// Deep Dive · AI
We ran 7 local LLMs on an RTX 5070 Ti — and a 35B model surprised us
Quantization, layer offload, and how much VRAM you actually need. We tested seven offline models and measured tokens/s — with every step to reproduce it yourself.
Security
Inside the 0-day that ran on 40,000 routers before the patch
Newsroom · 9 min
Science
Telescope captures the oldest galaxy ever seen — 290 million years after the Big Bang
Newsroom · 6 min
Cars
We tested the Chinese EV that costs half a Tesla — and almost no one saw it coming
Lucca Ortolan · 11 min