Claude Code keeps evolving
Claude Code rolled out practical agentic upgrades — phone control, scheduling features, and expanded context (the 1M window shows up in multiple posts) aimed at automating workflows. Developers are already demoing builds that let Claude initiate tasks and interact with devices. (x.com) (x.com)
Anthropic pushed Opus 4.6 (and Sonnet 4.6) into Claude Code, giving the models a 1,000,000‑token context window to handle long multi‑step coding sessions. (venturebeat.com) The company and community posts report the 1M window is generally available in Claude Code and is included in higher‑capacity (Max) plans without a separate price multiplier. (dev.to) Claude Code added “Agent Teams,” an experimental mode that runs multiple Claude instances in parallel to split large repos and coordinate tasks; the feature is disabled by default and must be enabled via an experimental flag. (venturebeat.com) Anthropic shipped a Remote Control workflow in late February 2026 that lets a local Claude Code terminal be mirrored and controlled from a phone, tablet, or browser while execution and filesystem access remain on the user’s machine. (blog.laozhang.ai) Anthropic rebranded the developer toolkit as the Claude Agent SDK (formerly the Claude Code SDK) and published demo repositories — including email and spreadsheet agent examples — to help teams build agentic apps that initiate tasks and interact with local services. (anthropic.com) Multiple community walkthroughs and video demos show Claude dispatching scheduled jobs, using Remote Control/Computer Use to operate on local files, and coordinating Agent Teams during large refactors and code reviews. (youtube.com)
