Andrew Bosworth has built a reputation inside Meta as the executive willing to say the quiet part out loud. This week he did again.
When Flaviu Radulescu set out to build a text-to-image company in 2023, he kept running into the same wall: the cost of inference.
Infosys pairs OpenAI's models with its Topaz fabric as systems integrators become the frontier-model enterprise channel.
At AIPCon 10, Palantir paired a multi-tiered Google Cloud deal with enterprise AIP deployments in law, construction and insurance.
Defakto raised $30.75M to secure the identities of AI agents and other non-human actors that now outnumber humans on networks.
Nadella's first-ever long-form essay argues the AI economy will be won by owning the loop, not the model.
In the seventy-two hours after Moonshot pushed Kimi K2.7 Code, two more coding models landed and a leaderboard reshuffled twice.
A new paper argues you can improve expressive robot policies by steering them at test time, not fine-tuning them with brittle RL.
A new paper argues the most capable mobile agents won't just tap screens; they'll reach past the interface to a shell or tool call.
Modern AI systems are spectacular pattern-matchers, but a new paper argues genuine causal reasoning is foundational, not optional.
Most AI research agents see the world through a narrow window: a retriever's ranked list. Dr-DCI lets them touch the corpus.
The DOJ asked a court to toss the NAACP's pollution suit against xAI, calling the company's Memphis turbines vital to national security.
Roughly 200 Stanford graduates walked out on Sundar Pichai's commencement speech over Google's Project Nimbus and ICE ties.
Colorado's first-in-nation AI discrimination law drew a DOJ intervention and xAI suit, freezing its June 30 deadline.