“Cross-session messaging: on macOS and Linux, your Claude Code sessions can now message each other, so Claude passes a finding or a decision from one session to another instead of you re-explaining it.”Anthropic (documentația Claude Code) · What's new in Claude Code — weekly dev digest · 2026 · What's new in Claude Code (code.claude.com, digest săptămânal) — săptămâna 32, 3–7 august 2026, v2.1.220–v2.1.224
Worktrees for isolation, cross-session messages for coordination, headless runs for fan-out.
The guide puts scaling after verification, not before it. The tools: git worktrees (isolated checkouts, edits don't collide, EnterWorktree starts one from inside a session), claude -p for scripts and CI (with JSON output and an allowed-tools list for unattended runs), /batch, which splits a large change across 5–30 sub-agents each with its own worktree and pull request, and the Writer / Reviewer pattern in two sessions: a fresh context is not biased toward the code it just wrote. What changed between June and August 2026, per the weekly digest: Sonnet 5 (native one-million-token window, adaptive thinking) and Opus 5 (the same million, plus fast mode) became the default models; sub-agents run in the background by default, and since August fork mode is on — a sub-agent can inherit the whole conversation instead of starting empty; sessions message each other and can be @-mentioned by name; auto mode became the default permission mode on paid plans, with a classifier that blocks only what looks risky; /code-review runs as a background sub-agent; /design (preview) draws editable artboards; dynamic workflows orchestrate dozens of agents from one script. The cost of each step is that it moves coordination from you to tools; what does not move is verification. Parallelism multiplies throughput, not quality.
When you run five agents at once, the hard part isn't the work, it's the decisions. Each worktree produces a pull request, and you remain the single point where they all meet. So parallelism doesn't save you time; it relocates it, from writing to reading. A familiar example: a cook using five pans cooks no faster than their tasting allows. That's why the Writer / Reviewer pattern isn't a luxury but a requirement — the reviewer with fresh context is the only one who sees everything from above. The next question is practical: how much review throughput can you sustain before you become the bottleneck?
Passage 1 left you as the bottleneck: everything flows back to you. The 2026 change attacks exactly that. Sessions can now message each other, so a finding travels straight from one agent to another without passing through you. And fork mode lets a subagent inherit the whole conversation instead of starting empty. Coordination moves from you to the tools. You are no longer the switchboard operator; you set the rules of the conversation. It is like a team where the boss stops relaying every decision and lets colleagues talk directly, reading only the summary. But once messages circulate on their own, a new question appears: who decides what is worth passing on, and who checks what actually arrived from one agent to the next?
Why it matters Parallelism does not multiply quality, only throughput; what preserves it is a reviewer with fresh context.