Anthropic shipped a cluster of updates to Claude Code that change the tool's defaults for most existing users. Sonnet 5 is now the default model rather than an opt-in choice, Claude in Chrome (a browser-automation integration) reached general availability, and background subagents are now on by default, meaning Claude Code can spin off and run subagent tasks in the background rather than blocking the main session on every delegated piece of work. Claude Desktop for Linux also entered beta, filling a platform gap for developers who had been running Claude Code without a native desktop client on Linux machines. Separately, auto mode -- which lets Claude Code pick the right model tier automatically rather than requiring an explicit flag -- became the default on Bedrock, Vertex AI, and Microsoft Foundry, and Bedrock's available model was updated to Claude Opus 4.8. The team also fixed a longstanding annoyance where terminals would freeze during long streamed output, along with assorted worktree and plugin bugs, and changed the agent view to show colored state words and short classifier-written headlines instead of a raw feed of tool calls, plus automatic linking between a session and any pull request it touches. None of these individually is dramatic, but together they mark Claude Code shifting from a single-threaded assistant you babysit toward a tool that runs background work asynchronously and surfaces only the parts that need attention -- relevant for any team that runs many parallel Claude Code sessions or subagents, since default-model and default-mode changes affect cost and behavior across every existing script or CI job that did not pin a specific model.