When Brain is processing a request, the input is locked until the response completes. If I realize mid-generation that I forgot context, want to redirect the approach, or need to clarify something, I have to wait for the full response, then correct course in a follow-up, which wastes both my time and the context window.

Requested behavior: Let me send a new message while Brain is actively generating. The new message should be treated as real-time steering input (additional context, a correction, or a redirect) that Brain incorporates into its current response or gracefully pivots from.

Why it matters:

Thinking evolves faster than responses generate. By the time Brain finishes, I've already spotted the gap.

Longer multi-tool responses (task creation, research, bulk updates) take 30-60+ seconds. That's a lot of dead time when I already know it needs a course correction.

Current workaround is stopping generation and re-prompting from scratch, which loses all progress.

Example use cases:

Brain starts creating a task in the wrong list; I want to say "actually put that in X" before it finishes

I forgot to mention a constraint and can see Brain heading down the wrong path

I want to add "also do Y" while it's still working on X