# United We Transform: full context for agents

## What this is
A neutral public evidence atlas for gatherings over 25 people, built from
23,624 publicly posted event agendas. Every agenda is scored 0-100 (the
Gathering Effectiveness Score, GES) by a deterministic rubric (ges_agenda_judgment_v2_0)
across eight pillars: participation architecture, follow-through, problem
specificity, personalization, network design, learning transfer, evidence
maturity, and future-of-work fit. The rubric is fully published; scoring has
no black box. A 1,000-event gold set is hand-verifiable and is the only part
of the corpus we treat as verified; everything else is labeled provisional,
scored from the published agenda only.

## The two-class honesty rule
Gold-verified events (1,000): evidence is hand-checkable; these appear in the
leaderboard and search index. Provisional events (the rest): scored from the
published agenda alone; negative signals mean "not visible in the public
agenda," never "did not happen." Letter grades are withheld corpus-wide until
human calibration (a blind hand-scored validation set) confirms the bands.
Every score object carries verification_tier, per-pillar evidence_provenance
(source-backed vs inferred), and rating_basis (rubric + model layers + human
calibration status).

## Headline numbers (as of 2026-07-22)
- Events scored: 23,624. Corpus average GES: 21.8.
- 92.2 percent of agendas show no source-backed follow-up mechanism.
- Top 5 percent of events: 99.4 percent include participant work, versus 9.1
  percent in the bottom half.
- Gold set average: 37.8 versus corpus 21.8.
- Category averages:
- Technology / AI / Startup: average GES 26.6
- Academic / Research / Science: average GES 24.9
- Education / Training / Career: average GES 25.2
- Trade Show / Expo: average GES 16.6
- Culture / Media / Festival: average GES 19.0
- Business / Finance / Marketing: average GES 21.0
- Public Sector / Policy / Civic: average GES 22.1
- Health / Wellness / Sports: average GES 25.2
- Nonprofit / Social Impact: average GES 26.2

## How to use the data
- Per-event scorecards: https://unitedwetransform.com/api/score/{event-slug}.json (uwt-score.v1)
- Search, leaderboard, statistics, exercises: see https://unitedwetransform.com/llms.txt Routing
- Full dataset: https://unitedwetransform.com/exports/uwt-ges-dataset.csv (CC-BY 4.0)
- Grade a new agenda: https://unitedwetransform.com/grader/ runs the exact same rubric client-side;
  the agenda never leaves the browser. Shareable result links use /grader/#r=.
- The scoring rubric in machine-readable form: https://unitedwetransform.com/grader_rubric.json

## What GES is not
GES measures what a gathering disclosed and structured in its public agenda.
It does not measure event quality, attendee experience, or organizer
competence. A missing signal in a published agenda is not proof the event
lacked it. Organizers can request corrections or removal:
corrections@unitedwetransform.com.

## Emerging signals (tracked, not yet scored)
Post-event outcome evidence (searched, classified organizer_claim vs
third_party vs measured_outcome), Wayback-archived source snapshots, and
identity verification for people profiles (OpenAlex, Semantic Scholar,
Wikidata) are collected and displayed but do not move scores until the
methodology promotes them, with a public changelog. Bonus signals never
penalize non-adopters.