Claude Code Thinking Summaries Now Render as Markdown and Stay Visible for 3 Seconds
Claude CodeClaude Code v2.1.152 ships three connected thinking-display improvements: collapsed thinking summaries now render as markdown, stay on screen for a minimum of 3 seconds before collapsing, and are capped at 10 lines with Ctrl+O to expand the full transcript.
Key Takeaways
- Thinking summaries now render as markdown, making code snippets, bold text, and structured output readable in the collapsed group view.
- A 3-second minimum display time prevents fast-thinking steps from flashing and disappearing before they can be read.
- Collapsed summaries cap at 10 lines, keeping the UI clean without losing information — Ctrl+O shows the full thinking output.
- The fullscreen thinking counter now counts up live, giving a real-time sense of reasoning duration rather than a static snapshot.
- When thinking is interrupted, the counter preserves its last value, showing how much reasoning had occurred before the interruption.
- These improvements are most impactful for extended-thinking workflows where understanding the model's reasoning process is as important as the final output.
Sources & Mentions
3 external resources covering this update
Thinking Summaries Get a Readability Overhaul
For developers using Claude Code with extended thinking enabled, v2.1.152 delivers a trio of improvements that make thinking output meaningfully easier to follow during long sessions.
Markdown Rendering in Collapsed Summaries
Previously, thinking summaries shown in the collapsed group appeared as raw plain text. Code snippets, formatted output, and structured reasoning were flattened into a single unformatted stream. With this update, collapsed summaries render full GitHub-Flavored Markdown — inline code, bold, headers, and lists all display correctly. This is particularly valuable when the model's thinking includes pseudocode, file paths, or structured plans.
The 3-Second Minimum Display Rule
Fast operations — quick file reads, short computations, brief API calls — would previously produce thinking summaries that flashed on screen and immediately collapsed, leaving no time to read them. The new 3-second minimum ensures every thinking step remains readable long enough to scan. For slower operations, the summary stays visible as long as the step takes.
10-Line Cap with Ctrl+O Expansion
Long thinking sequences can now produce verbose summaries that would otherwise overwhelm the collapsed view. The new 10-line cap keeps the display compact during active work, while pressing Ctrl+O opens the full thinking transcript at any time — a shortcut that was already available for the full conversation log and now extends naturally to the thinking detail view.
Bonus: Live Thinking Counter in Fullscreen Mode
A related improvement in the same release: the "Thinking for Ns" indicator in fullscreen mode now counts up live while the model is thinking, rather than showing a static value. If thinking is interrupted mid-stream, the counter preserves its last value instead of resetting — giving a clearer picture of how much reasoning effort a task consumed.
These changes collectively make thinking output a practical tool for understanding model behavior during complex multi-step tasks, rather than a fast-moving overlay that's too brief to act on.