diff --git a/Architecture.md b/Architecture.md
index b3fee32..e985dbf 100644
--- a/Architecture.md
+++ b/Architecture.md
@@ -204,6 +204,16 @@ just as well on a screen that froze a minute ago. The drawn-frame counter is the
 `__gm.renderStats()` gives it, and step 33 is the check that the editor really does go quiet and
 really does wake up again.
 
+**Work that costs more per pointer event than a frame is worth waits for the frame** —
+`onFrame(key, fn)` runs `fn` once before the next frame is drawn, however often it was asked for
+under that key, and `cancelFrameWork(key)` drops it. A drag writes the model on every pointer
+move, and a model event rebuilds the panels, the run in the editor and any picture drawn from the
+model: on a busy scene that measured 36 ms per pointer move, while a mouse reports around 125
+times a second and the screen takes 60. Without coalescing the pointer outruns the drawing and the
+editor only catches up once the hand stops. The work runs **before** `app.render()`, not after,
+so what it writes is in the frame it waited for. The Vector art tab's drag uses it; while the
+gesture runs, its own copy of the rings is the truth and it never reads the model back.
+
 The toolbar's right edge shows the frame rate in edit and play mode alike; `__gm.fps()` exposes
 the same number to tests. It counts **drawn** frames over the time the renderer was **awake**:
 idle seconds are not slow seconds, so the readout answers "how smooth is it while it moves" and