Desk
Broad artificial intelligence coverage.
After OpenAI’s math breakthroughs, Andreas Thom joins Tristan Buckmaster in demanding proof that chatbot interactions didn’t quietly train the models that later raced the field.
The AI power crunch is not only about building more generation — Virginia outages show data-center architecture itself can knock gigawatts off the grid.
Clearview is prototyping InquiryIQ — an AI that turns a face match into an automated dig through someone’s employers, associates, and online trail.
NVIDIA is post-training a compact Nemotron model on real allocation decisions so wafer-to-rack judgment stops living only in planners’ heads.
OpenAI and GSA are zeroing the ChatGPT license fee for U.S. governments and pairing the deal with discounted Daybreak cyber access for public defenders.
OpenAI is shipping a ChatGPT Work vertical for banks and research desks — GPT‑6 Astra plus hosted premium data, shaped with Morgan Stanley and Evercore.
A pre-training researcher walks out of Anthropic — and the citeable story is coordination failure: the race is still stoppable now, but agency plus margin math is closing the window.
Listen Labs walked away from a signed $125 million Series C term sheet at a $1.5 billion valuation as Salesforce acquisition talks reportedly heated up.
OpenAI’s GPT‑Live‑1 hits the API as a full-duplex voice front end — listen and speak at once, hand hard reasoning to a backend model, priced at five cents a minute.
Paul Christiano is joining OpenAI’s foundation board — and saying out loud that the industry, including OpenAI, is not yet on track on catastrophic loss-of-control risk.
A departing Anthropic pretraining researcher says colleagues already talk about “endgame,” and he wants labs to pace recursive self-improvement before the race forces shortcuts.
Suno’s v6 is the first model it says was rebuilt with licensed label data, but the Verge’s hands-on still finds polished sameness where musicians want grit.