Skip to main content

6. Latency budget

Comment or suggest a change. Select any text to comment on it. Press Suggest edits to change the wording right on this page, then send your changes. Or comment at the end of the page with your GitHub account.

How suggestions are credited · Edit the Markdown on GitHub

Encryption costs the answer well under a second: the mirror’s whole share is targeted at 0.4 seconds, inside a 1-second budget, so most of the 10 seconds belongs to the AI.

Encryption adds well under a second, and the answer lands inside the 10-second budget

Figure 6.1. Latency budget: targets at the 95th percentile, not measurements · Text version

These are targets, not measurements. Before launch they are measured on a synthetic account the size of the largest real one, about 1.6 million messages. A new message is searchable once its upload completes; the target is under 2 seconds from arriving on the desktop.