Set your own conditions and the page recalculates. Everything that could be measured was measured by a run; everything else is honestly marked as an assumption you can change.
We measured neither objects per meeting nor bytes per object — those are assumptions tied to conversation length. What is measured here is something else, and more important: the system counts its own volume and growth per organisation, and shows the slope rather than just the size.
It is the growth slope, not the starting size, that bankrupts a long-lived memory — so the system measures volume and growth itself and reports them per organisation. Decay is typed: contextual links (mentioned near each other) weaken, while structural facts — decided, assigned — never decay.
Only measured results below. Every row names the run it came from; arms marked invalid do not appear in this table.
| what was compared | baseline | ours | run |
|---|