Desk
Coverage tied to OpenAI products and research posts.
OpenAI’s specialist security model scores 48.5% on ExploitGym, helped produce exploits for 948 Linux kernel bugs and 81 new Firefox issues, and is rated Critical. It also stayed in scope on honeypot tests 99.93% to 100% of the time. Access breadth is the open question.
Integration platform Merge runs GPT-6 Astra, ChatGPT, and Codex across product reviews, marketing, and a "dinner bot" that cut executive-event prep from dozens of hours to about 30 minutes.
OpenAI is wrapping its next compute campuses in a National Community Compact — pay-your-own-way energy, public reporting, and a policy blueprint asking lawmakers to make the rules stick industry-wide.
Franklin Templeton is wiring Codex and OpenAI tooling into a "connected intelligence" layer so portfolio managers see diverse research views and risks — not a narrower feed of what they already like.
OpenAI's Chris Lehane casts the 2028 race as AI's New Deal moment — five policy priorities that sound public-spirited while still clearing the runway for wider adoption and more data centers.
OpenAI’s Images 2.6 update lets ChatGPT generate and edit animated GIFs in one flow — a small consumer feature sitting on top of 4 billion images a week.
Uber says third-party AI agents booked just four rides and about 500 food orders last quarter — a rounding error that still tells us how the disintermediation fight will go.
Sam Altman says his RAM remark was taken out of context, but the joke on X doesn't change the fact that AI data centers are soaking up the world's memory supply.
OpenAI is buying Pierre Computer Company, whose Trees and Diffs libraries already power review inside Codex. With its Pando source-control backend, OpenAI is moving to own the layer where agent-written code is stored, tracked and reviewed.
OpenAI’s cyber pipeline validated hundreds of Linux local privilege escalations, wrote patches, and argued that unpatched low-priv access is already root for a serious attacker — defense now means isolation and faster fixing, not hope.
OpenAI's cyber model is moving into the government cloud, and the useful part is less the model than the gate around it: verified defenders, monitoring, and data that stays put.
DevRev is wiring GPT-6.1 Sol into Computer and Operator so enterprise service agents can finish hard requests — with reversible actions, permissions, and answers that point back to sources.