Claude Code says "No response after 60s — continued without an answer"? Here's what happened

You asked Claude Code to do something, it asked you a clarifying question, and you went to make a coffee. When you came back, the question was gone and your transcript had this in it:

● No response after 60s — continued without an answer

Claude didn't wait. It picked an answer for you and kept working.

What was actually happening

In early July 2026, Claude Code's AskUserQuestion tool — the dialog it uses to ask you things like "which of these three approaches do you want?" — quietly gained a 60-second timeout. If you didn't respond in time, the tool returned this to the model, verbatim:

"No response after 60s — the user may be away from keyboard. Proceed using your best judgment based on the context so far; you can re-ask this question later if it's still relevant."

So the model wasn't going rogue. It was doing exactly what the harness told it to do: assume you're away, and decide for itself. One user in the thread even asked Claude to explain the message, and Claude correctly reported there's no timeout parameter in the tool's schema — nothing you or the model could have set. The behaviour was baked in.

The part that stung most: it shipped silently. I went through the changelog — there is no entry announcing the timeout. The only changelog line that mentions this behaviour is the one that removed it.

Issue #73125 went up on 2 July and pulled 138 comments almost immediately, because plenty of workflows treat those questions as checkpoints. If your rule is "ask me before choosing the approach," a tool that self-answers after 60 seconds of silence isn't a convenience feature. It's your checkpoint deciding it doesn't need you.

What it didn't do

Worth being precise here, because the scarier version of this story travelled further than the true one. The timeout did not bypass the permission system. Bash approval prompts, deny rules, file-edit confirmations — all of that still worked normally. AskUserQuestion is the clarifying-question dialog, not the permission gate.

What it did do is arguably still bad enough: scope and approach decisions — "delete these or archive them?", "quick patch or proper refactor?" — got answered by "use your best judgment" instead of by you. Both things are true: not a permission bypass, and a real break in what people trusted the tool to do.

Anthropic's response was genuinely fast

Thariq Shihipar from the Claude Code team replied the same day the issue was filed:

"Hi all, my apologies. Will change this to be opt-in by default for now and experiment with making it an arg for the model to call, thanks for the feedback!"

And a few hours later:

"We've cut a release that should go out tomorrow morning PT — this will appear in /config and the timeout will be configurable, but will default to not being on."

The fix landed in v2.1.200, with a proper changelog entry this time: "Changed AskUserQuestion dialogs to no longer auto-continue by default; opt into an idle timeout via /config." I've confirmed the setting (askUserQuestionTimeout) exists in the current build. Reported 2 July, apology same day, fixed next morning, closed 4 July. Whatever you think of the feature shipping silently, the turnaround was excellent.

What to do

1. Check your version with claude --version. If you're below 2.1.200, update: claude --update or npm update -g @anthropic-ai/claude-code.
2. On current versions, questions block until you answer — the old behaviour is back as the default. Nothing to configure.
3. If you actually liked the auto-continue (it has real uses for long unattended runs), it's now opt-in: open /config and set the idle timeout deliberately.

What not to do

Skip the prompt-engineering arms race. When this hit, people started adding CLAUDE.md rules like "NEVER proceed without my answer" to fight the harness. That's patching the wrong layer — the model was handed an instruction saying you were away, and your prompt rules are competing with the tool result itself. A version update fixes it properly; a prompt rule just argues with it.

The transferable lesson is about silent behaviour changes. Claude Code ships multiple releases a week, and this one changed a safety-relevant default with no announcement. If your workflow depends on the tool blocking, waiting, or asking — know your version, and skim the changelog when behaviour shifts under you. The issue tracker usually knows before you do.

The receipts

• Issue #73125 — filed 2 July 2026, closed completed 4 July, 138 comments. The maintainer quotes above are in the thread; the verbatim timeout message is in the issue body.
• Changelog v2.1.200 carries the fix entry; no earlier entry introduced the behaviour.
• Verified against Claude Code v2.1.202 on 7 July 2026, including the askUserQuestionTimeout setting in the current binary.