1) A graph of latency on FTX over time:
Conversation
4) Instead, we're going to do something a big more ambitious: re-writing the bulk of the matching engine.
The downside is this won't pay dividends until it's done, probably in two months or so.
The upside is that it should, hopefully, give us 2x or better improvements.
In two months the 2x improvements won't be enough. You need to have solution for 4x already by than.
Better have more than two workers this time around—enjoy the journey
GIF
read image description
ALT









