Preflight Checklist
What's Wrong?
When a response is actively being generated in a conversation, navigating to a different Project or thread in the left sidebar interrupts and discards the in-progress response. Upon returning to the original conversation, a message indicates that Claude's response was interrupted.
This makes it impossible to work across multiple Projects concurrently — a natural workflow given response latency. Users must remain on a single conversation until generation completes, effectively serializing all work.
What Should Happen?
Generation should continue in the background regardless of which conversation is currently in focus. The completed response should be present upon returning to the original conversation — consistent with how claude.ai behaves in separate browser tabs, where background tabs continue receiving streamed output.
Error Messages/Logs
Steps to Reproduce
- Open a conversation in any Project
- Submit a prompt that doesn't produce an immediate response and wait for generation to begin
- While generation is in progress, click on a different Project or conversation in the left sidebar
- Submit a message or simply view the second conversation
- Navigate back to the original conversation
Claude Model
Opus
Is this a regression?
Yes, this worked in a previous version
Last Working Version
No response
Claude Code Version
Claude 1.1.3830 (d2af56) 2026-02-20T15:45:56.000Z
Platform
Other
Operating System
macOS
Terminal/Shell
Other
Additional Information
Claude 1.1.3830 (d2af56) 2026-02-20T15:45:56.000Z
Mac OS 15.4 (24E248)
Preflight Checklist
What's Wrong?
When a response is actively being generated in a conversation, navigating to a different Project or thread in the left sidebar interrupts and discards the in-progress response. Upon returning to the original conversation, a message indicates that Claude's response was interrupted.
This makes it impossible to work across multiple Projects concurrently — a natural workflow given response latency. Users must remain on a single conversation until generation completes, effectively serializing all work.
What Should Happen?
Generation should continue in the background regardless of which conversation is currently in focus. The completed response should be present upon returning to the original conversation — consistent with how claude.ai behaves in separate browser tabs, where background tabs continue receiving streamed output.
Error Messages/Logs
Steps to Reproduce
Claude Model
Opus
Is this a regression?
Yes, this worked in a previous version
Last Working Version
No response
Claude Code Version
Claude 1.1.3830 (d2af56) 2026-02-20T15:45:56.000Z
Platform
Other
Operating System
macOS
Terminal/Shell
Other
Additional Information
Claude 1.1.3830 (d2af56) 2026-02-20T15:45:56.000Z
Mac OS 15.4 (24E248)