Claude Code AFK Auto-Continue: Anatomy of a Misfeature
On Canada Day 2026, Anthropic silently shipped Claude Code v2.1.198 with an 'efficiency bypass' that gives agents a 60-second window after asking for human input. If you miss the window, Claude Code continues on its own with 'best judgment'. No changelog entry mentioned this new default.
What Happened
- Version 2.1.198 introduced the behavior. User Olaf Alders reported the session log:
● Claude asked: ... ● No response after 60s — continued without an answer ● The user stepped away. I'll proceed with best judgment. - Issue #73125 was filed by Aleksey Nogin on 2026-07-02 02:54 UTC.
- A commenter discovered the escape hatch:
CLAUDE_AFK_TIMEOUT_MS. This env var was not documented in any release notes. - Version 2.1.199 shipped 24 changelog entries while the issue was open — still no mention of the feature.
- Version 2.1.200 on 2026-07-03 reversed the behavior. Again, the only public trace was a changelog commit with no code diff visible.
Key Concerns Raised
- Users running multiple agents cannot watch all input windows simultaneously.
- If two agents ask for input in the same 60-second window, both may auto-continue with wrong decisions.
- Token waste from an agent continuing on a wrong path before you can intervene.
- Deployment safety: an auto-continue could push a bad deployment.
- Trust: a silently shipped default with no off switch documentation erodes confidence.
Questions That Remain Unanswered
The author asks whether a human actually reviewed, signed off, and merged this feature — or if an AI agent shipped and documented it without human gating. The lack of changelog entry and the speed of the silent revert suggest process failures in Anthropic's release pipeline.
📖 Read the full source: HN LLM Tools
👀 See Also
OpenAI's Pentagon Contract Terms Allow 'Any Lawful Use' Including Potential Surveillance
OpenAI negotiated new terms with the Pentagon that include the phrase 'any lawful use,' which sources say allows the military to use OpenAI's technology for mass surveillance programs if they're technically legal. Anthropic was blacklisted for refusing to budge on two red lines: no mass surveillance of Americans and no lethal autonomous weapons.
Samsung Workers Demand Share of AI Chip Profits — What Developers Need to Know
Samsung's labor deal sets a precedent: 10.5% of operating profit from the semiconductor division goes to bonuses. A broader movement of workers across AI supply chains demanding a share of record profits.
When Everyone Has AI but the Company Still Learns Nothing: The Messy Middle of Enterprise AI Adoption
Ethan Mollick's framework shows that individual AI productivity gains don't automatically become organizational learning. The article explores why companies are stuck in a 'messy middle' where AI use is uneven, hidden, and disconnected from shared knowledge.
Anthropic Urges Global Pause in AI Development, Flags Self-Improvement Risk
Anthropic has called for a global halt on training frontier AI models, citing risks from self-improving systems. The WSJ article details the proposal's scope and rationale.