. It draws the 90-day strip on every row. Days through September 27, 2026 were rebuilt in one pass from git history with today’s rules (the registry did not exist before then), using only the publisher checks dated by each day; since then the daily job appends each day after it closes. Recorded days are never rewritten.

Publisher: scripts/build-freshness-history.mjs, run daily by GitHub Actions
Holds: Every day since the repo began (May 24, 2026); through September 27, 2026 rebuilt from git history
Last 90 days: Jul 14 – Sep 27: no data · Sep 28 – Oct 9: live · Oct 10: not recorded yet · Oct 11: live
Cadence: Daily, 14:37 UTC — agent-freshness.yml records the day before
Engine: Freshness check: GitHub Actions, agent-freshness.yml, daily 14:37 UTC
Who approves: Commits straight to main, unreviewed; its gate is the only check
Last written: Oct 9, 2026
Lands in: public/data/freshness_history.json

FeedsData

Larder Watch delivery basketsRefreshed 8 days agoClaude agents · 1 page

Two sworn baskets priced to the same door every week: item prices, delivery and service fees, tax. Never ordered, never tipped, no promos. Set against CPI Food at Home and Food Away From Home from the FRED pipeline.

Publisher: DoorDash storefront prices, priced weekly by a mission-control agent (a Smith’s grocery basket, a Lucky Slice pizza order)
Holds: 2 weeks of prices; grocery last priced 2026-10-03, pizza last priced 2026-10-03
Last 90 days: Jul 14 – Sep 6: no data · Sep 7–16: missed a run · Sep 17–18: live · Sep 19 – Oct 2: missed a run · Oct 3–9: live · Oct 10: not recorded yet · Oct 11: live
Cadence: Weekly — patrol Wednesdays; mission-control larder-feeder.yml exports Thursdays 15:15 UTC
Engine: Claude agents: mission-control GitHub Actions: housing daily, JD openings Mondays, Larder patrol Wednesdays
Who approves: Commits straight to main, unreviewed; its gate is the only check
Last written: Oct 3, 2026
Lands in: public/data/larder.json

FeedsEconomic Dashboard

Ogden-metro homes for saleGatedRefreshed yesterdayClaude agents · 2 pages

For-sale listings within about 10 miles of Ogden, UT, read deterministically from Redfin’s listing data. scripts/housing-merge.mjs owns the keys, price history and status lifecycle (active means seen in the last 7 days), so no model writes the catalogue directly. Since Oct 5, 2026 each map pin is the one Redfin sends; listings no scrape has seen since still sit on their city’s centroid, and the map leaves them off.

Publisher: Redfin listing data (scripts/scrape-buy.mjs); earlier rows from Zillow and Redfin pages read by a Claude agent
Holds: Cumulative since June 2026, with price history
Last 90 days: Jul 14–22: live · Jul 23–26: missed a run · Jul 27 – Aug 1: live · Aug 2–12: missed a run · Aug 13–15: live · Aug 16–21: missed a run · Aug 22 – Sep 28: live · Sep 29 – Oct 2: missed a run · Oct 3–9: live · Oct 10: not recorded yet · Oct 11: live
Cadence: Daily — mission-control ogden-housing-feeder.yml, 09:15 UTC
Engine: Claude agents: mission-control GitHub Actions: housing daily, JD openings Mondays, Larder patrol Wednesdays
Who approves: Commits straight to main, unreviewed; its gate is the only check
Gate: scripts/housing-validate.mjs refuses a corrupt catalogue before every push
Last written: Oct 10, 2026
Lands in: public/data/homes.jsonpublic/data/housing_gate.json

FeedsOgden Home Finder The Scrape Gate

Ogden-metro rental listingsGatedRefreshed yesterdayClaude agents · 2 pages

Rental listings within about 10 miles of Ogden, UT, parsed deterministically from rent.com and merged by scripts/housing-merge.mjs into a cumulative catalogue with first-seen dates, price changes and a status lifecycle: active means a scrape saw it in the last 7 days, then missing, then removed. Since Oct 5, 2026 each map pin is the one rent.com sends; listings no scrape has seen since still sit on their city’s centroid, and the map leaves them off.

Publisher: rent.com (scripts/scrape-rent.mjs); earlier rows from apartments.com and Zillow
Holds: Cumulative since May 2026, with price history
Last 90 days: Jul 14–22: live · Jul 23–29: missed a run · Jul 30 – Aug 1: live · Aug 2–12: missed a run · Aug 13–15: live · Aug 16–21: missed a run · Aug 22 – Oct 9: live · Oct 10: not recorded yet · Oct 11: live
Cadence: Daily — mission-control ogden-housing-feeder.yml, 09:15 UTC
Engine: Claude agents: mission-control GitHub Actions: housing daily, JD openings Mondays, Larder patrol Wednesdays
Who approves: Commits straight to main, unreviewed; its gate is the only check
Gate: scripts/housing-validate.mjs refuses a corrupt catalogue before every push
Last written: Oct 10, 2026
Lands in: public/data/apartments.json

FeedsOgden Home Finder The Scrape Gate

Page view and download countersCounted as people read; this page polls it every few secondsVisitor counters · 1 page

One row per page: views, chart downloads, and when it last moved. No cookies, no identifiers, and local or preview visits never count. The live pulse on this page reads it.

Publisher: This site’s visitors, counted in the mission-control Supabase (page_stats)
Holds: Since July 2026
Last 90 days: Jul 14 – Oct 9: no data · Oct 10: not recorded yet · Oct 11: live
Cadence: Live — one bump per production page view
Engine: Visitor counters: Supabase bump_page_stat RPC, production visits only
Who approves: Read live at build or in the browser; never committed
Lands in: Supabase view page_stats

FeedsData

Release watch receiptRefreshed 6 days agoRelease watcher · 1 page

What each watched publisher had out at the last check, and what the watcher did about it: a pull request for datasets it can refresh itself, one issue for those that need a person. The board’s “last check” dates and verdicts read it, so they are never older than a week.

Publisher: scripts/release-watch.mjs, run weekly by GitHub Actions
Holds: The latest weekly check
Last 90 days: Jul 14 – Sep 27: no data · Sep 28 – Oct 9: live · Oct 10: not recorded yet · Oct 11: live
Cadence: Weekly, Mondays 14:05 UTC — release-watch.yml
Engine: Release watcher: GitHub Actions, release-watch.yml, Mondays 14:05 UTC
Who approves: Opens a pull request; nothing lands until Paul merges it
Last written: Oct 5, 2026
Lands in: public/data/release_watch.json

FeedsData

Strava activity log — long runsGatedRead live from the gated views at every build; fallback snapshots refreshed dailydbt gate · 2 pages

My own Strava activities, synced nightly into the private mission-control database and published only through a dbt view that keeps shapes and drops everything else: local date and start hour, sport, distance, climb, moving and elapsed time, rounded heart rate. Activity names, ids and the HR histogram never leave the database; the leak gate fails the build if any text but the sport enum appears. Race labels are curated in the site repo and joined on date.

Publisher: Strava, via the mission-control Supabase (anonymized showcase_long_runs view)
Holds: June 2020 – present, every Run / TrailRun of 13.1 miles or more (weekly roll-up from March 2020)
Last 90 days: Jul 14 – Sep 12: no data · Sep 13 – Oct 9: live · Oct 10: not recorded yet · Oct 11: live
Cadence: Read live at every site build; fallback snapshots refreshed daily by the ultra-refresh GitHub Action
Engine: dbt gate: GitHub Actions, dbt-gates.yml, daily 13:17 UTC and on every model change
Who approves: Read live at build or in the browser; never committed
Gate: The dbt leak gate: allow-listed columns only, and no text but the sport enum
Last written: Oct 10, 2026
Lands in: Supabase view showcase_long_runsSupabase view showcase_run_weekspublic/data/long-runs.jsonpublic/data/run-weeks.json

FeedsPast Thirteen Miles Race Finder

Task handling, agents vs. meGatedRead live from the gated views at every builddbt gate · 1 page

Who closed each task (an agent or me), how fast, and how often it came back, by week. The views emit only counts and rates; the leak gate checks that no task text or contact name escapes.

Publisher: Mr. Meseeks task database (mission-control Supabase), via anonymized dbt views
Holds: Rolling 90 days
Last 90 days: Jul 14 – Oct 9: live · Oct 10: not recorded yet · Oct 11: live
Cadence: Read live at every site build; views rebuilt daily by dbt-gates.yml
Engine: dbt gate: GitHub Actions, dbt-gates.yml, daily 13:17 UTC and on every model change
Who approves: Read live at build or in the browser; never committed
Gate: The dbt leak gate: allow-list, PII patterns and a contacts oracle
Last written: Jul 22, 2026
Lands in: Supabase view showcase_scorecard_handlerSupabase view showcase_scorecard_weeklypublic/data/task-scorecard.json

FeedsThe Agent-vs-Human Scorecard

U.S. macro indicatorsGatedRefreshed 2 days agoFRED pipeline · 2 pages

Treasuries, mortgage rates, employment, housing, and market series behind the live economic dashboard. Per-series provenance is shown in the dashboard’s own Data Table view; the validator’s verdict is published as data_status.json.

Publisher: FRED — Federal Reserve Bank of St. Louis
Holds: Full history through current month
Last 90 days: Jul 14 – Oct 9: live · Oct 10: not recorded yet · Oct 11: live
Cadence: Weekdays 13:15 UTC — update-data.yml (GitHub Actions)
Engine: FRED pipeline: GitHub Actions, update-data.yml, weekdays 13:15 UTC
Who approves: Commits straight to main, unreviewed; its gate is the only check
Gate: scripts/validate-economic-data.mjs checks every series; a failure opens an issue instead of shipping
Last written: Oct 9, 2026
Lands in: public/data/economic_indicators.csvpublic/data/data_status.json

FeedsEconomic Dashboard Eval Evidence

UltraSignup event finder — upcoming races near OgdenRefreshed yesterdayUltra refresh · 2 pages

The same public search endpoint UltraSignup’s own event finder calls. The script keeps only what the race finder shows — event name, dates, city and state, coordinates, the advertised distance list, registration state, and the event’s UltraSignup link — and drops banner ids and images so the site hotlinks nothing.

Publisher: UltraSignup (service/events.svc/search)
Holds: Rolling twelve months within 250 miles of Ogden, UT
Last 90 days: Jul 14 – Sep 12: no data · Sep 13 – Oct 9: live · Oct 10: not recorded yet · Oct 11: live
Cadence: Daily — scripts/fetch-race-events.mjs, run by the ultra-refresh GitHub Action
Engine: Ultra refresh: GitHub Actions, ultra-refresh.yml, daily 13:45 UTC
Who approves: Commits straight to main, unreviewed; its gate is the only check
Last written: Oct 10, 2026
Lands in: public/data/race-events.json

FeedsRace Finder Past Thirteen Miles

Current 20

Holds the publisher’s latest release, by a dated check or its release calendar.

American Community Survey 1-year detailed tablesWatchedHolds the latest release (2024), checked Oct 5, 2026Release watcher · 1 page

Women’s median age at first marriage, never-married shares by age, bachelor’s degrees and labor-force participation at 25–34, residents born in their state, and median home value against median household income, for every state: the explanations the birth-rate page tests across states.

Publisher: U.S. Census Bureau, via data.census.gov
Holds: 2010–2024
Last 90 days: Jul 14 – Sep 27: no data · Sep 28–29: current · Sep 30 – Oct 4: refresh due · Oct 5–9: current · Oct 10: not recorded yet · Oct 11: current
Cadence: Annual — node scripts/fetch-fertility-acs.mjs after the September ACS release
Engine: Release watcher: GitHub Actions, release-watch.yml, Mondays 14:05 UTC
Who approves: Opens a pull request; nothing lands until Paul merges it
Release watch: Oct 5, 2026: 2025 is not out yet: https://data.census.gov/api/access/data/table?id=ACSDT1Y2025.B12007&g=0100000US answered 400. Refreshes unattended: node scripts/fetch-fertility-acs.mjs
License: Public domain (U.S. government work)
Lands in: public/data/fertility-state-indicators.csv

FeedsBelow Replacement, Even in Utah

American Time Use Survey microdata: parents’ time with each childWatchedHolds the latest release (2025), checked Oct 5, 2026Release watcher · 1 page

Minutes a day a parent spends in the room with one of their own household children, by the child’s age, weighted with the survey’s final weights. No published table has this cut, so the page computes it; run on OWID’s 2010–2024 pool, the same pipeline reproduces OWID’s published time-alone figures exactly. BLS lost 2025’s Sept 30 – Nov 11 diaries to the federal shutdown and reweighted the rest of that quarter.

Publisher: U.S. Bureau of Labor Statistics, ATUS multi-year microdata (respondent, roster, activity and who files)
Holds: 2025
Last 90 days: Jul 14 – Sep 27: no data · Sep 28 – Oct 9: current · Oct 10: not recorded yet · Oct 11: current
Cadence: Annual — node scripts/atus-parent-time.mjs <unzipped files>, then scripts/build-four-thousand-weeks.mjs
Engine: Release watcher: GitHub Actions, release-watch.yml, Mondays 14:05 UTC
Who approves: Opens a pull request; nothing lands until Paul merges it
Release watch: Oct 5, 2026: 2026 is not out yet: https://www.bls.gov/tus/datafiles/atusresp-0326.zip answered 301. Refreshes unattended: node scripts/fetch-atus.mjs
License: Public domain (U.S. government work)
Lands in: scripts/sources/atus-parent-time-by-child-age.csvscripts/sources/atus-parent-time-meta.jsonpublic/data/four-thousand-weeks.json

FeedsFour Thousand Weeks

Annual statistical reportsWatchedHolds the latest release (2025)Hand-verified · 2 pages

Membership, convert baptisms, new children of record, full-time missionaries, wards and branches, stakes and dedicated temples, one row per year and checked against the original reports. Rounded or estimated years, the 1989–90 counting change and the 2008 children-of-record redefinition are flagged in the CSV and handled in src/lib/ldsGrowth.ts.

Publisher: The Church of Jesus Christ of Latter-day Saints (general conference archive, Church Newsroom, Liahona); 1974–1997 membership from the 2013 Church Almanac
Holds: 1974–2025
Last 90 days: Jul 14 – Sep 22: no data · Sep 23 – Oct 9: current · Oct 10: not recorded yet · Oct 11: current
Cadence: Annual — append one row to public/data/lds-church-growth.csv after April general conference
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
Release watch: Calendar only: no machine-readable signal, so the watcher flags it when the release date passes By hand: Append the April statistical report’s year to public/data/lds-church-growth.csv.
Lands in: public/data/lds-church-growth.csv

Feeds50 Years of LDS Growth Below Replacement, Even in Utah

Births: Final Data (national and state birth and fertility rates)Tied outWatchedHolds the latest release (2024)Build scripts · 2 pages

Births, crude birth rates, general fertility rates, total fertility rates and 5-year age-group birth rates. Extracted from the report PDFs by scripts/extract-nchs-natality.py into public/data/fertility-us.csv and fertility-by-state.csv; every row with age rates is checked against TFR = 5 × Σ age rates at extraction and again at build.

Publisher: CDC National Center for Health Statistics, National Vital Statistics Reports
Holds: 2005–2024
Last 90 days: Jul 14 – Sep 27: no data · Sep 28 – Oct 9: current · Oct 10: not recorded yet · Oct 11: current
Cadence: Annual — add the new final report to scripts/extract-nchs-natality.py when NCHS publishes it (mid-year)
Engine: Build scripts: scripts/*.mjs in this repo, rerun when the publisher releases
Who approves: A person commits it through a pull request Paul merges
Tie-out: Total fertility rate, recomputed as five times the sum of the age-specific rates: 565 of 565 rows recompute
Release watch: Calendar only: no machine-readable signal, so the watcher flags it when the release date passes By hand: Add the new “Births: Final Data” report to REPORTS in scripts/extract-nchs-natality.py and rerun it.
License: Public domain (U.S. government work)
Lands in: public/data/fertility-us.csvpublic/data/fertility-by-state.csv

FeedsBelow Replacement, Even in Utah 50 Years of LDS Growth

CEO pay ratio disclosures, company by companyTied outWatchedHolds the latest release (2025), checked Sep 27, 2026Hand-verified · 1 page

The CEO’s total pay, the median employee’s pay and the disclosed ratio, typed from each filing with its link. The ratio is not XBRL-tagged, so no SEC API returns it; it has to be read from the proxy text.

Publisher: SEC EDGAR — each company’s definitive proxy statement (DEF 14A), Item 402(u)
Holds: FY2025–2026
Last 90 days: Jul 14 – Sep 27: refresh due · Sep 28 – Oct 9: current · Oct 10: not recorded yet · Oct 11: current
Cadence: Annual — node scripts/fetch-proxy-ratios.mjs prints each new proxy’s pay-ratio paragraph to re-read; proxies are due within 120 days of fiscal year-end
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
Tie-out: Each disclosed pay ratio, recomputed as the CEO’s pay over the median employee’s: 16 of 16 rows recompute
Release watch: Oct 5, 2026: Every company’s newest proxy on EDGAR is the one the CSV links. By hand: node scripts/fetch-proxy-ratios.mjs prints each newer proxy’s pay-ratio paragraph; retype those rows in public/data/ceo-pay.csv.
Last check: Sep 27, 2026: Every row is the company’s newest definitive proxy on EDGAR as of this check (node scripts/fetch-proxy-ratios.mjs --check: 0 of 16 newer). Source ↗
License: Public domain (SEC filings)
Lands in: public/data/ceo-pay.csv

FeedsHow Far CEO Pay Pulled Away

CEO-to-typical-worker compensation ratioTied outWatchedHolds the latest release (2025), checked Sep 27, 2026Release watcher · 1 page

EPI’s ratio of CEO pay at the largest U.S. firms, stock options counted when realized, to the pay of a typical worker in their industries. The headline chart on the CEO pay page.

Publisher: Economic Policy Institute, State of Working America Data Library (ceo_pay_ratio, realized measure)
Holds: 1965–2025
Last 90 days: Jul 14 – Sep 14: current · Sep 15–27: refresh due · Sep 28 – Oct 9: current · Oct 10: not recorded yet · Oct 11: current
Cadence: Annual — node scripts/fetch-epi.mjs downloads EPI’s bulk release and rebuilds
Engine: Release watcher: GitHub Actions, release-watch.yml, Mondays 14:05 UTC
Who approves: Opens a pull request; nothing lands until Paul merges it
Tie-out: CEO-to-worker ratio (realized), every year, against EPI’s bulk release: 61 of 61 rows tie
Release watch: Oct 5, 2026: Economic/data’s latest release is 2026.9.22, the version the files hold. Refreshes unattended: node scripts/fetch-epi.mjs
Last check: Sep 27, 2026: EPI’s latest bulk release (v2026.9.22, September 23, 2026) ends at 2025: 324.7 to 1 on the realized measure. The same release revised 2024 from 280.0 to 290.2. Source ↗
Lands in: scripts/sources/epi-ceo-pay-ratio.csvpublic/data/pay-gap-ratio.csv

FeedsHow Far CEO Pay Pulled Away

GDP per capita (PPP, constant)WatchedHolds the latest release (2025), checked Oct 5, 2026Release watcher · 1 page

Purchasing-power-adjusted income per person, used to test whether happiness tracked economic growth over the decade.

Publisher: World Bank (NY.GDP.PCAP.PP.KD)
Holds: 2015–2025
Last 90 days: Jul 14–30: current · Jul 31 – Sep 27: refresh due · Sep 28 – Oct 9: current · Oct 10: not recorded yet · Oct 11: current
Cadence: Annual — node scripts/fetch-wb-gdp.mjs, after the July WDI update
Engine: Release watcher: GitHub Actions, release-watch.yml, Mondays 14:05 UTC
Who approves: Opens a pull request; nothing lands until Paul merges it
Release watch: Oct 5, 2026: The World Bank API (updated 2026-07-13) carries NY.GDP.PCAP.PP.KD through 2025. Refreshes unattended: node scripts/fetch-wb-gdp.mjs
License: CC BY 4.0 (World Bank Open Data)
Lands in: scripts/sources/wb-gdp-pc-ppp.csvpublic/data/decade-decoupling.csvpublic/data/decade-trajectories.csv

FeedsThe Happiness Reports

HCA Healthcare CEO pay, Summary Compensation TableTied outWatchedHolds the latest release (2025), checked Oct 3, 2026Hand-verified · 1 page

HCA files no Form 990, so the for-profit comparison takes its CEO’s total from the Summary Compensation Table, which values stock and options when granted, with HCA’s own disclosed CEO pay ratio beside it.

Publisher: SEC EDGAR — HCA Healthcare’s definitive proxy statements (DEF 14A)
Holds: FY2024–2025
Last 90 days: Jul 14 – Oct 3: no data · Oct 4–9: current · Oct 10: not recorded yet · Oct 11: current
Cadence: Annual — typed from each spring’s proxy into scripts/fetch-nurse-ceo-pay.py
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
Tie-out: Summary Compensation Table total, re-added from its seven columns: 2 of 2 rows recompute
Disclosed pay ratio, recomputed as that total over the median employee’s pay: 2 of 2 rows recompute
Release watch: Calendar only: no machine-readable signal, so the watcher flags it when the release date passes By hand: Type the CEO’s row of HCA’s newest Summary Compensation Table and its CEO pay ratio into HCA_PROXIES in scripts/fetch-nurse-ceo-pay.py and re-run it.
Last check: Oct 3, 2026: HCA’s newest DEF 14A, filed March 13, 2026, covers fiscal 2025 ($26,456,606). The page uses fiscal 2024 to match the 990s’ calendar-2024 pay. Source ↗
License: Public domain (SEC filings)
Lands in: public/data/nurse-ceo-sec.csvpublic/data/the-200-year-nurse.csv

FeedsThe 200-Year Nurse

Hospital system CEO pay, Form 990 Schedule JWatchedHolds the latest release (2024), checked Oct 3, 2026Build scripts · 2 pages

Each CEO’s Schedule J, Part II row: base salary, incentive and other pay, retirement and deferred pay, benefits, the total, and the deferred pay earlier returns already reported. Read from the XML of the return, with every other return that reports the same CEO re-read as a cross-check.

Publisher: IRS Form 990 e-file XML (TEOS bulk downloads), with filer pages on ProPublica’s Nonprofit Explorer
Holds: 2024
Last 90 days: Jul 14 – Oct 3: no data · Oct 4–9: current · Oct 10: not recorded yet · Oct 11: current
Cadence: Annual — pin the next year’s returns in scripts/fetch-nurse-ceo-pay.py and re-run it
Engine: Build scripts: scripts/*.mjs in this repo, rerun when the publisher releases
Who approves: A person commits it through a pull request Paul merges
Release watch: Calendar only: no machine-readable signal, so the watcher flags it when the release date passes By hand: Find each system’s next tax-year return in the IRS TEOS index CSVs (apps.irs.gov/pub/epostcard/990/xml/<year>/index_<year>.csv), pin its object id and zip in scripts/fetch-nurse-ceo-pay.py, bump PAY_YEAR and re-run python3 scripts/fetch-nurse-ceo-pay.py.
Last check: Oct 3, 2026: All fifteen tax-year-2024 returns are in the IRS TEOS XML through 2026_TEOS_XML_08A; Mass General Brigham’s (year ending September 2025) was the last. No system has a tax-year-2025 return covering calendar 2025 yet. Source ↗
License: Public domain (IRS filings)
Lands in: public/data/nurse-ceo-990.csvpublic/data/the-200-year-nurse.csv

FeedsThe 200-Year Nurse Less Than Inflation

Membership by countryWatchedHolds the latest release (2025)Hand-verified · 1 page

U.S. membership for the U.S.-vs-world split. Years with no exact published U.S. total (before 1999, 2001–02, 2004–05, 2007, 2020) stay blank rather than estimated; every row carries its own source link in the CSV.

Publisher: The Church of Jesus Christ of Latter-day Saints country facts pages (archived copies via the Internet Archive) and the Deseret News Church Almanac
Holds: 1999–2025
Last 90 days: Jul 14 – Sep 22: no data · Sep 23 – Oct 9: current · Oct 10: not recorded yet · Oct 11: current
Cadence: Annual — append to public/data/lds-us-membership.csv with the statistical-report refresh
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
Release watch: Calendar only: no machine-readable signal, so the watcher flags it when the release date passes By hand: Append the new U.S. year-end membership to public/data/lds-us-membership.csv.
Lands in: public/data/lds-us-membership.csv

Feeds50 Years of LDS Growth

Nuclear warhead stockpiles and inventoriesTied outWatchedHolds the latest release (2026), checked Oct 5, 2026Release watcher · 1 page

Warhead stockpiles by country and year since 1945, plus the 2026 split into deployed, reserve and retired warheads.

Publisher: Federation of American Scientists (Status of World Nuclear Forces), via Our World in Data
Holds: 1945–2026
Last 90 days: Jul 14 – Oct 9: current · Oct 10: not recorded yet · Oct 11: current
Cadence: Annual — node scripts/build-nuclear-arsenal.mjs after OWID’s spring update
Engine: Release watcher: GitHub Actions, release-watch.yml, Mondays 14:05 UTC
Who approves: Opens a pull request; nothing lands until Paul merges it
Tie-out: Warheads, every country and year, against the FAS stockpile series: 820 of 820 rows tie
Release watch: Oct 5, 2026: Our World in Data’s nuclear-warhead-stockpiles runs through 2026. Refreshes unattended: node scripts/fetch-nuclear.mjs
License: CC BY 4.0 (Our World in Data)
Lands in: scripts/sources/nuclear-inventories.csvscripts/sources/nuclear-stockpiles.csvpublic/data/nuclear-arsenal.csvpublic/data/nuclear-breakdown.csv

FeedsNuclear Weapon Inventory

Popular Baby Names: national dataWatchedHolds the latest release (2025), checked Oct 10, 2026Build scripts · 1 page

First names from Social Security card applications. scripts/extract-baby-names.browser.js keeps every year of the 84 names that pass the comeback rule in src/lib/babyNames.ts (count and rank by sex) plus the yearly total of babies of each sex in the file, the denominator for “per 1,000”.

Publisher: U.S. Social Security Administration, Office of the Chief Actuary
Holds: 1880–2025
Last 90 days: Jul 14 – Oct 9: no data · Oct 10: not recorded yet · Oct 11: current
Cadence: Annual — the SSA adds the previous year each spring; re-run the browser extract
Engine: Build scripts: scripts/*.mjs in this repo, rerun when the publisher releases
Who approves: A person commits it through a pull request Paul merges
Release watch: Calendar only: no machine-readable signal, so the watcher flags it when the release date passes By hand: Each spring the SSA adds the year before to names.zip. Its server refuses scripted downloads, so open https://www.ssa.gov/oact/babynames/limits.html in a browser, paste scripts/extract-baby-names.browser.js into the console, run `copy((await extractBabyNames()).comebacks)` and paste into public/data/baby-name-comebacks.csv, then the same with `.totals` for baby-name-totals.csv. npm test re-checks the rule and the pinned headline numbers in src/lib/__tests__/babyNames.test.ts.
Last check: Oct 10, 2026: names.zip read on October 10, 2026 runs 1880–2025. Source ↗
License: Public domain (U.S. government work)
Lands in: public/data/baby-name-comebacks.csvpublic/data/baby-name-totals.csv

FeedsBack from the 1900s

Productivity and typical-worker pay since 1948Tied outWatchedHolds the latest release (2025), checked Sep 27, 2026Release watcher · 1 page

Net productivity against the compensation of production and nonsupervisory workers, as cumulative change since 1948, plus real median hourly wages by gender and race since 1973. epi.org blocks scripts, so the files come from EPI’s GitHub bulk release.

Publisher: Economic Policy Institute, State of Working America Data Library (productivity_and_pay_indexes, hourly_wage_median)
Holds: 1948–2025
Last 90 days: Jul 14 – Sep 14: current · Sep 15–27: refresh due · Sep 28 – Oct 9: current · Oct 10: not recorded yet · Oct 11: current
Cadence: Annual — node scripts/fetch-epi.mjs downloads EPI’s bulk release and rebuilds
Engine: Release watcher: GitHub Actions, release-watch.yml, Mondays 14:05 UTC
Who approves: Opens a pull request; nothing lands until Paul merges it
Tie-out: Net productivity, cumulative change since 1948, against EPI’s index (1948 = 100): 78 of 78 rows tie
Typical-worker compensation, cumulative change since 1948, against EPI’s index: 78 of 78 rows tie
Release watch: Oct 5, 2026: Economic/data’s latest release is 2026.9.22, the version the files hold. Refreshes unattended: node scripts/fetch-epi.mjs
Last check: Sep 27, 2026: EPI’s latest bulk release (v2026.9.22) carries annual productivity–pay measures through 2025, and quarterly through 2026 Q1. Source ↗
Lands in: scripts/sources/epi-productivity-pay-indexes.csvscripts/sources/epi-median-wages-by-group.csvpublic/data/pay-gap-productivity.csvpublic/data/pay-gap-demographics.csv

FeedsProductivity Has Left Wages in the Dust

Registered nurse wages by metro (OEWS)WatchedHolds the latest release (2025), checked Oct 3, 2026Build scripts · 2 pages

Employment and the mean, median and 90th-percentile annual wage of registered nurses in each hospital system’s headquarters metro. The median is the denominator of every ratio on the page.

Publisher: U.S. Bureau of Labor Statistics, Occupational Employment and Wage Statistics, via the BLS public API
Holds: 2025
Last 90 days: Jul 14 – Oct 3: no data · Oct 4–9: current · Oct 10: not recorded yet · Oct 11: current
Cadence: Annual — re-run scripts/fetch-nurse-ceo-pay.py after each spring release
Engine: Build scripts: scripts/*.mjs in this repo, rerun when the publisher releases
Who approves: A person commits it through a pull request Paul merges
Release watch: Calendar only: no machine-readable signal, so the watcher flags it when the release date passes By hand: After BLS publishes the next OEWS release each spring, re-run python3 scripts/fetch-nurse-ceo-pay.py; the BLS API serves only the newest release.
Last check: Oct 3, 2026: The BLS API returns May 2025 estimates for every registered-nurse series and holds no earlier release; bls.gov refuses scripted downloads of the flat files. Source ↗
License: Public domain (U.S. government work)
Lands in: public/data/nurse-rn-wages.csvpublic/data/the-200-year-nurse.csv

FeedsThe 200-Year Nurse Less Than Inflation

S&P 500, monthlyTied outWatchedHolds the latest release (2025), checked Sep 27, 2026Release watcher · 1 page

Each year’s December index level, overlaid on the CEO pay ratio because most CEO pay arrives as stock.

Publisher: Robert Shiller (Yale), via datahub.io
Holds: 1965–2025
Last 90 days: Jul 14 – Sep 14: current · Sep 15–27: refresh due · Sep 28 – Oct 9: current · Oct 10: not recorded yet · Oct 11: current
Cadence: Rebuilt with the EPI series — node scripts/fetch-epi.mjs
Engine: Release watcher: GitHub Actions, release-watch.yml, Mondays 14:05 UTC
Who approves: Opens a pull request; nothing lands until Paul merges it
Tie-out: December S&P 500 level, every year, against Shiller’s monthly file: 61 of 61 rows tie
Release watch: Calendar only: no machine-readable signal, so the watcher flags it when the release date passes Refreshes unattended: node scripts/fetch-epi.mjs
Last check: Sep 27, 2026: scripts/sources/sp500-shiller.csv runs to May 2026, so December 2025 is the latest full year; the chart stops at the EPI series’ last year and moves with it. Source ↗
Lands in: scripts/sources/sp500-shiller.csvpublic/data/pay-gap-ratio.csv

FeedsHow Far CEO Pay Pulled Away

Suicide mortality rateTied outWatchedHolds the latest release (2021), checked Oct 5, 2026Build scripts · 1 page

Deaths from self-harm per 100,000 population, every year the indicator carries. Merged with the Cantril ladder into the tidy country-year core panel (happiness-suicide-panel.csv) that the happiness views derive from. The IHME/OWID series is license-blocked from CSV redistribution, so the World Bank indicator is used.

Publisher: World Bank (SH.STA.SUIC.P5) · WHO age-standardized estimates
Holds: 2000–2021
Last 90 days: Jul 14 – Sep 26: snapshot · Sep 27 – Oct 9: current · Oct 10: not recorded yet · Oct 11: current
Cadence: Periodic (WHO estimate cycles) — scripts/fetch-suicide-panel.mjs
Engine: Build scripts: scripts/*.mjs in this repo, rerun when the publisher releases
Who approves: A person commits it through a pull request Paul merges
Tie-out: Suicide rate per 100,000, every country and year, against the World Bank extract: 4,070 of 4,070 rows tie
Release watch: Oct 5, 2026: The World Bank API (updated 2026-07-13) carries SH.STA.SUIC.P5 through 2021. By hand: node scripts/fetch-suicide-panel.mjs, then build-happiness-panel, build-relative-deprivation and build-suicide-decline.
License: CC BY 4.0 (World Bank Open Data)
Lands in: scripts/sources/suicide-who-panel.csvpublic/data/happiness-suicide-panel.csvpublic/data/suicide-decline-trajectories.csvpublic/data/suicide-decline-changes.csv

FeedsThe Happiness Reports

Total fertility rate and population, every countryWatchedHolds the latest release (2024), checked Oct 5, 2026Release watcher · 2 pages

The world line and fourteen countries for the world chart, and the share of the world’s people living in a country with a TFR under 2.1 (every country with both series, weighted by population). The U.S. line uses NCHS from 1970 because the Bank carries provisional U.S. figures for the latest years.

Publisher: World Bank (SP.DYN.TFRT.IN, from UN World Population Prospects and national statistics; SP.POP.TOTL)
Holds: 1960–2024
Last 90 days: Jul 14 – Sep 27: no data · Sep 28 – Oct 9: current · Oct 10: not recorded yet · Oct 11: current
Cadence: Annual — node scripts/fetch-fertility-world.mjs
Engine: Release watcher: GitHub Actions, release-watch.yml, Mondays 14:05 UTC
Who approves: Opens a pull request; nothing lands until Paul merges it
Release watch: Oct 5, 2026: The World Bank API (updated 2026-07-13) carries SP.DYN.TFRT.IN through 2024. Refreshes unattended: node scripts/fetch-fertility-world.mjs
License: CC BY 4.0 (World Bank Open Data)
Lands in: public/data/fertility-world.csvpublic/data/fertility-below-replacement.csv

FeedsBelow Replacement, Even in Utah 50 Years of LDS Growth

Who Americans spend their time with, by ageWatchedHolds the latest release (2024), checked Sep 27, 2026Build scripts · 1 page

Hours a day alone and with a partner, coworkers, children, friends and other family, by single year of age, for everyone, men and women. OWID pools the survey a year behind BLS, so this series runs one year short of the parent-time curve until OWID’s next update.

Publisher: Our World in Data, from the American Time Use Survey (BLS)
Holds: 2024
Last 90 days: Jul 14 – Sep 22: no data · Sep 23–27: snapshot · Sep 28 – Oct 9: current · Oct 10: not recorded yet · Oct 11: current
Cadence: Annual, when OWID updates (each March) — re-download the CSV, then scripts/build-four-thousand-weeks.mjs
Engine: Build scripts: scripts/*.mjs in this repo, rerun when the publisher releases
Who approves: A person commits it through a pull request Paul merges
Release watch: Oct 5, 2026: OWID’s time-spent-with-relationships-by-age-us was last updated 2026-03-10; next update due 2027-03-10. By hand: Re-download the CSV (URL in scripts/build-four-thousand-weeks.mjs), bump OWID_WINDOW there to the new pooled years, rerun it, and move this probe’s since date.
Last check: Sep 27, 2026: OWID’s chart (updated March 10, 2026; next update due March 10, 2027) still pools 2010–2024: a fresh download matched the committed CSV byte for byte. Source ↗
License: CC BY 4.0 (Our World in Data)
Lands in: scripts/sources/atus-time-with-by-age-owid.csvpublic/data/four-thousand-weeks.json

FeedsFour Thousand Weeks

World Happiness Report — Cantril ladderTied outWatchedHolds the latest release (2025), checked Oct 5, 2026Build scripts · 1 page

Self-reported life satisfaction on a 0–10 ladder, the headline metric of the World Happiness Report. Feeds the tidy country-year core panel (happiness-suicide-panel.csv). The latest snapshot uses the 2023–2025 mean to match the WHR 2026 three-year window; the per-country factor decomposition comes from the official WHR 2026 “Data for Figure 2.1” download, which carries every edition 2011–2025.

Publisher: Gallup World Poll / UN SDSN, via Our World in Data
Holds: 2005–2025
Last 90 days: Jul 14 – Oct 9: current · Oct 10: not recorded yet · Oct 11: current
Cadence: Annual
Engine: Build scripts: scripts/*.mjs in this repo, rerun when the publisher releases
Who approves: A person commits it through a pull request Paul merges
Tie-out: Cantril ladder score, every country and year, against OWID’s download: 2,089 of 2,089 rows tie
Release watch: Oct 5, 2026: Our World in Data’s happiness-cantril-ladder runs through 2025. By hand: Download the OWID ladder CSV and the WHR “Data for Figure 2.1” workbook into scripts/sources/, then run build-happiness-panel, build-relative-deprivation, build-happiness-factors, build-decade-of-happiness and build-suicide-decline.
License: CC BY 4.0 (Our World in Data)
Lands in: scripts/sources/happiness-cantril-owid.csvscripts/sources/happiness-whr26-figure21.csvpublic/data/happiness-suicide-panel.csvpublic/data/happiness-factors.csvpublic/data/relative-deprivation.csvpublic/data/relative-deprivation-latest.csvpublic/data/decade-trajectories.csvpublic/data/decade-decoupling.csv

FeedsThe Happiness Reports

World population growthWatchedHolds the latest release (2025), checked Oct 5, 2026Hand-verified · 1 page

The benchmark behind the per-member growth chart; it replaces an arbitrary 2% line.

Publisher: World Bank (SP.POP.GROW)
Holds: 1974–2025
Last 90 days: Jul 14 – Sep 22: no data · Sep 23 – Oct 9: current · Oct 10: not recorded yet · Oct 11: current
Cadence: Annual, with the statistical-report refresh
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
Release watch: Oct 5, 2026: The World Bank API (updated 2026-07-13) carries SP.POP.GROW through 2025. By hand: Add the new year’s world_pop_growth_pct to public/data/lds-church-growth.csv.
License: CC BY 4.0 (World Bank Open Data)
Lands in: public/data/lds-church-growth.csv

Feeds50 Years of LDS Growth

Snapshot 12

Kept by hand, with no release calendar to hold it to.

Canyonlands Ultra 50 Mile courseBuilt once from the official 2026 course fileBuild scripts · 1 page

The official 50-mile course as a track of coordinates (no elevation or timestamps in the file). The build script measures the course by haversine, finds every pass of each aid station, simplifies the track for drawing, and samples 300 points against the Copernicus 90 m digital elevation model for the profile — so the climb figure is terrain-sampled, not GPS-recorded.

Publisher: Mad Moose Events course GPX (CalTopo export) · Open-Meteo elevation API (Copernicus DEM GLO-90)
Holds: 2026 edition (October 24, 2026), 2,179-point course track
Last 90 days: Jul 14 – Sep 12: no data · Sep 13 – Oct 9: snapshot · Oct 10: not recorded yet · Oct 11: snapshot
Cadence: Static — scripts/build-course.mjs canyonlands-50-mile <gpx>
Engine: Build scripts: scripts/*.mjs in this repo, rerun when the publisher releases
Who approves: A person commits it through a pull request Paul merges
Last written: Sep 13, 2026
Lands in: scripts/sources/canyonlands-ultra-50-mile.gpxpublic/data/courses/canyonlands-50-mile.json

FeedsPast Thirteen Miles

Company filings behind the briefingsVerified per page, June–September 2026; each quarter’s filings move themHand-verified · 1 page

The real figures on the company briefings, each with its filing and as-of date. The funnels on the job-description briefings are illustrative, labeled so on the page, and not on this board.

Publisher: SEC EDGAR 10-K and 10-Q filings, shareholder letters and earnings releases
Holds: Latest quarter as of each page’s verification date
Last 90 days: Jul 14 – Oct 9: snapshot · Oct 10: not recorded yet · Oct 11: snapshot
Cadence: Manual, per briefing
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
Lands in: src/components/dashboards/upbound-growth/growth-data.tssrc/components/dashboards/upbound-intelligence/data.tssrc/components/dashboards/upbound-progressive/progressive-data.tssrc/components/dashboards/upbound-dave/dave-data.tssrc/components/dashboards/upbound-affirm/affirm-data.tssrc/components/dashboards/upbound-sezzle/sezzle-data.tssrc/components/dashboards/husqvarna-sales-analyst/data.ts

FeedsBriefings

Friction and payoff scoresScores set and revised by hand; the page labels them judgment, not measurementHand-verified · 1 page

Prioritization scores (friction 0–100, expected payoff 0–100) for personal interventions, not measured effect sizes.

Publisher: My own read of the evidence, weighted for my situation
Holds: Scored September 2026
Last 90 days: Jul 14 – Sep 12: no data · Sep 13 – Oct 9: snapshot · Oct 10: not recorded yet · Oct 11: snapshot
Cadence: Manual
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
Lands in: src/components/dashboards/leverage-map/data.ts

FeedsLeverage Map

Layoffs, H-1B approvals, LCAs and PERM by employerWatchedResearch handoff of October 8–9, 2026; USCIS FY2026 employer data is expected in late 2026 and would add the next point to every H-1B seriesHand-verified · 1 page

Six tables from the research behind the October 8, 2026 PERM suspension: 419 company-year facts (layoffs by scope, H-1B initial, continuing and total approvals, LCA positions and applications, PERM filings, 10-K headcount) each with source, scope, year basis and confidence; 37 market aggregates; 56 companies; nine public claims checked; 35 dated policy events; 50 offshore signals. src/lib/layoffsVisas.ts picks one row per company-year-metric by confidence and value type.

Publisher: USCIS H-1B Employer Data Hub as tabulated by NFAP and EPI; DOL OFLC LCA Selected Statistics; SEC 10-Ks; WARNTracker; Senate press releases; news reports
Holds: FY2012–2026
Last 90 days: Jul 14 – Oct 8: no data · Oct 9: snapshot · Oct 10: not recorded yet · Oct 11: snapshot
Cadence: One-off — refresh by hand when USCIS publishes FY2026 employer data
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
Release watch: Calendar only: no machine-readable signal, so the watcher flags it when the release date passes By hand: Replace the secondary figures with primary files: USCIS H-1B Employer Data Hub CSVs (initial and continuing approvals by employer), DOL OFLC LCA and PERM disclosure files, and state WARN databases (WA, CA, TX, NY, OR), then re-check src/lib/__tests__/layoffsVisas.test.ts.
License: Public records and published analyses; each row carries its source URL
Lands in: public/data/layoffs-visas-facts.csvpublic/data/layoffs-visas-aggregates.csvpublic/data/layoffs-visas-companies.csvpublic/data/layoffs-visas-claims.csvpublic/data/layoffs-visas-events.csvpublic/data/layoffs-visas-offshore.csv

FeedsLaid Off, Still Sponsoring

LDS General Conference CorpusSearched September 23, 2026, by decadeHand-verified · 1 page

Word counts and phrase frequencies behind the fair recount of names in general conference: Christ’s names and titles and Joseph Smith’s full name per 10,000 words, with the Church’s full name and each talk’s closing line subtracted. It sets the Deseret News tally of one conference beside four decades.

Publisher: Mark Davies, Brigham Young University
Holds: Talks by decade, 1980s–2020s; searched Sept 23, 2026
Last 90 days: Jul 14 – Sep 22: no data · Sep 23 – Oct 9: snapshot · Oct 10: not recorded yet · Oct 11: snapshot
Cadence: Manual
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
Lands in: public/data/lds-conference-mentions.csv

FeedsCounting Christ

Period life tableRead by hand (ssa.gov blocks scripts); SSA revises it with each Trustees ReportHand-verified · 1 page

The odds of reaching each age, by sex, behind the life-in-weeks grid’s survival curve.

Publisher: Social Security Administration, Office of the Chief Actuary
Holds: 2023
Last 90 days: Jul 14 – Sep 22: no data · Sep 23 – Oct 9: snapshot · Oct 10: not recorded yet · Oct 11: snapshot
Cadence: Manual — read once in a browser and saved to scripts/sources/
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
License: Public domain (U.S. government work)
Lands in: scripts/sources/ssa-period-life-table-2023.csv

FeedsFour Thousand Weeks

Religious Landscape Study: state estimatesOne-off survey waves; the next is unscheduledHand-verified · 1 page

The Latter-day Saint share of Utah adults: 58%, 55%, 50%. The 2023–24 wave moved from phone to mostly online and paper questionnaires, which Pew flags as a mode change.

Publisher: Pew Research Center
Holds: 2007, 2014 and 2023–24 waves (Utah n = 323, 315, 340; margins ±6.0, ±7.1, ±7.3 points)
Last 90 days: Jul 14 – Sep 27: no data · Sep 28 – Oct 9: snapshot · Oct 10: not recorded yet · Oct 11: snapshot
Cadence: One-off; refresh with the next Religious Landscape Study
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
Lands in: src/data/birthRates.ts

FeedsBelow Replacement, Even in Utah

State population estimates by age and sexPinned to the 2008 and 2024 ends of the birth-rate analysis; moves when NCHS 2025 landsBuild scripts · 1 page

The weights for splitting a general fertility rate change into birth rates and age mix (Kitagawa). Rebuilt GFRs match NCHS’s within a point in every state.

Publisher: U.S. Census Bureau (intercensal estimates 2000–2010; Vintage 2024 estimates)
Holds: Women by 5-year age group, every state, 2008 and 2024
Last 90 days: Jul 14 – Sep 27: no data · Sep 28 – Oct 9: snapshot · Oct 10: not recorded yet · Oct 11: snapshot
Cadence: Manual — node scripts/fetch-fertility-women.mjs
Engine: Build scripts: scripts/*.mjs in this repo, rerun when the publisher releases
Who approves: A person commits it through a pull request Paul merges
License: Public domain (U.S. government work)
Lands in: public/data/fertility-women-by-age.csv

FeedsBelow Replacement, Even in Utah

Super PAC committee filingsCommittee facts read by hand from FEC pagesHand-verified · 1 page

Committee ids, types and filing facts behind the structure picker’s worked examples.

Publisher: Federal Election Commission committee pages
Holds: Three Utah committees and national examples
Last 90 days: Jul 14–23: no data · Jul 24 – Oct 9: snapshot · Oct 10: not recorded yet · Oct 11: snapshot
Cadence: Manual
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
License: Public domain (U.S. government work)
Lands in: src/pages/tools/super-pac/index.astro

FeedsSuper PAC Structure Picker

Tech & AI compensationRefreshed by hand with npm run update:salary; the source adds rows continuouslyBuild scripts · 1 page

Base salaries from ai-jobs.net postings by title, experience level and location, behind the salary explorer. Roles are regex-classified, salaries trimmed to each role’s 5th–95th percentile, and any year/role or country/role group under 30 rows is dropped from medians and trends.

Publisher: ai-jobs.net salary postings (foorilla/ai-jobs-net-salaries on GitHub)
Holds: 2020–2025
Last 90 days: Jul 14 – Oct 9: snapshot · Oct 10: not recorded yet · Oct 11: snapshot
Cadence: Manual snapshot — npm run update:salary (scripts/update-salary-data.mjs)
Engine: Build scripts: scripts/*.mjs in this repo, rerun when the publisher releases
Who approves: A person commits it through a pull request Paul merges
Lands in: public/data/salary_aijobs_summary.jsonpublic/data/salary_aijobs_clean.csv

FeedsWhat the roles pay

The Brigham nurses’ strike: proposals, MGB’s finances, Massachusetts CEO payWatchedThe public record as of October 9, 2026, five days before the strike date; the two sides’ positions are their October 8 proposalsHand-verified · 1 page

Twenty-nine typed figures with their source: MGB’s audited fiscal-2025 lines and layoff numbers, each side’s costing of the dispute, Klibanski’s pay by calendar year, seven Massachusetts system CEOs’ 2024 pay as the Globe read the 990s, and Brigham’s top-step hourly rates under MGB’s offer. The CEO’s 2024 pay and the Boston nurse wage come from the-200-year-nurse.csv.

Publisher: Massachusetts Nurses Association releases; Mass General Brigham’s nursing-updates pages; The Boston Globe, WBUR, NBC Boston, CommonWealth Beacon; Becker’s and Fierce Healthcare on the audited statements; Lown Institute and the Paddock Post on earlier 990s
Holds: Fiscal 2025 results; calendar 2019–2024 CEO pay; the dispute through October 9, 2026
Last 90 days: Jul 14 – Oct 9: no data · Oct 10: not recorded yet · Oct 11: snapshot
Cadence: One-off — refresh by hand when the strike settles or the next 990s post (August)
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
Release watch: Calendar only: no machine-readable signal, so the watcher flags it when the release date passes By hand: After the strike settles (or MGB posts fiscal-2026 results, in December), re-check the figures in src/lib/brighamStrike.ts against the sources it names, regenerate public/data/mgb-nurse-strike.csv (csvRows() in that file) and re-run src/lib/__tests__/brighamStrike.test.ts.
License: Public records and published reports; each row carries its source URL
Lands in: public/data/mgb-nurse-strike.csv

FeedsLess Than Inflation

Tradition and Change Among Latter-day Saints in the U.S.One-off survey; refresh with the next Religious Landscape StudyHand-verified · 2 pages

Retention of Americans raised Latter-day Saint (70% → 64% → 54%), the share now unaffiliated, retention by age, switching into and out of the Church as shares of all U.S. adults, and family-life ratings by religious group — the survey figures set beside the Church’s own counts. The birth-rate page uses its family and education figures: Latter-day Saints with a child under 18 at home (49% → 37%, against 35% → 28% for all adults) and women with a bachelor’s degree (25% → 40%).

Publisher: Pew Research Center (Religious Landscape Studies 2007, 2014, 2023–24)
Holds: Released Sept 22, 2026; 2023–24 fieldwork July 17, 2023 – March 4, 2024 (565 Latter-day Saints, 316 former members)
Last 90 days: Jul 14 – Sep 22: no data · Sep 23 – Oct 9: snapshot · Oct 10: not recorded yet · Oct 11: snapshot
Cadence: One-off; refresh with the next Religious Landscape Study
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
Lands in: src/data/ldsClaims.tssrc/data/birthRates.ts

Feeds50 Years of LDS Growth Below Replacement, Even in Utah

Archive 9

Finished. Nothing upstream will change it.

Faith and evolution interviewsPublished study dataHand-verified · 1 page

Coded interviews on how BYU students hold faith and evolution together.

Publisher: BYU ScholarsArchive, dataset 98 (student interview transcripts)
Holds: 48 students, 38 quoted passages
Last 90 days: Jul 14 – Oct 9: archive · Oct 10: not recorded yet · Oct 11: archive
Cadence: Finished
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
Lands in: public/data/faith-evolution-students.csvpublic/data/faith-evolution-quotes.csv

FeedsFaith and Human Evolution: Mental Gymnastics

FIFA World Cup 2026 bracketTournament over: Spain won the final on July 19, 2026Hand-verified · 1 page

The 48 qualified nations and each knockout round, cross-checked against CBS Sports and Yahoo Sports, plus every champion since 2006. Powers the World Cup lens on the happiness paradox chart — each round becomes a country grouping with its own extremes, cluster mix, and trend vs. the world — and the "does winning make a country happier" strip. England and Scotland join to United Kingdom data.

Publisher: FIFA, via Wikipedia
Holds: 2026 tournament, group stage through the final (as of Jul 19, 2026)
Last 90 days: Jul 14 – Oct 9: archive · Oct 10: not recorded yet · Oct 11: archive
Cadence: Finished — updated by hand as rounds completed (scripts/sources/world-cup-2026.json)
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
Lands in: scripts/sources/world-cup-2026.jsonscripts/sources/world-cup-champions.jsonpublic/data/world-cup-2026.json

FeedsThe Happiness Reports

Mortality on the Mormon Trail, 1847–1868Historical record, 1847–1868Hand-verified · 1 page

Deaths on the overland emigration by age, cause, year and company, and arrivals by way of travel.

Publisher: BYU Studies (“Mortality on the Mormon Trail, 1847–1868”) and Demographic Research vol. 52, article 10
Holds: Companies arriving 1847–1868
Last 90 days: Jul 14 – Oct 9: archive · Oct 10: not recorded yet · Oct 11: archive
Cadence: Finished
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
Lands in: public/data/mormon-pioneers-age.csvpublic/data/mormon-pioneers-mortality.csvpublic/data/mormon-pioneers-causes.csvpublic/data/mormon-pioneers-arrivals.csvpublic/data/mormon-pioneers-method.csvpublic/data/mormon-pioneers-companies.csv

FeedsMormon Pioneer Trek to the Salt Lake Valley

Personal listening historyThe export ends in July 2023Build scripts · 1 page

Every play with artist, album and time of day, rolled up into artists, eras and listening hours.

Publisher: My own streaming export
Holds: September 2019 – July 2023, 47,117 plays
Last 90 days: Jul 14 – Oct 9: archive · Oct 10: not recorded yet · Oct 11: archive
Cadence: Finished — npm run build:music rebuilds it from the export
Engine: Build scripts: scripts/*.mjs in this repo, rerun when the publisher releases
Who approves: A person commits it through a pull request Paul merges
Lands in: scripts/sources/music-listening-history.csvpublic/data/topology-of-music-meta.jsonpublic/data/topology-of-music-artists.csvpublic/data/topology-of-music-albums.csvpublic/data/topology-of-music-clusters.csvpublic/data/topology-of-music-loyalty.csvpublic/data/topology-of-music-stream.csvpublic/data/topology-of-music-hours.csvpublic/data/topology-of-music-monthly.csvpublic/data/topology-of-music-tracks.csv

FeedsTopology of Music

Semantic-layer benchmarkPublished paper, April 2026Hand-verified · 1 page

Accuracy of model-written SQL with and without a semantic layer, from the published results.

Publisher: Cube, arXiv 2604.25149 (April 28, 2026), and its public benchmark repository
Holds: April 2026 paper
Last 90 days: Jul 14 – Aug 26: no data · Aug 27 – Oct 9: archive · Oct 10: not recorded yet · Oct 11: archive
Cadence: Finished
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
Lands in: src/components/dashboards/the-page-was-worth-twenty-points/PageWorthTwentyPoints.tsx

FeedsThe Page Was Worth Twenty Points

Swedish vacations and antidepressant dispensingPublished study, 2013Hand-verified · 1 page

The study’s ARIMA coefficients, recomputed into its seasonal curve; no raw data is redistributed.

Publisher: Hartig et al. (2013), Society and Mental Health — the published model coefficients
Holds: Monthly series, January 1993 – March 2005
Last 90 days: Jul 14 – Oct 9: archive · Oct 10: not recorded yet · Oct 11: archive
Cadence: Finished
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
Lands in: src/components/dashboards/vacation-antidepressants/VacationAntidepressants.tsx

FeedsWhen Sweden Vacations Together

The Soviet continuous work weekHistorical record, 1929–1940Hand-verified · 1 page

The colour order, rest-day rotation and adoption figures (72.9% of workers on October 1, 1930) from the surviving 1930 calendar and the historical record.

Publisher: Wikipedia (Soviet calendar); Cabinet Magazine, issue 61; History.com
Holds: 1929–1940
Last 90 days: Jul 14 – Oct 9: archive · Oct 10: not recorded yet · Oct 11: archive
Cadence: Finished
Engine: Hand-verified: Read from the filing or report itself
Who approves: A person commits it through a pull request Paul merges
Lands in: src/components/dashboards/nepreryvka-week/NepreryvkaWeek.tsx

FeedsThe Week That Had No Weekend

World country outlinesReference geometryBuild scripts · 2 pages

Country outlines projected once at build into the SVG paths the world maps draw.

Publisher: Natural Earth 1:110m, via the world-atlas package
Holds: 1:110m admin-0 countries
Last 90 days: Jul 14 – Oct 9: archive · Oct 10: not recorded yet · Oct 11: archive
Cadence: Static — scripts/build-world-paths.mjs
Engine: Build scripts: scripts/*.mjs in this repo, rerun when the publisher releases
Who approves: A person commits it through a pull request Paul merges
License: Public domain (Natural Earth)
Lands in: scripts/sources/countries-110m.jsonpublic/data/world-paths.json

FeedsThe Happiness Reports Nuclear Weapon Inventory

World Happiness Report 2019 (combined)The 2019 snapshot is fixed by designBuild scripts · 1 page

The original combined file — happiness score, six explanatory factors, plus suicide rate, Gini, and life expectancy — behind the 2019 snapshot of the Relative Deprivation dashboard.

Publisher: Kaggle / UN SDSN
Holds: 2019 report (147 countries)
Last 90 days: Jul 14 – Oct 9: archive · Oct 10: not recorded yet · Oct 11: archive
Cadence: Static
Engine: Build scripts: scripts/*.mjs in this repo, rerun when the publisher releases
Who approves: A person commits it through a pull request Paul merges
Lands in: scripts/sources/happiness-2019.csv

FeedsThe Happiness Reports

The pulse

Page views as they land, from the same counters every page’s info bar reads. A check that the site is being read, not a measure of anything above.

Live

— page views counted across the site. Latest: listening…

Polled every few seconds from the same counters every page’s info bar reads. No cookies, no identifiers; local and preview visits never count.

How the verdicts work

A feed is fresh while the stamp its own job wrote is inside a grace window (two days for a daily scrape, eight or nine for a weekly one). Past the window it has missed a run, and thedead-man’s switch alerts me a day later.

A release is current when the file holds the latest period the publisher has put out. That period comes from a dated check of the source when there is one, and from the publisher’s release calendar once the calendar moves past that check. Until the next release date, the check outranks the calendar: a publisher running late turns nothing amber, and one running early does. After it, the check has gone stale and the calendar takes over until someone looks again.

The release watcher asks every watched publisher what it has out each Monday: an API row, a chart’s update date, a release tag, whether next year’s file exists. A check that sees the publisher’s latest period replaces the typed one, so these verdicts are never more than a week old (last run Oct 5, 2026). When something new lands and the dataset has a committed fetch script, the watcher runs it, runs the tests and the build, and opens a pull request; what needs a person goes into one issue. A release the watcher refreshes reaches the site only when someone merges its pull request. The feeds are different: each scheduled job commits straight tomain, unreviewed, and its gate is the only check in the way. Every row on the board says which path its data took.

The strips are the same verdicts, one per day. Each day is judged on what the data files on main held when that day closed (UTC), and a daily job appends it to a committed ledger (public/data/freshness_history.json); recorded days are never rewritten. The last mark is this build’s own verdict. Days through Sep 27, 2026 were rebuilt in one pass from git history, because the registry didn’t exist before then: they apply today’s rules to the files as they were, with only the publisher checks dated by each day, so a release with no calendar reads as having no clock until the day it was first checked. A day with no file to read is an outline, never green; a dataset that lives only in a database has no history git can see.

The record counts dataset-days, not datasets: every dataset with a clock adds one verdict a day, and the share up to date is the days that were live or current over all of them. A lapse is a run of days at refresh due or missed a run. It is fixed on the first up-to-date day after it, and fixed on time when that day falls inside its window, counted from the first day it was flagged: 3 days for a feed, 7 for a receipt, 30 for a release. A lapse still open on the build’s day is an open finding. The traffic tile weighs each page’s views to date by whether every dataset it reads is up to date today. The counters keep totals, not dates, so it is a weight on today’s verdicts, not a history.

A tie-out reads two files at every build and compares them row by row: the file a page reads and the publisher’s own download committed beside it (a reconcile), or a printed total and the printed parts it should equal (a cross-foot). Its tolerance covers rounding and nothing else, and one row out of tolerance fails the test suite before a deploy. Where a page reads a figure the build derived, the tie-out runs against the derived file, so the build script is checked too.

The registry (src/data/sources.ts) is the only list. A test fails the build if a page reads a data file the registry doesn’t claim, if an entry names a file or page that doesn’t exist, or if the alarm thresholds drift from the grace windows on this board. Where a dashboard needs a derived field (a percentile, a cluster, a ratio), a committed script re-derives it, so every number can be reproduced.

Every number here has a source, an engine and a clock.

Open findings

The record

The controls

Tie-outs 11 checks, 7,842 rows, run at every build

Gates 6 datasets

Not covered yet 18 datasets on a clock

The map

Claude agents scheduled, unattended

FRED pipeline GitHub Actions

Ultra refresh GitHub Actions

dbt gate tested before publish

Release watcher GitHub Actions

Build scripts rerun on each release

Hand-verified read from the source

The board

The pulse

How the verdicts work