Tabletop exercise · Saturday, July 4, 2026 · Washington, DC
Exercise time (EDT)
11:30
Crowd on site
—
est. from zone sensors
Heat index
—
NWS Sterling
Open incidents
0
awaiting decision: 0
Patients treated
0
DC FEMS + mutual aid
EMS units free
—
of 48 staged
Metro load
—
Mall-area stations
Readiness
92
exercise score / 100
Crowd: comfortable
Crowd: dense
Crowd: critical density
Drag to orbit · scroll/pinch to zoom
What-if
Change a planning assumption. The map and numbers update as you drag.
12:0014:0016:0018:0020:0022:00
How this was built
One request typed during the meeting started it. A day of follow-up requests, testing and AI review finished it.
The request
Build me a 3D model of America 250 event over the 4th of July in DC with lots of event scenarios, scans of places in DC of the event, etc.
Interview answers: AI as a thinking partner + "my team could do this" · ready by the next break · live exercise framing · Ask-the-AI, after-action report, aerial stills · make it interactive.
The receipt
30 minfrom the request to a first working link. A rough first version.
+1 dayof follow-up requests, testing and AI review before it looked like this
~1,800lines of code, all written by AI
9scripted injects. Any what-if the room invents runs live inside Claude.
What the AI did
Modeled the Mall from public geography: 20+ landmarks, ~800 city blocks of buildings, the Potomac, Metro stations, a crowd model of ~16,000 tracked points.
Wrote a 9-inject master scenario events list (MSEL) spanning ESF-1, 6, 8, 9, 12, 13 and 15, each with a decision point, consequences and guidance.
Wired in an AI advisor that answers questions with the current exercise picture, critiques each decision, invents new injects on request, and drafts the after-action report. It runs live when this page is opened in Claude; on this public copy it falls back to built-in guidance.
AI reviewers did realism passes and tested the page on desktop and phone. No emergency manager or agency has reviewed it. Treat it as a first draft for your planners, not a validated MSEL.
Added over the next day, on request: a phone layout, What-if dials for planning assumptions, Your City, which drafts a six-inject exercise for any city, and the Ad Astra AI look. Each was tested on a phone-sized screen before going live.
What it would take for your city
Swap the map footprint and the event, keep the engine. Your venues, your MSEL, your ESF structure. A first version is a working session, not a procurement. Talk to Ad Astra AI.
Your city, your exercise
Name a city and its signature event. The AI drafts a six-inject tabletop with local venues, agencies and decision points.
Open it on your phone
Point your phone camera at the code. It opens in any browser with no account. The live AI features run when the page is opened in Claude.