Revolut cut AI costs 90% by moving 2,000 engineers from expensive OpenAI and Anthropic models to self-hosted Chinese GLM models. It reports 60% faster coding and 70% shorter engineering lead times (from starting work to being ready to merge). But the productivity story is more interesting.

Show Post

it’s probably worth noting the GOP is basically doing the exact same thing in Nevada but the Times is not reporting on it

Show Post

A decade after beginning the project, Ryan McGee, Olivia Kosterlitz, Artem Kaznatcheev, Ben Kerr, and I have a paper in PNAS this week that is one of the best pieces of work I've ever been involved in. Without paywall: carlbergstrom.com/publications... Below, a thread about (part of) the paper.

Show Post

This is me whenever someone complains that Americans are just sitting around and not protesting this administration:

Show Post

“There has been a proliferation of karaoke rooms … in the densely populated neighborhood of Koreatown” — right. I’ve been saying: if there’s anything L.A. has too much of right now, it’s nightlife. (*literally facepalms*)

Show Post