[Desktop] Short notes between tool calls are folded into "Ran N commands", while longer ones show only as a rewritten summary

Summary

In the Code tab, the notes Claude writes between tool calls end up in one of two places, and neither shows the user what Claude wrote:

• A note of one or two sentences arrives as a text block and is folded into the "Ran N commands" group. The user has to open the group to read it.
• A longer note arrives as a progress-update thinking block (documented for Claude Opus 5.5 and Claude Fable 5.1: "notes longer than a sentence or two come back as progress-update thinking blocks"). The app shows it in the chat, but as a rewritten summary, not as written.

So the only notes that stay visible are the ones the user never sees verbatim. This is #90425 (still open), plus the progress-update path that has appeared since.

Environment

• Claude desktop app 2.19675.0 on macOS, Code tab, bundled Claude Code 2.1.286
• Model: claude-opus-5-5, effort high
• Transcript view: Normal (Thinking and Verbose don't change it)

Evidence from one session's .jsonl

| Shown to the user | Block type in the .jsonl |
|---|---|
| In the chat, as a paraphrase | thinking (progress update) |
| Folded inside "Ran N commands" | text, e.g. a one-sentence note before a web fetch |
| Folded inside "Ran N commands" | text, a one-sentence note before a status check |

While the turn runs, the folded notes appear as rows inside the expanded group, mixed with tool rows. When the user asks a question mid-turn, the answer can land in a folded row.

Expected

Render text blocks written between tool calls in the chat, outside the tool group, as before. Fold only the tool calls.

Why it matters

Users who don't read the tool rows (the group exists so they don't have to) miss what Claude found and what it's about to do. Writing longer notes to get a visible summary trades the user's exact words (commands, numbers, file names) for a paraphrase.

Related: #90425, #94354, #97504, #96288.