Technology
Google’s Gemini 4 Argon targets long-horizon work
Google is putting a new frontier model into the hands of trusted cyber defenders first, with a million-token context and a deliberately phased rollout.
Fieldnote topic file / 5 briefings
Artificial Intelligence
The models, research and safety questions moving AI forward. Concise, sourced briefings on the developments worth keeping up with.
Recently published
Artificial Intelligence, in focus
Technology
Claude Sonnet 5.5 puts speed and task cost in focus
Anthropic’s latest Sonnet release is positioned as a faster everyday model, with the company reporting lower costs for common work.
Technology
Claude Opus 5.5 makes an efficiency pitch for complex work
Anthropic says its newest Opus model approaches the previous generation’s performance on many tasks at a lower operating cost.
Technology
OpenAI proposes a reporting framework for model misalignment
A new incident-reporting process aims to make unexpected model behavior easier to document, investigate and learn from.
Technology
OpenAI describes an AI-assisted approach to Navier–Stokes
A coordinated system of AI agents produced a proposed solution that OpenAI says was formally checked, prompting fresh discussion about AI in mathematical research.