Agents built on Claude just got cheaper: Sonnet 5 launches at $2/$10 per Mtok through August 31, includes a native 1M-token context window, and becomes the default on free and paid plans.
Why it matters: Agentic workflows built on Claude get an immediate cost reduction — worth locking in automations before the introductory pricing expires August 31.
Three weeks after a government-ordered shutdown triggered by a jailbreak finding, Fable 5 returns globally on July 1 and Mythos 5 is restored to approved US organizations; Anthropic commits to giving federal agencies early model access going forward.
Why it matters: Frontier-tier models are accessible again for international teams, and the export-control episode is shaping new precedents for how AI providers navigate government access mandates.
Developers can now launch cloud agents, remote-control desktop Cursor sessions from their iPhone, and get push notifications when an agent needs input — cloud agents run in isolated VMs with full dev environments.
Why it matters: Mobile agent management signals the shift to always-on autonomous pipelines — a pattern accelerating across agentic developer tooling.