I encourage you to read more of Dan Luu's posts, because he talks about things like what you mention, and consistently makes insightful points about them, e.g.

I encourage you to read more of Dan Luu's posts, because he talks about things like what you mention, and consistently makes insightful points about them, e.g. https://mastodon.social/@danluu/109332405719026619 https://mastodon.social/@danluu/109332405719026619 https://mastodon.social/@danluu/109339567207009389 https://mastodon.social/@danluu/109339567207009389 The point here isn't just that the features are cool, but that they provide a good return on investment. At Twitter's scale, a 1% improvement in CPU usage or user retention is massive, and could be worth having a team around for in the hypothetical case that's all they ever did. If, in the lifetime of the specialist team like the kernel team, a single person found something that persistently reduced TCO by 0.5%, that would pay for the team in perpetuity, and Twitter’s kernel team has found many such changes. [1] The "number of calls" metrics that Elon Musk have been throwing around without much understanding are probably themselves a product of Twitter's investment into performance - someone showed him that you can get a call graph of all services involved in producing the front page, and what he got out of it was "1000 is a big number". Presumably it will get rewritten until performance visibility is sufficiently reduced that he can't be bothered to micromanage it any more. [1] https://danluu.com/in-house/ https://danluu.com/in-house/