I'm not sure if 'picking apart Gemini's technical setup' is the right way to describe it, I tend to think of 'picking apart' being negative while the comments are entirely positive.

I meant it as in he scrutinized it.

He's really just looking at the frontend side of things.

I would not consider going over their frontend assets and request headers anything close to picking apart their technical setup though. For anyone a little more knowledgeable that post is just a collection of random facts about the apps visible front-end.

Yes, that's true - it was a loosely categorized collection of my thoughts while going through the frontend. The frontend architecture is very similar to that of BitMEX, the exchange that I built, so I am very interested in the choices they've made, why they chose them, and what's different.

Gemini is a spot exchange (simple buy/sell) while we're a derivatives exchange with much more complex requirements, so you'd expect a different set of decisions and tradeoffs, which is what I found. Gemini's real value (at this point) is in its ability to navigate regulatory capture, not necessarily in its technology. But their technology is a cut above what you usually see in Bitcoin exchanges. The exchange landscape has been plagued with unreliable/buggy exchanges, like the late Mt.Gox and the still-limping Bitfinex (which is much more complex).

On the whole, basic spot exchanges without leverage are relatively easy to create. I would love to do a more complete analysis but of course I don't have any inside information. I would be very interested in their backend, which appears to (possibly) be Scala. No clues as to whether they're using a SQL database or something more specialized like KDB+, which we use and love.

I find this kind of analysis way more interesting than highscalability. Different audiences, but these kind of frontend-centric articles are incredibly actionable.

interesting... still doesnt talk about the actual API stack though. Conformal is using golang and Coinbase is using Ruby... wonder what these guys are using.

Hey - author of that post here. I can't tell for sure what the stack is by looking at its output, but judging by common experience in the LinkedIn profiles of the engineers, it looks most likely that it's Scala.