Empowered Vote · Volunteer Briefing · October 20, 2026
Every number on this page comes from the live database, verified git history, or our product analytics, not from estimates. We build tools for shared facts; our progress report should be one. Each briefing is written fresh and kept at its own dated address, so you can read what we said before and check it against what we say now.
This is a seven-day cycle, and Georgia is the shape of it. No edition published on Friday, so this one covers a week rather than the usual three or four days. 11,167 governments now have budget data, up 2,246 — and every one of those 2,246 is a government that had none before. Georgia was swept end to end: 440 cities, 152 counties and 103 towns, which is the state complete. New York was finished: 933 towns, 561 villages and 57 counties, on top of the school districts loaded last cycle. Line items went from 5.3 million to 10.4 million, a doubling for the second cycle in a row.
And not one of those 207,113 budgets carries a provenance grade. Every single one is filed as unknown. Last cycle this page praised the opposite: the 717 school districts all landed graded, and the pile of budgets whose provenance is simply unknown sat unmoved at 38,491. That pile is now 245,604. The old 38,491 did not move — the growth is this week’s load, to the row, and the school districts still carry their grade. We took on five times more ungraded provenance in a week than the platform had accumulated in its entire history. The governments are real and the figures are real; what is missing is the line saying where each number came from and whether anyone audited it. That line is most of what separates this platform from a spreadsheet, and this week it is what got dropped to move fast.
The stance line restarted, and it is smaller than it sounds. Last edition reported that nothing had been written to the stance table since October 9 and called restarting it the one job that mattered. 72 positions landed — at 06:01 this morning, in a single batch, covering 15 Georgia officials across 11 topics. The eleven days before it were empty. It is the right state and the right work, and it is one batch rather than a week of research — written, as it happens, in the same minute this page began gathering its numbers. Eleven days of nothing, ended an hour before publication, is not yet a restart.
Each update is a dated snapshot. We do not edit one after publishing it.
This page is the current briefing, and it always lives at empowered.vote/briefing. Every edition also keeps a permanent address of its own, so a number you read here can be checked against the date it was published on. There is a picker at the top of this page that jumps to any of them. Each update is written from scratch: work that runs for a month stays the headline, but the sentences around it are rewritten every time. We never edit a published edition. When an earlier one got something wrong, the next one says what it corrects.
No claim in the October 13 edition was found to be wrong, and that is a result rather than an absence. Five of its figures were re-derived from scratch this week and every one reproduced exactly: the 717 school districts still carry a real provenance grade across all 82,564 of their budgets; the unknown-grade pile it printed at 38,491 is still 38,491; the last stance row before this morning was indeed October 9; Season 3 is still a draft holding exactly 17 answers; and 70 of Puerto Rico’s 78 alcaldes still have no portrait. Re-deriving a claim is the only way to find out whether it was wrong, so this page says plainly when the answer is that it was right.
It also set a question, and this edition answers it. It said the remaining slot-C bias in freshly generated trivia questions could only be separated into “working as designed” and “still broken” by the trivia repository’s own tooling rather than a database query, and that doing it was the next job. It was done, and the generator is not broken. The detail, including what is still wrong, is in the Inform table below.
One figure is withdrawn, though, and it is one this page has now printed twice. The editions of October 10 and October 13 both said that “24 rungs” ask a genuinely different question in Season 2 than in Season 1. Nothing in the repository produces that number, and this edition could not derive it. Two related figures are documented and are not it — 29 of 60 topics whose rungs differ from the frozen table, and 13 whose rung text differs from what voters are shown, both measured in late September. A number this page cannot reproduce does not get carried forward just because it was here last time, which is the same rule that retired the ev-accounts version label last edition.
Open to everyone, no login. Evidence-based, engaging, anti-partisan.
| Feature | Status | Journey | Where it stands | What's next |
|---|---|---|---|---|
| Civic Trivia Championships | Live | alpha |
Where it standsThe slot-C question is settled: the generator works, and what is left is a property of the questions rather than a defect in the code. Of the 459 questions generated since the fix landed, 248 offer four free-form options, and those land A 63, B 60, C 60, D 65 — 24.2% in slot C, even to within a point of chance. The other 211 offer four comparable numbers and are deliberately sorted by value rather than shuffled, so the slot is the value’s rank; they sit at 54.5% slot C, and the sort held for 195 of the 211. The whole residual skew lives in the sorted half. The bank a player actually meets is even: across 3,322 live questions in 43 active collections, A 25.0%, B 25.8%, C 25.0%, D 24.2%. 12 commits; question generation moved to a newer model mid-cycle. | What’s nextIt is not a code defect and it is still a tell. The reason sorted questions pile into C is measurable: on a numeric question the true value is one of the two middle options 188 times out of 211, 89.1%, where chance is 50%. The models writing these questions bracket the right answer, offering values above and below it, so sorting puts it in the middle. A player who learned “pick a middle number” would beat chance on that half of the bank — which is the thing the original fix existed to prevent. The open question is no longer whether the generator works but whether numeric options should be sorted at all. The repository’s own placement checks pass 47 of 47. The content audit still owes one collection that has to be written rather than read. |
| Empowered Compass | Live | alpha |
Where it standsEvery topic now keeps a journal that outlives the flag that used to carry it. A per-topic journal shipped this cycle — spec, plan and first phase — so a question’s history of revisions is a record in its own right rather than a note attached to whichever flag happened to raise it. The library holds 64 topics and the open season serves 60, both unchanged. 5 commits, including a lens repair: the Education Lens fallback had been dropping gender identity and school budget from the questions it offered. | What’s nextSeason 3 is being assembled in the open, and none of it has reached a voter. Checked again, because a season opening is a launch that no other figure announces: Season 2 open, Season 1 closed, Season 3 still a draft holding exactly the same 17 answers it has held for two cycles. Behind it this week sat the real work — codebook versions 0.5 and 0.6, a school-police annex covering per-state duties, two single-topic draft ladders put up for review, and two decision records that park the education and school-budget ladders rather than dropping them. One figure the last two editions carried is withdrawn here: that “24 rungs” ask a genuinely different question in Season 2 than in Season 1. Nothing in the repository reproduces that number. What is documented is a pair of older measurements, both from late September — 29 of 60 topics whose rungs differ from the frozen table, and 13 whose rung text differs from what voters are actually shown. The rule they serve is unchanged and is the part that matters: a rung that changed cannot carry an answer across without a ruling. |
| Treasury Tracker | Live | alpha |
Where it standsThe biggest week this app has had, and the one with the worst paperwork. 16 commits took Georgia from three seeded governments to the entire state — 159 counties and 553 municipalities swept, consolidated city-counties handled as a kind of government the current census vintage cannot name, and seven apparent disagreements between the census roster and the state’s own list run down to one finding: all seven had been dissolved. New York’s towns, villages and counties completed in the same window, and the school districts were reconciled against their own audited statements. 2,246 governments gained budget data and all 207,113 of their budgets are filed with an unknown provenance grade. | What’s nextGrade the Georgia and New York load. It is a week’s worth of rows and it is the difference between holding a number and standing behind it. The oldest debt on this page is now five editions old and still unpaid: nobody has timed the real page, five runs before and after, so four cycles of performance work remain unproven — and the database has doubled twice since that work was done, which makes the measurement more urgent, not less. A faster call is not a faster page until the page is timed. The comparison gate still has not been asked whether a school district may be set beside a city. |
| Empowered Essentials | Live Alpha | alpha |
Where it standsZIP-to-representatives with funding, votes and stances. 7 commits, and the Georgia theme reaches here too: Atlanta, Fulton County and DeKalb County were seeded with 41 offices and 34 headshots, and the app took new city and state banner art to match. A profile page now groups a politician’s stances by issue category instead of listing them flat, and the Alpha Communities landing was redesigned as a state rail beside a detail panel. 12,435 curated officials and candidates are served, beside a 77,733-record discovery pool still in triage. | What’s nextStill owed, and unchanged for a fifth cycle: portraits for the 70 Puerto Rico alcaldes without one. Re-derived this week and exact — 78 municipios carry a mayor, 70 of them have no default image, and the island’s other 83 seats all do. The traffic the Compass was going to send has arrived: people looking up an official went from 35 to 42 in the quarter, the largest rise of the five per-app actions measured below. |
| 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 shared-library version bumps — no feature work for a second cycle, after the redesign landed. | What’s nextLast edition said the next pull would be the first that could deliver a verdict on the redesign. It cannot. 10 people revealed a ballot in the quarter against 20 race starts, where it was 9 on 20 — one more finisher, no more starters. Two cycles running, this figure has moved by exactly one person. At this volume that is noise, and we will not read a redesign’s success or failure out of it. It remains a start problem, and it is still gated on a stance pool that gained 72 positions in a single state this week. |
| 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, with names stripped before the model sees a word and every clip re-run with the speakers swapped. Quiet for an eighth cycle: no commits at all. It is still the one repository on this page with no remote — its state can only be read from the machine this report is written on, while every other repository here is read from a published branch. Two editions ago this page called giving it one “the smallest open task on this list”. It still has not been done. | What’s nextIts dependency moved only by maintenance: On the Record shipped two dependency updates and no code, though two more meetings were ingested. Clip sourcing at volume still depends on transcripts the quality gate will accept. |
Connected accounts, one human one voice. Parked during the election push; resumes after November.
This pillar is parked, and it logged 47 commits this cycle. Read that carefully before taking it for a restart. Eleven of them are automated maintenance — every commit in Empowered Listening, Focused Communities and Empowered Quests is a dependency bump, a build pipeline or a lint pass, raised and merged by tooling, with no human deciding to build anything. The other 36 are Civic Spaces, and that is real work: brand marks, a reply composer that works in dark mode, a discussion view that keeps its topic title in sight, and a contributor onboarding guide. It is polish and groundwork on a shell that is already built, not the v4.0 the pivot is waiting for. A parked pillar with a busy dependency robot is indistinguishable from a reactivated one in a commit count, and only reading the commits tells them apart — which is why this page counts them here rather than in the headline.
| Feature | Status | Journey | Where it stands | What's next |
|---|---|---|---|---|
| Civic Spaces | Live | alpha |
Where it standsGeo-scoped communities, neighborhood to federal. 36 commits, all of them finishing work already begun: the Inform tools now show their real wordmarks in the rail, the Civic Spaces mark was recoloured to match, fact chips became compact with their detail on hover and tap, and a thread keeps its topic title in view while you read the replies. A volunteer contribution guide was written and a stale onboarding document retired — the kind of thing that only gets done while a pillar is quiet. | 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+. Three commits, all dependency bumps. | What’s nextv1.6 planning. |
| Focused Communities | Paused | build |
Where it standsv1.0 shipped April 17: per-topic forums wired to the Compass. Four commits, all dependency bumps. Waiting on the post-election Connect pivot. | What’s nextv1.1: badge ratification and argument mapping. |
| Empowered Listening | Paused | build |
Where it standsStructured debate infrastructure; 5 of 6 phases done — timers, transcription, audience notes. Four commits: dependency bumps, a build pipeline, and a first pass at its lint warnings. | 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. |
The shared spine every app plugs into.
Auth, tiers, XP and gems, politician database, TIGER geofencing, elections. 48 commits, the busiest repository this cycle, and the through-line is the 2026 ballot. The California general ballot was finished — 110 people created and 11 more matched to records we already held, found by deduplication rather than by hand, and one of those matches was a person whose evidence had been sitting in our own database all along. A headshot sweep went after a failure mode worth naming: portraits that are not portraits, images with no face in them, alongside a ballot-photo column nothing had ever swept and candidate-to-politician links that had never been made. Atlanta, Fulton and DeKalb were seeded with 41 offices. 2,265 federal politicians carry finance records.
No version number is printed here. Earlier editions carried one; nothing in the repository produces that string, and a label this page cannot derive does not belong on it.
Speaker-attributed, searchable transcripts of public meetings, synced to the original video. Two commits, both dependency updates, and the dataset grew anyway — the normal shape for an ingest pipeline between code changes. It reads 182 meetings, 21,776 transcript segments and 1,448 identified speakers, including LA City Council with 7,200 recorded votes: two new meetings, 21 new segments and 6 new speakers this week. The standing caution is repeated rather than retired: segments carry no creation timestamp, so dated meetings are the evidence here — 18 of them in the last 30 days — and the two drops earlier this year are still unexplained. Built by volunteer Chris Andrews.
The invisible work. This is what "evidence-based" costs, and we paid it.
Fifteen of the sixteen rows below are identical to last week, and the whole cycle’s movement is in the sixteenth. That is the strongest form of this check: when every named jurisdiction reproduces to the row, the resolution order that assigns a politician to a place has not shifted under us. The table partitions stored rows rather than served positions, so it reconciles against the row count: 38,383 rows across 4,243 politicians and 64 topics; 1,864 of the 2026 candidates have some. 32,072 of them resolve to a position a voter is served, held by 4,223 politicians at 7.6 topics each, with 1,629 standing withdrawn — a withdrawal count that has not moved since the audit stopped. One figure keeps its date said out loud: 2,722 sources were independently re-checked in a sweep that finished in April 2026, and the number has not moved since, because every row in it was written then. It is a dated achievement, not a running total.
| Coverage area | Politicians | Stances |
|---|---|---|
| US Congress + national | 556 | 9,694 |
| 2026 candidates & discovered names (no seat) | 931 | 6,206 |
| California (state + local) | 373 | 5,048 |
| Massachusetts (state + local) | 302 | 3,887 |
| Texas (state + local) | 270 | 2,412 |
| Maryland (state + local) | 188 | 2,316 |
| Maine (state + local) | 186 | 1,796 |
| Virginia (state + local) | 166 | 1,009 |
| Oregon (state + local + county) | 154 | 826 |
| Utah (state + city + county) | 57 | 682 |
| Wisconsin (state + county + local) | 208 | 629 |
| Washington (state + county + city) | 168 | 502 |
| Arizona (state + local) | 129 | 438 |
| North Carolina (state + local) | 128 | 411 |
| Tennessee (state + local) | 119 | 231 |
| 38 more jurisdictions (incl. Georgia, Florida and DC) | 308 | 2,296 |
The remainder row absorbed all of it: 293 politicians and 2,224 stances last week, 308 and 2,296 now. That is exactly the 15 officials and 72 positions written this morning, and it says where they went — Georgia, which at 21 researched officials is still too small to earn a row of its own. Both totals reconcile against the headline counts of 4,243 politicians and 38,383 stances.
Nothing was added to the certified corpus, for a third cycle running. One stratum of one evidence class remains 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. News-sourced and statement-sourced evidence are not certified, and neither is the local tier.
Treasury Tracker’s database holds budget data for 11,167 government entities across all 50 states plus DC, spanning fiscal years 1976 to 2028:
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 — which is what a small town’s budget usually is, and is an honest answer. “Unknown” is not a grade on that scale. It means nobody recorded the answer. Someone reading one of these 2,246 new governments sees real figures with nothing behind them, and has no way to tell a town that published audited accounts from one that typed a number into a form.
Why it happened is not a mystery, and it is worth saying plainly: grading is per-source work, and the sweep was built for speed. The 717 school districts last cycle could be graded because they came from one publisher with one known practice — and this week they were reconciled against their own audited statements, which is the next step past a grade. Georgia’s municipalities and New York’s towns and villages arrive through forms whose provenance differs by the generation of the form, and one of this cycle’s commits exists precisely to read an older form generation. The work is identified, not done, and this page will report it the week it is. Until then the honest framing is that we hold these budgets and cannot yet tell you how good they are.
The third leg, alongside votes and stances. The platform holds 29.2 million individual campaign-finance contributions totaling $8.6 billion, covering 2,265 federal politicians and candidates across 24 election cycles (1980 to 2026). Every record traces to its original FEC filing, and superseded lines from amended filings are retired individually rather than counted twice. This cycle’s work was again subtraction: the California adjudication ratchet was banked at 437 confirmed independent-expenditure sources, down from 520, and the mechanism that had been quietly re-absorbing rows an editor already ruled on was closed. A money page that credits a contribution to the wrong person is worse than one that is missing it, which is why a figure going down counts as progress here.
Every state has at least one researched official, and all 50 congressional delegations are covered nationally. Shading shows the depth of state and local research. Exactly one tile changed this cycle, and it is the only one that could have: Georgia moved from Seeded to Growing, 6 researched officials to 21. That is this morning’s batch and nothing else — a tile shades only when an answer is published, and every answer published this week went to Georgia. Every other tile is where it was last week. Puerto Rico is still an outlined tile: 161 seats seeded, no stances yet.
Hover any state for its exact count. The seeded tier is mostly governors and statewide officers; deeper tiers add legislatures, counties and cities.
Who actually uses what we build — measured, not guessed.
773 people used the platform in the last 90 days, across every app, over 1,510 sessions and 3,636 page views. Sessions average five minutes fourteen and 66% do not bounce. 120 of the 369 who reached the main site clicked through into an app — 33% in the quarter. That share had read 31% for five pulls running, and the count itself had been pinned at exactly 105 for the last three. Both moved this cycle, which after three identical readings is worth noting even though fifteen people is a small number to draw anything from.
The same divergence as last cycle, and wider. Visitors went 724 to 773, while sessions fell 1,554 to 1,510 and views 3,741 to 3,636, with the average visit eighteen seconds shorter again. More people arriving, each doing slightly less once they do. Read every figure here as an order of magnitude, not a trend line — at this volume a couple of people either side is the whole swing, and the window slides. Over ten pulls the main site reads 306, 316, 312, 297, 300, 314, 337, 340, 344, 369. The hostname map was re-run first, as it is every cycle, and found no new host.
Everyone who arrived, against the people who did the one thing that app exists for. Bars are the 90-day rate; counts are people.
Essentials has passed the Compass on this measure for the first time in the seven editions this archive holds. 42 people looked up an official, against 38 who finished a calibration; in every previous edition the Compass led. Essentials has climbed 29, 29, 29, 31, 33, 35, 42 across those pulls, and this is its largest single step — consistent with compare having moved here from the Compass, though two pulls is not proof of a cause. The Compass remains the deepest funnel: 125 of its 145 visitors opened the quiz and 38 finished it. Civic Trivia went from 13 finishers to 14, on 21 starts rather than 20.
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 and both of Compass and Civic Trivia on two.
Treasury Tracker rose furthest for a fourth cycle running — 206 to 224 in the quarter and 110 to 133 in the month, which puts it within nineteen people of the main site over thirty days. Every app gained. Civic Trivia is the slowest, 112 to 116, and its 30-day figure recovered from 19 to 25. 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.
An earlier edition published a rate here — seven people in a quarter, as 5% of 152 visitors — 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. The direction of travel stays wrong too: this quarter those few people produced 173 category drills, 103 line-item views, 99 entity selections, 63 year changes, 27 chart changes and 18 searches — all but unchanged from last cycle, while visitors rose.
95 people reached a budget address and 28 fired any past-the-front-page event. The 95 splits the way it always has: 65 landed on our own organisation’s budget, one click from the donate page, which selects nothing and so fires nothing. Asked directly, 33 people opened some other government’s budget; asked by subtraction the same window gives 30, and the gap of three is people who looked at both. The ratio is narrowing for the first time: the donate-driven half went 60 to 65 while the other half went 25 to 33, so our own budget is now twice the rest rather than two and a half times. And the shape of that half changed. It used to be our own onboarding checks — Austin, Travis County, Los Angeles, Colorado Springs. This quarter it spreads across more than twenty governments nobody here seeded for testing, including Duvall, Bloomington, Durham, Ashfield and Seattle. It is still not a publishable rate and we are still not publishing one. It is the first cycle where the trend points at one.
How much further we have to go, honestly.
Grade the budgets, and keep the stance work going past one morning. Those are this cycle’s two debts and they are the same debt in different data: a figure we hold but cannot yet stand behind, and a figure we stand behind but have barely added to. With two weeks to the election, the 207,113 ungraded budgets are the larger risk, because they are already in front of readers. The audit’s tail bands are untouched — the withdrawal count has not moved — and the rungs that changed between seasons still need a ruling before any answer carries across. Then put the 8.4 million votes to work — that record feeds exactly the evidence class the one certified stratum rests on, and nothing has been certified in three cycles. Finish the 2026 candidate waves, decide whether numeric trivia options should be sorted at all, write the one collection an audit cannot fix, and time the Treasury page properly. 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.
Reactivate the parked pillar: Empowered Listening phase 6, Focused Communities v1.1, Civic Spaces v4.0 — whose shell, shared theme, banner library, dashboard and now its contributor guide 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.
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.
A volunteer-built platform, visible in the commit history.
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.