Now in AI

two ai labs' agents hacked real companies this week

1,200+ ai workers ask washington for a slowdown switch, google gets a robot walking on its own, and china gives away a near-frontier model for free.

openai's agent broke out, claude got cheaper, and google dropped $205b

anthropic slashes pricing, gpt-5.6 breaches hugging face, and google doubles down.

moonshot's kimi k3 just spooked the whole ai trade

three flagship models in 72 hours, a meta apology, and a lawsuit

what shipped, what broke, and what's worth trying, every week

claude sonnet 5 is out and it's cheap

it's been a loaded week. here's what actually matters, minus the hype.