As best I understand the Twitter spat, he was claiming the client was making 1000 RPCs to generate the home timeline - hence it being slower in India where mobile internet is slower. Except the Android dev pointed out it was actually only 6 (IIRC) calls from the client and slowness had nothing to do with that.
Possible he did, I suppose, but I never saw it in any of the recaps / threads. But also wouldn't make a lick of sense talking about India vs US for speed...
The Twitter product has not meaningfully changed or evolved in 10 years.
Revenue and user growth and product stagnated. So who cares about optimizing a further 100ms improvement or what OSS they maintained?
The “interesting stuff” you see in this post is what many engineers love to do. Make work for the sake of work. It doesn’t always translate into growth.
> The Twitter product has not meaningfully changed or evolved in 10 years.
In the last few years alone they made huge changes like Blue, reader mode, ad-free articles, super follows, birdwatch, spaces/ticketed spaces, the list goes on.
All under an FTC consent degree which required "that Twitter produce written privacy assessments of any new “product, service or practice” — or when Twitter updates those things — that could affect user data or put it at risk."[1] - not an environment where you can "move fast and break things"!