Core backend engineering patterns have remained unchanged for a decade.
ughh, i haven't had to worry about auth refresh patterns, and re-entrant reset sequences and globally aware database sharding schemas in a WHILE. in many ways AI research is so much simpler than normal 'FAANG Engineering' . at least nothing has changed in the last 10 years so it was easy to slide right back in. its just like riding a bike that actively tries to make you crash
Some users express surprise at how fun AI work is compared to FAANG auth and sharding, while others sarcastically dismiss FAANG roles as inventing problems to justify headcount.
@_xjdr Just jam it all in one big postgres lol
ughh, i haven't had to worry about auth refresh patterns, and re-entrant reset sequences and globally aware database sharding schemas in a WHILE. in many ways AI research is so much simpler than normal 'FAANG Engineering' . at least nothing has changed in the last 10 years so it was easy to slide right back in. its just like riding a bike that actively tries to make you crash
AI research is more fun but also feels more rewarding because it's the world of the unknowns, so many things could work
in comparison, "normal" engineering feels more dull because it's always the same thing to some degree; however it can remain fun if you think of it from a design perspective rather than the more grey outlook
@_xjdr I can’t believe how much fun this side is
@_xjdr FAANG engineering: inventing problems just complicated enough to justify the headcount.
@_xjdr This is the underrated split. AI work feels harder intellectually, simpler operationally until you re-enter auth/state/sharding land. Infra complexity is mostly adversarial complexity: many boring edge cases, no mercy.