Seems to me like FANG is primarily sacrificing QA to increase productivity, with AI as the reason. Will be interesting to see if they too are 'peeing their pants' (warm and nice -> cold and not so nice), or if it actually will be sustainable.
LLM's were trailed as customer service bots, which backfired (for the most part). Now they're exploring different areas.
May be it's easier in FAANG because a lot of the work comes from the sheer (and unnecessary IMO) complexity of the stack/tools.
In fact, a common meme within Google is that Engineers are mere plumbers routing protos from one place to another.
(Given how brittle these systems are though, I'm not sure if I trust LLMs to be better; but they may be more suited for dealing with un-interactive delayed-feedback baroque systems that are the norm in such places.)
LLM's were trailed as customer service bots, which backfired (for the most part). Now they're exploring different areas.