Rosetta Intel

Daily developments at OpenAI / Anthropic / Google, with 🛡 marking security-related items.

OpenAI

Model arms race continues (carryover)— June saw OpenAI / Anthropic / Google / xAI move in lockstep (Gemini 3.5 Pro, Claude Mythos 1, a rumored Claude Sonnet 4.8, xAI Grok 5). OpenAI's IPO timing remains a watch item: it expects rival Anthropic to list first, and both have filed confidentially with the SEC.
No major product/model releases captured today.

Claude Mythos 1 released; Sonnet 4.8 rumored— Anthropic is among those shipping new models this round.
Colossus 1 compute agreement— Anthropic secured the entire capacity of SpaceX/xAI's Colossus 1 data center (Tennessee): 300+ MW and 220,000+ Nvidia GPUs, a significant expansion of training/inference compute.
Talent inflow— AlphaFold's John Jumperleft Google after nearly nine years and, following a break, joined Anthropic; part of a broader migration of senior researchers (echoing Google's earlier losses of Adler, Pritzel, and Shazeer). The industry-wide shift of research benches and model-safety capability is worth watching.

Gemini 3.5 Pro approaching— at I/O (5/19) a "next month" release was promised, and the Pro model ID has not yet been made public; Gemini 3.5 Flash is already live. Positioned as the strongest agentic / coding model.
Boston Dynamics × DeepMind— integrating Gemini Robotics-ER 1.6into the Spot robot dog and the Orbit AI visual inspection platform, advancing embodied/industrial inspection work.
Talent outflow— the same period saw the loss of Noam Shazeer (to OpenAI) and John Jumper (to Anthropic), drawing industry attention to the stability of its research bench.

Security-related items also appear in the main brief's "AI Security" section and are marked 🛡 here. Source configuration: see intel/sources.yaml.