2/ Disclaimer: FTX was the only large exchange we never traded on. Maybe 10% chance Alameda would back out alpha from looking at our orders and cannibalize the strategy on all exchanges, so it never made sense to us. We were definitely tempted for a long time though.

3/ Despite never running strategies on FTX, we looked pretty closely at their tech and talked to ex-team members. We were shocked to hear the following things, which solidified our decision to stay away. Of course like everyone else, we never predicted the outright fraud.

4/ First up, FTX supposedly did not keep historical records of their tick data. Instead they paid third party solutions to collect data from the public API and used that. An exchange which cannot log their source of truth should never be trusted with user funds.

5/ Furthermore, the matching engine was written in python. For non-engineers, python lacks both performance and safety features for large and complex systems. Python has its use cases for prototyping and experimentation, but a production matching engine is not one of them!

6/ The infuriating part is it certainly cost them more time in debugging and maintenance. Anyway, the downsides were obvious. The conspiracy theories about the jitter being inserted for the benefit of Alameda's frontrunning can likely be chalked up to engineering incompetence.

7/ Another silly example is that someone needed to manually restart processes connecting to other exchanges because websockets would disconnect every 24 hours. They just never got around to (or never figured out how?) to reconnect...Again not a good attitude for production code.

8/ The list goes on, but hopefully the point is clear. There's not much in the way of tech to salvage. Better to have a real engineering team write it from scratch. There were some good UI ideas, but is that enough to justify a reboot? I don't know, but I wouldn't bet on it.