Restore thinking output

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

Summary: A recent update hid the visible extended thinking output and replaced it with a token counter. Please restore the streamed thinking output.

Why the extended thinking output mattered

I relied on the visible extended thinking output to follow Claude's reasoning as it streamed. Seeing the reasoning gave me early signals about what Claude intended to do, which let me:

• Respond to permission prompts with confidence — I could judge whether Claude's planned action was sound before accepting or rejecting it.
• Save time — if Claude was heading in the wrong direction, I could cancel early instead of waiting for it to finish and return something unusable. It was a valuable tool for timely course correction.

A token counter provides no meaningful value other than giving you PTSD. It shows that Claude is working, but not what it's doing. Now I wait with no visibility, and too often the result is wrong, especially with short, one off prompts. The thinking output was the single most useful signal I had for steering toward a better outcome.

Proposed Solution

Restore the streamed thinking output, make it even more effective and prominent by adding relevant features. I don't want to work in a black box.

Alternative Solutions

_No response_

Priority

Critical - Blocking my work

Feature Category

Performance and speed

Use Case Example

_No response_

Additional Context

Related issue

#30660 requests the same feature for interactive mode, but I don't want this tied to any specific mode or setting. The thinking output should be visible on its own, but it may be turned off via a dedicated setting.