OpenAI says it shut down two false-front influence operations that used ChatGPT to staff a fake Latin America think tank and a ring of journalist personas. NVIDIA put science commitments it values at $1 billion over five years on the table at a Washington Genesis Mission event. RTX Spark laptop preorders opened next to Microsoft's new Windows agent containers. Codex CLI 0.162 added managed Git worktrees so local agents can branch without colliding.
In the news
OpenAI false-front operations
OpenAI published a Safety report on October 8. It describes two covert influence operations that used ChatGPT plus ordinary tradecraft to run false-front entities.
The Russia-origin cluster, which OpenAI nicknamed Dark Clark, used a fake persona named Mia Clark to control a Latin America think tank called the Social Research Center. The operators drafted reports, fake leaks, and audio scripts aimed at Ukraine and at politics in Argentina, Bolivia, and Ecuador. OpenAI assessed that case as Category 5 on the Brookings Breakout Scale. It is the first Category 5 disruption in 30 influence-operation reports since early 2024.
A second operation, which OpenAI traces to Iran, ran seven journalist personas pitching long-form pieces to small and medium outlets. It also ran comment batches on US-Iran war topics. OpenAI assessed that case Category 4. Both landed content in real media, not only on social platforms.1
NVIDIA $1 billion for US science
NVIDIA announced on October 8 commitments it values at $1 billion over the next five years. The commitments cover US research in quantum computing, healthcare, and energy security. The announcement was made at the Science: A New Golden Age event in Washington. The US Office of Science and Technology Policy marked an expansion of the Genesis Mission there.
NVIDIA said the mix covers higher-education research institutions, American quantum work, and cloud providers serving US government missions. The company is a collaborator on phase 2 Genesis awards in quantum computing, fusion, accelerator design, and microelectronics. It also points to DOE systems at Argonne and Los Alamos. NVIDIA describes the figure as commitments valued at $1 billion: compute, infrastructure, and support rather than a cash grant.2
RTX Spark, MXC, and DGX Station for Windows
NVIDIA and Microsoft used a Windows AI event in San Francisco on October 7 to show local Blackwell hardware for Windows. Microsoft said Microsoft Execution Containers (MXC) are generally available. The containers give agents OS-level isolation so they can run in the background under Windows control.
NVIDIA RTX Spark pairs Blackwell RTX with a Grace CPU in Windows laptops and compact desktops. NVIDIA says those machines reach up to 128 GB of unified memory and about one petaflop of FP4 performance. Laptop preorders are open, with availability on October 16. Compact desktops are due in November.
NVIDIA also previewed DGX Station for Windows, a deskside GB300 box. NVIDIA cites 748 GB of coherent memory and up to 20 petaflops of FP4. The company says that is enough to run trillion-parameter-class work locally.3
Codex CLI 0.162 managed Git worktrees
OpenAI tagged Codex CLI 0.162.0 on October 8. The release adds tools to create and list managed Git worktrees from trusted local projects when the worktrees feature is on. Parallel agents can then branch without sharing one working tree.4
A 0.162.1 follow-up on October 9 fixes a TUI crash on multiline async questions. It also fixes startup failures when a background server's feature flags differ from CLI defaults.5
