AGV Tracker — AI Governance Venues, 2015–present v0.1 · pilot

Analytical dashboard

Three default views filter the AGV population around different analytical questions (CLAUDE.md §6.2): continuous bodies only (institution-building signal), recurring convenings only (annual/biennial conferences and summits), and one-off included (full dataset). Switching the toggle re-renders every chart below; the explanatory paragraph directly under the toggle tells you what you should and should not read into the current view.



Current view contains venue out of in the full dataset.

1. Cumulative active population over time

How the live stock of AGVs in the current view has grown. A venue is "alive at year Y" if it was founded on or before Y AND is not current_state = terminated. absorbed and succeeded venues are counted as alive because their work continues inside their successor.

2. Founding rate by entity type (small multiples)

Annual foundings, faceted by entity_type. Empty facets indicate a type that is filtered out under the current view (e.g. one_off_summit is absent from the continuous view).

3. Founding rate by topic focus (small multiples)

Annual foundings, faceted by topic_focus_primary. Inflection points on specific topical rows often correspond to discrete external triggers (ChatGPT launch for genai_content, Bletchley Park for safety_frontier).

4. Lifecycle state distribution

Distribution of current_state across the venues in the current view. The presence of terminated / succeeded / absorbed buckets at the bottom is the population-ecology signature the dataset is built to capture.

5. Topic × entity-type heatmap

Where the governance effort concentrates. Clusters of cells pick out the current hot corners of the landscape; empty rows/columns show the structural gaps worth filling (CLAUDE.md §5.4 coverage matrix is the complementary metadata view).

6. Geographic scope and lead actor

Two complementary breakdowns of who and where: the jurisdictional reach (left) and the driving institutional actor (right).

7. Relations network

Force-directed graph of the agv_relation.csv edges (succeeds, absorbed_into, parent_of, coordinates_with, convenes_within, member_of, references_principles_of). Node colour encodes entity_type; edge stroke width encodes relation strength (hierarchy and succession are drawn thicker than referencing relations). Terminated venues are drawn semi-transparent.

The graph respects the current view toggle: nodes are restricted to the filtered set, and edges are kept only when both endpoints are visible.

Interactions. Drag a node to reposition it (release to let the force re-settle). Scroll / pinch to zoom (0.3× … 4×); drag empty canvas to pan. Hover a node or edge for a native tooltip.

Edge density note. The v0.3 seed ships only relations across venues; most nodes are therefore isolated dots. This is expected — relation data will densify as v0.4 backfills member_of / parent_of / coordinates_with edges (CLAUDE.md §3.7).

nodes · edges in current view.


All charts above re-render on toggle change. For a row-level inventory, see the Venues directory. For the per-AGV evidence trail, open any venue's detail page from that listing.