Claude Code’s New Update Lets Subagents Run Entirely In The Background
By 813 Staff
The latest development in AI and tech shows Claude Code’s New Update Lets Subagents Run Entirely In The Background, according to Boris Cherny (@bcherny) (on June 29, 2026).
Source: https://x.com/bcherny/status/2071647677591466098
Two weeks ago, internal Anthropic documents showed Claude Code was handling roughly 12 percent of all active terminal sessions on developer machines where it was installed. That number is about to climb sharply. Boris Cherny, a senior engineer at Anthropic, posted on June 29 that the next version of Claude Code will let subagents run in the background. Engineers close to the project say this means the AI coding assistant will be able to spawn child processes that persist after a user dismisses the main CLI session—effectively turning the tool from a single-task assistant into a persistent, multitasking background daemon.
The rollout has been anything but smooth. Earlier this month, users reported that Claude Code’s background polling for file changes was causing significant CPU drain on MacBooks, and internal slack logs leaked showing the team was “still tuning the wake-up intervals.” Those tuning issues appear to be resolved. According to Cherny’s post, the background subagents will now handle long-running jobs like test suites, code reviews, and dependency auditing without blocking the primary workspace. The key detail: these agents can write to disk, commit to git, and even trigger CI pipelines independently, though Anthropic has not yet confirmed whether they will have full network access or be sandboxed to the local filesystem.
This matters because it fundamentally shifts what a coding assistant can be. Until now, every major AI coding tool—GitHub Copilot, Cursor, Amazon Q—has operated on a prompt-and-response model. You ask, it answers, you’re done. Background subagents turn Claude Code into something closer to an automated junior developer: it can begin a refactor, get interrupted, and resume the moment you dismiss the main chat. The impact for teams shipping daily will be noticeable: fewer context switches, less time waiting for dependency installs or test reruns.
What happens next is the interesting part. Anthropic has not given a public release date for this update, though the x post suggests it is feature-complete and in internal testing. Competitors are already watching. Engineers at GitHub have been seen probing similar background process APIs in VS Code extension source code, but nobody has shipped yet. If Cherny’s timeline holds, Claude Code will be the first major AI development tool to blur the line between assistant and automatic background worker. That is a line that, once crossed, is hard to uncross.
