---
title: 'In the News: November 4, 2026 (Morning)'
description: 'Claude Code 2.1.318 moves long subagent shell commands to the background after three minutes and adds a ten-day warning before a headless sign-in ends.'
canonical_url: 'https://darkfactory.dev/news/2026-11-04-morning'
markdown_url: 'https://darkfactory.dev/news/2026-11-04-morning.md'
collection: news
date_published: '2026-11-04T07:04:00-05:00'
date_modified: '2026-11-04T07:04:00-05:00'
---

# In the News: November 4, 2026 (Morning)


Claude Code 2.1.318, released November 4, changes three things that matter to unattended runs: when a subagent's shell command is moved to the background, how early a headless job is told its sign-in is ending, and how far an MCP deny rule reaches.

## 1. Claude Code 2.1.318 backgrounds long subagent commands and warns ten days before a headless sign-in ends

**[Claude Code changelog, version 2.1.318](https://code.claude.com/docs/en/changelog)** · Anthropic · Claude Code Docs, November 4, 2026

Background subagents now move a shell command to the background after 3 minutes on all providers. The release notes give the reason: the prompt cache should not expire while the command runs. They report no measurements for the change. For `claude -p`, the release adds a one-line stderr warning starting 10 days before a stored claude.ai sign-in ends, and exposes the end date as `loginExpiresAt` in `claude auth status` and as `credential_expires_at` in `claude -p` JSON results.

`deniedMcpServers` URL entries now also block Anthropic-provided servers whose upstream URL matches, and connectors in cloud sessions and on self-hosted runners. The notes say to deny a connector by name otherwise. The same release stops unattended runs (`claude -p`, SDK apps) from rewriting `~/.claude.json` when nothing in it changed, and makes that file, settings files and the stored login less likely to be deleted, briefly emptied or replaced by an older copy on shared network home directories.

**Why it matters:** If you run scheduled or headless Claude Code jobs, the sign-in end date is now something a script can read ahead of time. And a deny rule written as a URL now covers more than it did, so check your list before the update reaches a runner.

---

## Thread watch

_Discussions gathering force. No primary read yet, so these are reported as discussions, not as findings._

- **[Ask HN: OpenAI seems to be silently re-enabling model training on Codex chats](https://news.ycombinator.com/item?id=50352757)** · Hacker News · 488 points and 148 comments at about 21 hours (observed 7:05 AM EST, November 4). The original poster says coworkers found the data-sharing setting in Codex switched back on after they had turned it off, and one commenter reports the same on their own account. We have not read an OpenAI statement or any settings history, so this is a discussion, not a finding.