This is an archived briefing. It was published on October 29, 2026 and has not been changed since. Every figure on this page was true on that date and is not kept current. For where the platform stands today, read the current briefing.

The provenance debt is paid. 207,185 budgets now say where their numbers came from.

Where the Empowered Vote platform stands, in numbers anyone can check. This is a two-day cycle. Almost all of it went into Treasury, and the rest went into finding that we had been counting some people twice.

31,626
stance positions a voter is served, up 33 — the first rise since the audit began
11,188
governments with budget data, up 21
38,486
budgets still ungraded, down from 245,604
3,298
playable trivia questions, down 8
8,494,248
recorded legislative votes, unchanged this cycle
~900
commits landed in the last 30 days

A debt we reported unpaid for three editions

207,185 budgets gained a provenance grade this cycle. The pile of ungraded budgets falls from 245,604 to 38,486. A provenance grade says where a budget figure came from and who, if anyone, checked it. “Unknown” was never a grade on that scale: it meant nobody had recorded the answer. We had printed that number unchanged since October 20. The governments holding at least one ungraded budget fall from 3,420 to 1,177. The pile is not 38,419, which is where the grading run left it. A separate load wrote 67 more ungraded budgets a few hours later, and the figure we print is the one after that.

They did not just get a label. All 207,494 rows the run touched gained a source address and a stated basis as well. Of those, 207,113 are graded self-reported and unaudited, and 381 are audited to government accounting standards. The arithmetic closes to the row against the 583,565 budgets we hold. We checked the run had finished before counting any of it, and waited out a second load that started later.

Stance research restarted, after a cycle that added nothing. 43 politicians gained their first researched position. The audit wrote 112 rows and 28 of them were withdrawals, against 459 rows and 252 withdrawals last cycle. Positions a voter is served rose 31,593 to 31,626, the first rise since the audit began. Two map tiles moved, both upward: Alaska from 2 researched officials to 10, and Pennsylvania from 5 to 19.

And we found that we had been counting seven people twice. Eight federal offices carry no chamber in our data: the DC Delegate, its two Shadow Senators, and the Delegates for five territories. The rule that finds officials belonging to no government was shading them onto a state tile, while the coverage table counted them in its national row. They sit in Congress, so the row was right and the tiles were wrong. The table had the opposite fault, and it is the more interesting one. We record a candidacy for the US Senate as a national office, so four sitting state officials who are running were filed as members of Congress. One of them is the Attorney General of Texas. Both faults are fixed, and the numbers they moved are below.

Coverage across the map

Three readings of the same 52 boxes. The boxes never move; only the fill changes. Put a tab beside another and the gap between them is the work left to do.

Officials on file

Shading shows how many officials each jurisdiction has on file, researched or not. This is the denominator the first tab is a fraction of, and every one of the 52 tiles is shaded. Seating a jurisdiction is the cheap half; researching it is the expensive half. Puerto Rico is the clearest case: 160 officials here, an empty outline on the stance map. California holds 2,030 and has researched 376. Fifteen jurisdictions have researched every official they hold, and all fifteen hold nine or fewer.

AK
ME
VT
NH
WA
ID
MT
ND
MN
IL
WI
MI
NY
RI
MA
OR
NV
WY
SD
IA
IN
OH
PA
NJ
CT
CA
UT
CO
NE
MO
KY
WV
VA
MD
DE
AZ
NM
KS
AR
TN
NC
SC
DC
OK
LA
MS
AL
GA
HI
TX
FL
PR
Deep: 100+ officials (30 jurisdictions) Growing: 10 to 99 (3) Thin: 1 to 9 (19) None on file (0)

Hover any state for its officials on file and how many of them are researched. The 52 tiles hold 9,893 officials, 2,830 of them researched. Federal seats are on none of these maps. They are a national row, not a state one, and the coverage table below counts them. 1,481 researched people sit on no tile at all. 548 hold a federal seat and 1,327 are candidates in a race. Those overlap, because a member running again is both. Only 34 are neither.

Stance research

Every state has at least one researched official. Shading shows the depth of state and local research, and federal seats are counted in the national row of the coverage table below rather than here. Two tiles moved this cycle, both upward: Alaska from 2 researched officials to 10, and Pennsylvania from 5 to 19. One tile moved for a different reason. The District of Columbia reads 14 where the last edition said 16, because two of the people behind that number sit in Congress. Puerto Rico is still an outlined tile, with 160 officials on file and no researched positions yet.

AK
ME
VT
NH
WA
ID
MT
ND
MN
IL
WI
MI
NY
RI
MA
OR
NV
WY
SD
IA
IN
OH
PA
NJ
CT
CA
UT
CO
NE
MO
KY
WV
VA
MD
DE
AZ
NM
KS
AR
TN
NC
SC
DC
OK
LA
MS
AL
GA
HI
TX
FL
PR
Deep: 100+ officials (12 jurisdictions) Growing: 10 to 99 (12 jurisdictions) Seeded: statewide officials researched (27) Seats seeded, research not started (1)

Hover any state for its exact count. The seeded tier is mostly governors and statewide officers; deeper tiers add legislatures, counties and cities.

Treasury Tracker coverage

Shading shows how many governments in each state have a published budget in Treasury Tracker. Every state has at least one, and in 21 of them only one: the state government. The depth is wildly uneven by design. A state whose comptroller publishes every municipality at once arrives in one piece. That is why Pennsylvania reads 2,620, New York 2,331 and Michigan 1,857 while their neighbours read single digits. This is a map of what has been published to us, not of what exists. The District of Columbia and Puerto Rico have no budgets yet, so both are outlined.

AK
ME
VT
NH
WA
ID
MT
ND
MN
IL
WI
MI
NY
RI
MA
OR
NV
WY
SD
IA
IN
OH
PA
NJ
CT
CA
UT
CO
NE
MO
KY
WV
VA
MD
DE
AZ
NM
KS
AR
TN
NC
SC
DC
OK
LA
MS
AL
GA
HI
TX
FL
PR
Deep: 100+ governments (11 jurisdictions) Growing: 10 to 99 (6) Thin: 1 to 9 (33) No budgets published yet (2)

Hover any state for its exact count. The 52 tiles hold 11,187 governments between them; the federal government makes 11,188, which is the roster counted in the Treasury section below.

All three maps were queried on the same day as every other figure on this page. The two added on October 29 carried their own date for one edition, because they were added to a page that had already been published. That is over: one edition, one as-of, and our refresh script now refuses to run while any such note is on the page.

Corrections

What this edition fixes in the one before it. A published edition is never edited; the next one says what it corrects.

Four figures from October 27 are corrected here. Its stance map read 16 researched officials for the District of Columbia; the right number is 14. It read 374 for California; the right number is 376. In the coverage table, the national row read 556 and should read 552, and Texas read 270 and should read 272. Nobody left Congress. The lead above says why.

One of the seven is worth naming, because only making two numbers agree would have caught her. Aisha Wahab left the California state Senate for Congress. Our records still carry her old seat, and the map read it. A current term now beats a stale one. With all of it fixed, every one of the 51 state rows in the coverage table matches the map person for person. That was not true before, and nothing on the page would have told you.

We said on October 27 that we would report the provenance debt the week it was paid. It was paid, and that is the headline above. Two other claims from that edition were re-derived and both held. The certified corpus is still a single stratum of 57 state-level record-based stances. Puerto Rico still has 78 alcaldes, and 70 of them still have no portrait.

Our numbers were taken against a quiet database. The last stance answer was written seven hours before we counted, the last budget row six hours before, and the last politician record a day before. That check exists because the October 20 edition counted during a load and published two partial figures. This cycle it mattered: the grading run below wrote 198,068 rows in a single hour, and counting inside that hour would have published any number we liked.

Inform Pillar: Shared Facts

Open to everyone, no login. Evidence-based, engaging, anti-partisan.

FeatureStatusJourneyWhere it standsWhat's next
Empowered Compass Live
alpha
Where it standsSeason 3 moved, and it is still a draft. It held 74 answers on October 22 and holds 107 now. Those answers are the newest on their pairs. So they are already inside the served figure at the top of this page, written under a season no voter has been told about. We check the season table every cycle because an opening is a launch no other figure announces; this is the quieter version of the same risk. Season 2 is open, Season 1 closed. The library carries 64 topics and the open season serves 60. Compass logged 1 commit. What’s nextThe Season 1 shadow fell again, and again barely. 87.9% of served positions come from the closed season, against 88.1% two days ago: 27,806 positions of 31,626. Season 2 supplies 3,713 and the draft Season 3 supplies 107. Decide what a draft season is allowed to serve. Either those 107 answers belong to a season a voter can see, or they should not be counted as positions a voter is served. Today they are counted, which is the honest reading of our own definition and probably not the one we want. The “24 rungs” figure we withdrew two editions ago stays withdrawn, because nothing in the repository reproduces it.
Empowered Essentials Live Alpha
alpha
Where it standsZIP-to-representatives with funding, votes and stances. No commits this cycle. A researched person with no answers now gets the breakdown on its own. A chosen light or dark theme is published to a shared cookie, so the next Empowered Vote app a reader opens inherits it. 13,396 curated officials and candidates are served, beside a 77,729-record discovery pool still in triage. What’s nextStill owed, and unchanged for a seventh cycle: portraits for the 70 Puerto Rico alcaldes without one. Re-derived this week and exact. 78 municipios carry a mayor, 70 have no default image, and the island’s other 83 seats all do. Essentials leads the Compass on its own action for a third pull, 48 people to 43.
Treasury Tracker Live
alpha
Where it standsNo budgets were loaded, for a second cycle running. Every Treasury figure here is unchanged to the row: 11,188 governments and 10.4 million line items. 14 commits, the busiest repository this cycle. A scheduled sweep now fails the run when a database function is executable by anonymous callers. Search results say what kind of place each one is and where it sits, so two Atlantas can be told apart. The page itself is timed at last, and those numbers are at the top of this report. What’s nextGrade the budgets. Not one row moved, for the second edition running. 245,604 budgets carry an unknown provenance grade, the same figure to the row, and zero were touched this cycle. With seven days to the election these are in front of readers now. We can say we hold them and we cannot say how good they are. The comparison gate still has not been asked whether a school district may be set beside a city.
Civic Trivia Championships Live
alpha
Where it standsThe question we raised two editions ago has been answered, and answered well. A numeric question can now carry a seven-mark ascending scale, and the server rolls which four of those marks a player sees. The options still read in order, and the right answer lands on each letter about equally often across players. 509 of the live questions carry a scale. The bank fell to 3,298 from 3,323: we wrote 40 questions and 57 expired. What’s nextOne open item, and it is the same one. The content audit still owes the single collection that has to be written rather than read. The answer-slot figure needs restating now that the scales exist. Across the 2,797 questions with no scale the right answer sits at A 25.0%, B 26.7%, C 24.9% and D 23.4%. The 509 scaled ones are even by construction, so the old bank-wide figure no longer describes what a player meets.
Read & Rank In Development
build
Where it standsA blind taste test for candidates: the question, not the race, is the unit of comparison. Two commits, both dependency work, and no feature work for a fourth cycle. What’s nextThe finisher count finally moved. 14 people revealed a ballot in the quarter, against the 10 that stood for two pulls, on 18 race starts rather than 19. More people finished and fewer started. At this volume that is four people and no conclusion, but it is the first movement in three pulls. It is still gated on a stance pool that shrank again.
Fallacy Finders In Development
build
Where it standsv1.1 is 4 of 6 phases done and built the Fallacy Sleuth: curation-time AI proposing timecoded fallacy tags, blind to the speaker. Quiet for a tenth cycle: no commits at all, and none since August 12. It is still the one repository here with no remote, so its state can only be read from the machine this report is written on. Giving it one was called the smallest open task on this list four editions ago. What’s nextIts dependency is barely busier than it is. On the Record shipped 1 commit and ingested 2 meetings. Clip sourcing at volume still depends on transcripts the quality gate will accept.

Connect Pillar: Shared Values

Connected accounts, one human one voice. Parked during the election push; resumes after November.

The parked pillar logged 1 commit, which is what parked looks like. The signed-out arrival it shipped on October 27 stands: anyone may look at any community without an account. We count these commits here rather than in the headline, because a parked pillar with a busy week reads exactly like a reactivated one in a total.

FeatureStatusJourneyWhere it standsWhat's next
Civic Spaces Live
alpha
Where it standsGeo-scoped communities, neighborhood to federal. 1 commit this cycle. A reader with no account can now look anywhere, which is what the rest of the platform already allows. Explore shipped per slice, the post composer got a dark mode it had never had, and a place picker that swallowed clicks on its middle row was fixed. The production walk is written down now, so a deploy can be verified rather than assumed. What’s nextIt stays parked until after November. The groundwork laid early is the thing v4.0 will be built on.
Empowered Quests Live
alpha
Where it standsFact-verification quests with consensus scoring; verified facts already feed Essentials. Red gem quests unlock at VR 90+. One commit, a dependency wave. What’s nextv1.6 planning.
Focused Communities Paused
build
Where it standsv1.0 shipped April 17: per-topic forums wired to the Compass. Three commits, and one of them retires a whole service. The frontend now calls the shared engine, and the standalone backend is frozen. What’s nextv1.1: badge ratification and argument mapping.
Empowered Listening Paused
build
Where it standsStructured debate infrastructure; 5 of 6 phases done, with timers, transcription and audience notes. Two commits, both housekeeping: a security pin and a dependency wave. What’s nextPhase 6: voting and badges. Then Common Grounds can build on it.
Emparks Design
design
Where it standsPlatform-wide annotation system, five mark types. Full design doc; no build yet. What’s nextAwaits Connect-pillar capacity.
Common Grounds Design
design
Where it standsAdversarial-collaboration mediation. Full design doc; depends on Empowered Listening and Fallacy Finders arbiters. What’s nextBuilds after its dependencies land.

Platform Core & Infrastructure

The shared spine every app plugs into.

ev-accounts

Auth, tiers, XP and gems, politician database, TIGER geofencing, elections. 8 commits, against 72 last cycle. The contradiction check reported on October 27 was re-run for this report and still names 333 groups covering 2,283 rows, unchanged. Nothing adjudicated them this cycle. 2,277 federal politicians carry finance records.

The check cannot see the defect it most needs to find, and its own documentation says so. It reports disagreement, not error. A mapping that is wrong for everyone who cites it produces no disagreement at all. It can also compare only 28% of voter-facing rows, because a numeric bill needs a legislature to scope it and a descriptive phrase is not an instrument. A clean run would mean that no two rows disagree, never that the mappings are right.

One migration was reverted the next day because its premise did not hold. Four real display names had been replaced with pseudonyms, and the replacement was undone. No version number is printed on this card, because nothing in the repository produces one.

On the Record

Speaker-attributed, searchable transcripts of public meetings. The burst stopped as quickly as it started. The totals read 195 meetings, 22,336 transcript segments and 1,494 identified speakers, including LA City Council with 7,200 recorded votes. Four days added 2 meetings, 14 segments and 6 speakers, against 11 meetings in the three days before. The single commit is a dependency wave. The standing caution is repeated rather than retired: segments carry no creation timestamp, so dated meetings are the evidence, and 26 of them landed in the last 30 days. The two drops earlier this year are still unexplained. Built by volunteer Chris Andrews.

Data Coverage Deep Dive

The invisible work. This is what "evidence-based" costs, and we paid it.

Sourced stances, by where they serve

Two rows moved for a reason that has nothing to do with research. The national row falls 556 to 552 and Texas rises 270 to 272, for the reason in the lead. Eleven of the sixteen politician counts are unchanged. That is the check this table exists for. When the named jurisdictions hold to the row, the order that assigns a politician to a place has not shifted under us. The table partitions stored rows, not served positions: 39,042 rows across 4,311 politicians and 64 topics. Only 31,626 resolve to a position a voter is served, held by 4,289 politicians, with 2,197 standing withdrawn. 2,722 sources were re-checked in a sweep that finished in April 2026, and that number has not moved since, because every row in it was written then.

Coverage areaPoliticiansStances
US Congress + national5529,695
2026 candidates & discovered names (no seat)9296,218
California (state + local)3765,136
Massachusetts (state + local)3024,027
Texas (state + local)2722,544
Maryland (state + local)1882,412
Maine (state + local)1861,796
Virginia (state + local)1661,025
Oregon (state + local + county)154826
Utah (state + city + county)63688
Wisconsin (state + county + local)208629
Washington (state + county + city)168502
Arizona (state + local)129438
North Carolina (state + local)132415
Tennessee (state + local)119237
38 more jurisdictions (incl. Georgia, Kentucky and DC)3672,454

Both totals reconcile against the headline counts of 4,311 politicians and 39,042 stored rows. Note what this table cannot show you. It counts stored rows, so a withdrawal adds a row here rather than removing one. The reconciliation is also what makes a reorder of this table safe to publish. We moved the test that files someone as national below the two tests that find them a state, and the total still matches the headline to the row, with eleven named rows unmoved. A partition that still sums correctly after a change to its order is a partition whose order we have understood.

Nothing was added to the certified corpus, for a fifth cycle running. One stratum of one evidence class is still the only measured part of it: 57 state-level, record-based stances, certified October 3 at 0.830 and 0.932 with no severe disagreements. It was certified against codebook version 0.4, and the codebook has stood at 0.6 since October 15. A certification is a statement about a codebook as much as about a corpus. News-sourced and statement-sourced evidence are not certified, and neither is the local tier.

Budgets you can follow

Treasury Tracker’s database holds budget data for 11,188 government entities across all 50 states, spanning fiscal years 1976 to 2028. Every figure in this card moved, for the first time in three editions.

  • 3,103 cities, 2,787 townships, 1,828 towns, 949 boroughs, 923 counties, 814 villages, 50 states, and 734 others.
  • 10.4 million line items, 1.3 million transactions and 282,457 public salary records.
  • 583,565 budget rows, and 207,494 of them were rewritten this cycle. 376 are new.
  • 38,486 budgets still carry an unknown provenance grade, down from 245,604. The pile is spread across 1,177 of the 11,188 governments, where four days ago it was 3,420.
  • Graded, the rest reads 524,172 self-reported unaudited, 19,590 compiled from audited figures, 1,289 audited to government accounting standards, 28 to another recognised basis. The five grades sum to the 583,565 above. That is the check that makes this a partition and not a sample.
  • The class list above is a partition, not a sample. Our refresh script complains if the seven named classes and the catch-all stop summing to the headline.

The provenance debt, paid

A provenance grade is the sentence that says where a budget figure came from and who, if anyone, checked it. Ours run from an audited statement prepared under government accounting standards, through a compilation drawn from audited figures, down to a government’s own unaudited filing. That last one is what a small town’s budget usually is, and it is an honest answer. “Unknown” is not a grade on that scale. It means nobody recorded the answer.

Three editions reported this debt unpaid and promised to report the week it was paid. This is that week. 207,185 budgets were graded in a single run on October 28, most of it inside one hour. Every one of the 207,494 rows the run touched also gained a source address and a stated basis, so they are sourced rather than relabelled. 381 of them turned out to be audited to government accounting standards.

What is not fixed is the loader that built the pile. While we were writing this edition it created 67 more ungraded budgets, across three Hawaii counties. That is why the figure we print is 38,486 rather than the 38,419 the grading run left behind. Clearing a backlog is not the same as closing the hole it came out of, and the second job is the one still open.

Follow the money

The third leg, alongside votes and stances. The platform holds 29.2 million individual campaign-finance contributions totaling $8.6 billion, covering 2,277 federal politicians and candidates across 24 election cycles (1980 to 2026). Every record traces to its original filing, and superseded lines from amended filings are retired individually rather than counted twice. Nothing here moved this cycle. Only confirmed links are counted as money, and the disputed match reported on October 27 is still disputed rather than deleted.

Reach & Engagement

Who actually uses what we build — measured, not guessed.

More people arrived, and this time they read more too

894 people used the platform in the last 90 days, across every app, over 1,569 sessions and 3,907 page views. Sessions average five minutes thirty and 63% do not bounce. 137 of the 428 who reached the main site clicked through into an app, which is 32% in the quarter.

The divergence we called resumed two days ago has reversed. Visitors rose 845 to 894 and page views rose 3,746 to 3,907: 49 more people reading 161 more pages. Last time the same 49 people brought 28 pages between them. Pages per visitor is flat at 4.4, and the average session is thirteen seconds longer. Two days is not a trend, and we have now called this one wrong in both directions. Over twelve pulls the main site reads 316, 312, 297, 300, 314, 337, 340, 344, 369, 378, 398, 428.

Does the tool get used as a tool?

Everyone who arrived, against the people who did the one thing that app exists for. Bars are the 90-day rate; counts are people.

30d90d
Compass — completed a calibration (of 170)2148
28%
Essentials — looked up an official (of 196)3552
27%
Read & Rank — revealed a ballot (of 73)1416
22%
Civic Trivia — finished a game (of 107)412
11%
Treasury — anything past the front page (of 271 — see below)2231
n/a

Essentials leads the Compass for a fourth pull, 52 people to 48. Across ten pulls Essentials reads 29, 29, 29, 31, 33, 35, 42, 44, 48, 52. The Compass is still much the deepest funnel: 147 of its 170 visitors opened the quiz, which is 86%, and 48 finished it. Every row gained or held except Civic Trivia, which held at 12 finishers on falling arrivals.

How many arrive

Unique visitors per app, 30 days and the full quarter. Each app is counted once across every hostname that serves it — Treasury drew traffic on four this quarter, the main site on three, and Essentials, Compass and Civic Trivia on two each.

30d90d
main site196428
treasury165271
essentials125196
compass78170
civic trivia32107
readrank7373

Five rows gained and one lost ground, for a second cycle. The main site rose furthest, 398 to 428. Treasury rose for a seventh cycle, 250 to 271, Essentials 187 to 196, and Compass 160 to 170. Civic Trivia fell again, 108 to 107. Civic Spaces is absent from this table, and that is a measurement gap rather than a result. It serves no traffic our analytics can see.

Treasury is still unmeasured rather than low, and we had the reason wrong

An earlier edition published a rate here and withdrew it. It stays withdrawn. The events were sound; the denominator was not, because Treasury pushes its in-app navigation into the address bar without recording a page view, so every visitor is filed against the bare domain. This quarter those few people produced 170 category drills, 97 line-item views, 96 entity selections, 61 year changes, 31 chart changes and 18 searches.

271 people reached a Treasury address this quarter and 31 fired any past-the-front-page event. 130 of them arrived on financials.empowered.vote, which rewrites its own address to our own budget on load, so arriving there is enough to be counted. 160 arrived on the explorer hosts. We measured that split by hostname rather than by the budget opened, so it is not comparable with the two-to-one ratio printed before. It is still not a publishable rate and we are still not publishing one.

Corrected after publication: this list first went out without the searches. We checked which events Treasury sends, saw no search event, and took that for proof it had stopped existing. The list we read covers 30 days, and the last search was September 22. A quiet event looks identical to a dead one. The 31 above is unaffected: everyone who searched also did something else.

The Road Ahead

How much further we have to go, honestly.

Now to Nov

Election readiness

Stop the ungraded budgets arriving. Grading the backlog was the first item here for three editions and it is done. The loader that created it is not. Then rule on the rungs that changed between seasons, and on what a draft season may serve. 87.9% of what a voter reads still comes from a closed season. Put the 8.5 million votes to work. Nothing has been certified in five cycles, and the one certified stratum is two codebook versions behind. Finish the 2026 candidate waves, and give Fallacy Finders a remote. Success in November is one thing: any voter in a served area can look up every name on their ballot and follow the money behind it.

Post-election

The Connect pivot

Reactivate the parked pillar: Empowered Listening phase 6, Focused Communities v1.1, Civic Spaces v4.0 — whose shell, shared theme, banner library, dashboard, Explore view and now a signed-out arrival are groundwork done early. This is where the platform stops being a reference tool and becomes a community, and where our sustaining-donor model lives.

The long game

The Empower pillar

Designed but not built: the Gems stipend economy, Awareness Exchange, Badges, Symposium tournaments, Emparks, Common Grounds, and tools for citizens to run for office on the strength of their solutions. Roughly a third of the founding vision, waiting on the first two pillars.

The People Doing This

A volunteer-built platform, visible in the commit history.

Chris Cantrellfounder, systems design, data pipelines
Chris AndrewsOn the Record, Essentials campaign finance
Krishna PatelUX design; the Civic Spaces shell
Radhika JoglekarUX & logo design
Aditi DeshpandeUX design
Peter Torrestreasurer
Ryan Rinkelsecretary
Debra Maniscalcooutreach

The purpose, mapped

An interactive Purpose mind map walks the whole vision — the thesis, and the Why, How, Who and What — with a short explainer on every concept and pillar feature.