Empowered Vote · Volunteer Briefing · August 21, 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. Six apps are live or in alpha, two are in active development, and the deep data work is far ahead of what most of us realized.
Since June 7, nearly all effort has gone into the Inform pillar, on purpose. The midterms are under three months out, and the work that moved is what a voter touches in an election year: 2026 House candidate coverage in every state, audited budget pipelines in all fifty, campaign-finance records behind every federal candidate, city-by-city official seeding, and a public-meetings transcript system. The Connect pillar is parked until after November — sequencing, not abandonment.
Colorado Springs, end to end in one cycle. What took months for the first city now runs as a sequence: boundaries, seats, portraits, ballot, budgets, stances. All 100 members of the Colorado General Assembly are seated, alongside Colorado Springs’ ten-member council and El Paso County’s eleven — 126 Colorado seats, 125 of them with a portrait — and both the city and the county are loaded in Treasury Tracker. Stance research followed in the same week: 77 sourced positions across 28 people, every one of the 61 quotes verified against the raw bytes of its source, and 7 of the 35 people researched producing a documented, honest zero. Colorado moves off the palest tier of the map below. One caveat kept in view: Colorado’s legislature exposes essentially only the current session — a 13-year member returns 26 bills, 25 of them from 2026 — so the legislative half is worth re-running later.
What is on the ballot is a factual claim. Seeding the ballot a Colorado Springs voter actually receives — 17 races, 56 candidates, from the Secretary of State’s own list — exposed that our existing federal races carried only the major-party field: 16 U.S. House candidates where the state lists 38, every missing one minor-party or unaffiliated. A two-name race printed for a five-name ballot is a false statement about that ballot. Washington’s was not wrong so much as unfinished: the 2026 primary field had been seeded from a pre-certification feed and marked provisional, and the certified canvass has now closed that gate — 381 rows across all 147 races — 262 advanced, 119 not nominated, 5 withdrawn. Eight rows are deliberately still open: LD 42’s second and third place are four votes apart (13,125 to 13,121), inside the mandatory hand-recount margin, so the race stays unresolved until Whatcom County recounts. The check that caught it measures the margin; a guard that only looks for an exact tie passes that race green.
Open to everyone, no login. Evidence-based, engaging, anti-partisan.
| Feature | Status | Journey | Where it stands | What's next |
|---|---|---|---|---|
| Empowered Essentials | Live Alpha | alpha |
ZIP-to-representatives with funding, votes, stances. Colorado Springs and El Paso County are the newest served communities, certified with their own banners, on the heels of Puerto Rico (161 seats including all 78 municipios) and Austin and Travis County (23 seats); Washington came in before them, as Seattle, King County, Bainbridge Island and Kitsap. Counties are searchable by name and marked as such, so Los Angeles County can no longer be mistaken for the city. The habit that outlasts any one community: where research genuinely stopped, the page says so — 39 Collin County council members marked not-researched rather than left blank. | More served alpha communities; portraits for the 70 Puerto Rico alcaldes still without one, and the territory’s first stance research; extend the has-context chips beyond Collin County and the Washington entries. |
| Empowered Compass | Live | alpha |
44 topics, 220 authored stances, spectrum + invert mechanic, side-by-side politician comparison. 1,795 calibration changes tracked. Three lenses (general, Judicial, Educators) with an icon switcher on profile pages. No app changes this cycle; the work is in the ladders, which now run 28 topics at state scale and 22 at local. Two rulings from the Washington sweep are still open — a pair of chairs that forbid a combination a real enacted law delivers, and how far “structure over inferred purpose” reaches in criminal justice — logged rather than patched around. | Quick-compare polish; inferred-stance overlays as research matures. Getting the tool seen matters more than either — see Reach & Engagement. |
| Treasury Tracker | Live | alpha |
The sleeper giant. See the coverage panel below; the audited state-budget pipeline now covers all 50 states and reaches back to FY1976. | Three milestones shipped this cycle: Austin + Travis County, the Los Angeles repair described below, and Colorado Springs + El Paso County at 64 rows. The Colorado pass closed six defects that all tied at exactly $0 while wrong — one of them a shared loader hardcoding Texas’s fiscal calendar. No dollar figure moves, so nothing downstream notices: the same blind spot as the label defect Washington found, where a rule drawn in a PDF margin became part of a line item’s name. Six Washington years remain deliberately unloaded, every one a source-document refusal rather than a parser failure. WA-CITIES-01 continues across the six largest Washington cities; Tacoma (38 rows) and Spokane (40) are in at zero residues. |
| Read & Rank | In Development | build |
Rebuilt in July: mobile-first evaluate/rank surfaces, a race hub with relevance bands on real platform race data, a ballot-first election picker and a redesigned reveal & results. The LA Mayor questions were then repointed from the June primary to the November general, and the question — not the race — became the unit of comparison in the player path. | Candidate quote sourcing for 2026 races — now gated on the re-research worklist, since it draws from the same sourced-stance pool the audit is rebuilding. |
| Fallacy Finders | In Development | build |
v1.0’s core loop shipped June 18; v1.1 is 4 of 6 phases done and built the Fallacy Sleuth — curation-time AI that reads a meeting transcript and proposes timecoded fallacy tags. Its anti-partisan guarantees are mechanical, not promised: speaker names are stripped before the model sees a word, every clip is re-run with the speakers swapped and the results diffed, and nothing publishes without two independent High-confidence human agreements. Quiet this cycle, like the rest of the app tier. | Phase 10.1 built a Curation Studio so a moderator can mark and playtest windows by hand — a 24-lane timeline with the engine’s own tolerance halos and an in-browser playtest against the real game loop. Phase 10 then finishes the US House floor clip end to end before rights-clean clip sourcing at volume via On the Record. |
| Civic Trivia Championships | Live | alpha |
34 collections (federal, 21 local, 12 state). New brand identity, a Proficiency stat, a re-themed leaderboard, and a 5-question format, with volunteer-contributed features. | Content editor roles; civic-moment event nights. |
Connected accounts, one human one voice. Parked during the election push; resumes after November.
| Feature | Status | Journey | Where it stands | What's next |
|---|---|---|---|---|
| Civic Spaces | Live | alpha |
Geo-scoped communities (neighborhood to federal) with friends, posts and role-gated moderation; v3.0 shipped in April. | v4.0 planning, post-election. |
| Empowered Quests | Live | alpha |
Fact-verification quests with consensus scoring; verified facts already feed Essentials. Red gem quests unlock at VR 90+. | v1.6 planning. |
| Focused Communities | Paused | build |
v1.0 shipped April 17: per-topic forums wired to the Compass. Waiting on the post-election Connect pivot. | v1.1: badge ratification and argument mapping. |
| Empowered Listening | Paused | build |
Structured debate infrastructure; 5 of 6 phases done (timers, transcription, audience notes), plus a June security fix. | Phase 6: voting and badges. Then Common Grounds can build on it. |
| Emparks | Design | design |
Platform-wide annotation system, five mark types. Full design doc; no build yet. | Awaits Connect-pillar capacity. |
| Common Grounds | Design | design |
Adversarial-collaboration mediation. Full design doc; depends on Empowered Listening and Fallacy Finders arbiters. | Builds after its dependencies land. |
The shared spine every app plugs into.
Auth, tiers, XP and gems, politician database, TIGER geofencing, elections. This cycle: Colorado Springs and El Paso County end to end — state-legislative and place boundaries, council and commissioner districts, 100 General Assembly members, 32 new portraits behind a liveness verifier that cleared 7 dead URLs — plus the Washington certification cull, and a candidate dedupe key that now matches on surname alone and shows every match to a human, because “Bob Chew” and “Robert Chew” would otherwise have stood as two people in one Senate race — the same failure mode, run the other way, that once produced a fabricated mayor. Standing from earlier cycles: Puerto Rico end to end, its own government and 78 municipio boundaries verified to tile the territory with no gaps, parties stored verbatim in Spanish because PNP and PPD each contain both US national parties; Austin and Travis County; ZIP-code resolution in the API; the $999 contribution cap closed in the Massachusetts finance parser; 2026 House candidate coverage in every state; the enacted 2026 district polygons for TX, FL, NC, CA and OH; every California county seeded; CI that refuses a migration deleting an answer without declaring what happens to its reasoning; and the occupancy-evidence view that answers “seated official or discovered name?” structurally rather than by provenance. Underneath it all, ADR 0002 moved officeholder occupancy onto dated office_terms, so who holds a seat is a fact with a start and end date rather than a flag. 7,478 districts mapped, 121,178 legislative votes and 19,622 bills on record.
Speaker-attributed, searchable transcripts of public meetings, synced to the original video. Live in the pipeline: 153 meetings, 17,660 transcript segments, 946 identified speakers, including LA City Council with 7,200 recorded council votes. Built by volunteer Chris Andrews; it feeds the platform's meetings system. It was the fastest-growing dataset on the platform through July and has now held flat for three cycles — no ingest run while the stance work had the floor, reported plainly because a growth claim that stops being true is exactly what this page exists to correct. The Fallacy Finders hand-off is no longer a plan: a written transcript contract between the two apps, and an offline validator that fails loudly on a bad hand-off.
The invisible work. This is what "evidence-based" costs, and we paid it.
32,887 stance positions across 3,951 politicians and 44 topics. Every stance carries reasoning and cited sources; 2,722 sources independently verified. 1,665 of the 2026 candidates already have researched stances.
| Coverage area | Politicians | Stances |
|---|---|---|
| US Congress + national | 571 | 8,931 |
| 2026 candidates & discovered names (no seat) | 946 | 5,704 |
| California (state + local) | 373 | 4,401 |
| Massachusetts (state + local) | 302 | 3,084 |
| Texas (state + local) | 270 | 1,949 |
| Maryland legislature | 184 | 1,929 |
| Maine (state + local) | 186 | 1,521 |
| Virginia (state + local) | 166 | 801 |
| Oregon (state + local + county) | 154 | 735 |
| Wisconsin (state + county + local) | 208 | 565 |
| Utah (state + city + county) | 48 | 527 |
| Washington (state + county + city) | 161 | 434 |
| Arizona (state + local) | 129 | 396 |
| 40 more jurisdictions (incl. Colorado and DC) | 253 | 1,910 |
Why several states shrank. A source-supports-claim check re-reads every cited source and asks whether it actually supports the stance we wrote. It has now swept the map — Utah fell from 1,358 stances to 531 and out of the Deep tier — and a fabricated-source retirement closed inside it: 298 stances retired, 160 cited URLs denylisted, one cluster sourced to a newspaper that does not exist. Retirement is not the end state: a re-research worklist feeds those seats back through sourcing, and Oregon gained 38 stances in a cycle when the platform total fell.
How a stance gets seated. A member is placed on a topic because of a bill he sponsored, read in its session-law text — not a sentence someone else wrote about him. Three rules came out of the Washington sweep and now travel with every wave: a unanimous bill cannot evidence a position; cite the session law, not the introduced bill, because a bill can gain its substance during passage; and where a mechanism matches no chair clause by clause the answer is blank, not the closest chair. Blanks are recorded with a named reason — 49 in Washington, 7 more in Colorado Springs. And a blank is only as strong as the search behind it: Seattle’s ten “researched blanks” rested on a single source class, and re-research seated 48 rows across 11 of 11 offices.
Two corrections to this table, both worth keeping. Per-state counts once walked a single database link and so missed everyone seeded after occupancy moved onto dated office_terms: Wisconsin read 5 researched officials where it holds 208. Reading all three jurisdiction links fixed that and over-corrected in the same motion, sweeping in names discovered in campaign-finance filings that sit on placeholder offices with no geography. Migration 1702 settles it structurally — a real seat carries geography, a placeholder carries none. The durable lesson: the sum check this table relies on could not see either error, because a misattribution moves a row between buckets without changing the total.
Treasury Tracker's database now holds budget data for 2,458 government entities across all 50 states plus DC, spanning fiscal years 1976 to 2026:
The third leg of the transparency model, alongside votes and stances. The platform now holds 28.2 million individual campaign-finance contributions totaling $7.7 billion, covering 1,302 federal politicians and candidates across 24 election cycles (1980 to 2026). Every record traces to its original FEC filing. The pipeline handles amended filings — when a committee refiles, superseded lines are retired individually instead of counted twice, which removed 18,553 duplicate rows and $20.2 million of phantom money — and the work has moved from backfilling to interrogating: a settled verdict on how refunds net against contributions, and a found mechanism for the negative Schedule A rows that made some committee totals look wrong. This is the “who funds whom” data behind Essentials, presented without commentary.
Every state now has at least one researched official, and all 50 congressional delegations are covered at the national level. Shading shows the depth of state and local research per state. Colorado moves up a tier this cycle, on the Colorado Springs wave. Puerto Rico is an outlined tile — 161 seats seeded and mostly portraited, with no researched stances yet, so the territory is visible without being shaded as researched; that seeding landed on August 12 and this page missed it in two consecutive briefings before saying so.
Hover any state for its exact count. The seeded tier is mostly governors and statewide officers; deeper tiers add legislatures, counties and cities. The admin console goes county by county; this is the state rollup.
Who actually uses what we build — measured, not guessed.
591 people used the platform in the last 90 days. That is the whole audience, across every app, over 1,874 sessions. What they do once they arrive is genuinely good — the average session runs just under eight minutes and 7 in 10 don’t bounce — so the constraint is arrival, not engagement.
The front door works, and this is the first pull comparable to the last one. 99 of the 306 people who reached the main site clicked through into an app — 32% in the quarter, 39% in the last 30 days. Last cycle changed the basis, counting each app once across every hostname that serves it after two apps turned out to be counted as several; re-running that hostname breakdown today surfaced no new host, so these figures rest on the same basis as the last briefing and can be read against it. The main site went 299 to 306. It gets re-run every cycle regardless: a new hostname is how the next undercount arrives.
Read every figure here as an order of magnitude, not a trend line. At this volume a couple of people either side is the entire swing. No work this cycle was aimed at traffic.
For each app: everyone who arrived, against the people who did the one thing that app exists for. Bars are the 90-day rate; counts are people, over 30 days and the full quarter.
Two of these are start problems, not finish problems. Only 9 people started a trivia game in a quarter and 6 finished. Read & Rank is the same shape: 18 races started, 9 ballots revealed. The Compass is the opposite — 79 of its 114 visitors opened the quiz and 32 saw it through, so two thirds start and better than a quarter finish.
Unique visitors per app, last 30 days and the full quarter. Bars scale to the 90-day count. Each app is counted once across every hostname that serves it — Treasury Tracker draws traffic on four, the main site on three.
Treasury Tracker is the second most-visited thing we run and holds the largest dataset on the platform — 2.3 million budget line items. Whether those 165 people reached a budget is the question the note below is about; the instrumentation that answers it shipped on the 16th and is now capturing.
An earlier version of this section said seven people in a quarter opened any of Treasury Tracker, as 5% of 152 visitors. The rate was not sound and it is withdrawn. The events were fine; the denominator was not. Treasury pushed its in-app navigation into the URL but never captured a pageview for it, so every visitor was recorded against the bare domain and nothing else — and the app opens on an entity picker rather than a budget. The direction was wrong too: those few produced 56 drills, 29 line-item views, 41 entity selections and 13 year changes — the handful who reach a budget use it hard.
The fix shipped on the 16th and is capturing. Budget views now arrive as their own URLs — ?entity=austin-tx&year=2025 and the like — and 10 of the 13 people on those hosts in the five days since reached one. Still not a publishable rate, and the sample says why: the entities in it are Austin, Travis County, Los Angeles and Colorado Springs — mostly us checking this cycle’s own onboarding. Treasury stays unmeasured rather than low until traffic that is not ours accumulates, but it is now measurable, which it was not when this card was written. The rule it bought applies to every app here: a low count is a claim about instrumentation until you have read the code path.
How much further we have to go, honestly.
Run the city build again on the next served community, close the last 16 uncovered Washington legislators, and re-run Colorado’s legislative half once the state exposes more than the current session. Then: finish the 2026 House candidate waves, keep re-sourcing what the source-supports check retired, land Read & Rank candidate quotes, and get Fallacy Finders v1.1 real clips flowing through On the Record. Success in November looks like: 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. This is where the platform stops being a reference tool and becomes a community, and where our sustaining-donor model actually lives.
Designed but not yet 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 to stand under it. We are honest about this distance; it is the point of the whole journey.
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. Also in the site footer.