[BUG] Verbose mode not displaying thinking blocks in 2.1.42 (regression since 2.1.29)

Found 3 possible duplicate issues:

1. https://github.com/anthropics/claude-code/issues/22977
2. https://github.com/anthropics/claude-code/issues/24285
3. https://github.com/anthropics/claude-code/issues/22692

This issue will be automatically closed as a duplicate in 3 days.

• If your issue is a duplicate, please close it and 👍 the existing issue instead
• To prevent auto-closure, add a comment or 👎 this comment

🤖 Generated with Claude Code

Doesn't work in 2.1.70 either

Doesn't work in 2.1.70 either

@ashwin-ant @bcherny mess up thinking blocks AGAIN?

Data point: Thinking display WAS working in 2.1.71, broken again in 2.1.72

Contrary to this issue's timeline (broken since 2.1.29), I can confirm that thinking blocks WERE displaying correctly in 2.1.71 — as italic colored text in the terminal during interactive sessions. This display disappeared after upgrading to 2.1.72 (published 2026-03-09T23:29:16Z).

Additionally, I've discovered a more severe regression: thinking block content in session JSONL files is now empty in 2.1.72. The "type": "thinking" structure and "signature" field are present, but "thinking" is an empty string "". On 2.1.71, the same field contained full thinking text. I'm filing this as a separate issue since it affects session storage, not just display.

Empirical data (same project, same user, sequential sessions):

• Mar 8–9 sessions (2.1.71): 307 thinking blocks, 0 empty, 541,788 chars of content
• Mar 10 sessions (2.1.72): 112 thinking blocks, 99 empty (88%), only 17,702 chars

The transition from filled to empty happened between sessions on the version boundary.

I used to see thinking normally a month ago, but then it kind of disappeared mostly. I would see it randomly, very occasionally. That is still the case on 2.1.80.

I wish this could be fixed because the thinking was valuable to see if claude was on the right track before burning all those tokens...

and if it's bleeding too much proprietary information from the thinking techniques, maybe the 'thinking' output could be a summary of the actual internal thinking, more like "Insights" blurbs every so often.

2.1.84, not showing anything:

vova@Vovas-MacBook-Pro playgram % claude --verbose
 ▐▛███▜▌   Claude Code v2.1.84
▝▜█████▛▘  Opus 4.6 (1M context) with high effort · Claude Team
  ▘▘ ▝▝    ~/Documents/GitHub/playgramai

❯ think of something for a while

                                                                                                                                                              08:52 AM claude-opus-4-6
⏺ I thought about it for a while. Here's what came to mind:

[redacted]

  Showing detailed transcript · ctrl+o to toggle · ctrl+e to show all

Update: Here's one line that fixed it — my Claude Code had to scour the entire minified code to figure it out!

In settings.json:

{
  "showThinkingSummaries": true
}

Result:

vova@Vovas-MacBook-Pro playgram % claude
 ▐▛███▜▌   Claude Code v2.1.84
▝▜█████▛▘  Opus 4.6 (1M context) with high effort · Claude Team
  ▘▘ ▝▝    ~/Documents/GitHub/playgramai

❯ think of something

∴ Thinking…

  The user said "think of something" — this is very open-ended. They're probably just casually prompting me. I'll give a brief, useful response related to their project context.

⏺ Looking at your staged git changes, you're [etc. - redacted]

Still experiencing this on v2.1.9x (macOS, Anthropic API, Claude Opus). Verbose output is set to true via /config, but thinking blocks are not rendered — only the "Thinking" indicator icon is shown. The thinking content is confirmed to exist in the session JSONL file.

This has been open since February and spans many versions now. Would appreciate a fix or at least an update on the timeline. Thanks!