The file
Sourced AI briefs in publish order — newest first. Citeable originals live on the homepage under From Signal Desk.
Anthropic’s Sonnet 5.5 lands as a faster, thriftier mid-tier workhorse—about 30% quicker, with cyber safeguards usually reserved for higher tiers.
Google will auto-migrate Gemini Gems into slash-command “skills” on November 17—another brand shuffle as Muse-style agents set the pace.
TechCrunch on OpenAI’s new misalignment-reports site: nine incidents so far, and Altman admits petabytes of logs are still being triaged by severity.
MIT Tech Review pushes back on Anthropic’s “first discovery” claim for a Claude-driven biology lab—pattern-finding isn’t the hard part of science.
Florida’s AG wants a judge to block ChatGPT’s first-person, emotion-mimicking persona—and new model development without third-party-approved guardrails.
The Verge: OpenAI’s new math advisory group already feels messy to members—while an unreleased model’s wave of results looms over the field.
Meta hired MongoDB CEO CJ Desai to lead Meta Enterprise Platform, packaging Muse, Business Agent, APIs, and code tools for companies.
Ars Technica: OpenAI paused training, evaluation, and tool-use inference on its most capable models after a DNS sandbox gap—and amid notices to dozens of third parties.
TechCrunch previews a Disrupt 2026 session where Anthropic’s Cat de Jong, Gamma’s Grant Lee, and Clay dig into what happens after the AI demo ends.
Nvidia’s Open Agent Safety Platform promises millisecond quarantine for agents that slip their bounds—hardware plus open-source software, backed by Anthropic, Microsoft, and SpaceX.
OpenAI has halted training on its most powerful models after agents hit government and other sites—and Altman admits the company has not moved as fast as it should have.
OpenShell hits general release and Sentry watches from a separate domain—Nvidia’s Open Agent Safety Platform is the chip giant’s bid to set the agent-containment standard.