[AINews] not much happened today
Anthropic, Google, and others are standardizing persistent, asynchronous agent runtimes as the default UX.
“one conversation, many cloud threads”— Aravind Srinivas
A roundup of agent-infrastructure moves shows Anthropic's Claude Code Projects, Google's Antigravity-based Gemini harness with Credentials and Files APIs, and constrained-output classifier models like TypeSafe's Jev all converging on persistent, scoped-permission, asynchronous agents. This signals that multi-session orchestration and structured-decision layers—not chat-plus-tools—are becoming the default architecture for production AI systems.