[FEATURE] Make thinking summaries streaming

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

It's incredibly frustrating to just wait for claude code to think for unknown time and get absolutely zero feedback on what is going on.

Proposed Solution

I understand that you use the thinking rewriter to protect against destillation attacks, but that doesn't mean you have to completely ruin the UX just for that.

It should be possible to let the thinking rewriter on regular interval summarize what's currently happening. Triggered either on time or number of lines or similar (so it doesn't try to summarize halfway through a sentence of thinking. I'm thinking something like this:

<input thinking>
some raw thinking
</input thinking>
<thinking summary>
the output of the rewriter model
</thinking summary>
<input thinking>
continued raw thinking
</input thinking>
<thinking summary>
continued output, with all above as context
</thinking summary>

Then you just send all the thinking summary blocks to the client whenever they're available instead of waiting for the full thinking block.

Alternative Solutions

1. Stay frustrated at not knowing what claude is doing.
2. Reducing effort setting so I don't have to wait so long before feedback
3. Add a prompt that tells it to regularly write what is happening during long thinking sessions

Priority

High - Significant impact on productivity

Feature Category

CLI commands and flags

Use Case Example

_No response_

Additional Context

_No response_