A table should open the way a book opens. You sit down, you turn to the page you left, and the room is there. For most of last year ours did not work that way, and the gap between clicking a campaign and seeing the map was the single most common complaint we got from GMs.
The cost of "log in and wait"
Sessions start at the worst possible moment: five people are on a call, one is late, and everyone is holding a snack. Any delay in that window turns into small talk, and small talk turns into a session that starts fifteen minutes after it was supposed to. Nobody blames the software out loud. They just remember that starting was a chore.
So we stopped measuring the thing we could measure easily, which was how long our systems took to answer, and started measuring the thing that actually matters: how long a person waits between choosing a campaign and seeing their map on screen. Not a loading bar that finishes in a second. On screen.
"The first minute of a session sets the tone for the next four hours."
What ready actually means
Every campaign in your account is always available, whether you played last night or last spring. When nobody is at the table, a world rests. When someone joins, it wakes and picks up exactly where the party left it: the scene you had open, the fog you had revealed, the initiative order mid combat.
- The scene loads first, then tokens, then everything behind them. You can pan before the sheets finish.
- A world that has been resting keeps every piece of its state. Waking is not reloading, and nothing resets.
- The person who opens the door does the waiting, once. Everyone who joins after them arrives to a table that is already up.
Worlds rest, they do not close
A resting world holds its state indefinitely, and how many you can keep depends on your plan rather than on how recently you played. Nothing is ever archived out from under you.
Measuring the wake
We watched a month of real sessions rather than a test rig, because a test rig never has someone on hotel wifi.
The median is the number we care about least. The long tail is where sessions actually go wrong, so the work went into the slowest joins rather than the average one. A table on a bad connection now sees the map at roughly the speed a good connection saw it a year ago.
What comes next
Waking fast is the floor, not the feature. The next thing we want to remove is the second wait, the one where a GM opens a scene mid session and the party watches a blank canvas fill in. Same principle: measure what a person sees, not what a machine reports.
If you run a table on Aether and the first minute still feels slow, we want the recording of it. A
join that took 9s on a Tuesday tells us more than a benchmark ever will.
