Source Intelligence
Follow Discord

DisclaimerUnofficial, and not affiliated with Anthropic. Nearly all of this is read straight out of what ships: npm bundles, captured prompts, published docs. Anthropic's own notes go in verbatim, marked as theirs. The rest is my reading, and every entry carries the strings behind it. If one looks wrong, vote it down and say why.

← All of v2.0.0 Home All releases ←olderv1.0.128 v2.0.1newer→
Claude Code v2.0.0

Extended Thinking Improvements

What: Simplified thinking trigger system with new programmatic control

How to use:

# Use "ultrathink" anywhere in your prompt for maximum thinking
claude
> Please ultrathink about this architectural design

# Toggle thinking with tab key during interactive sessions
# Or use /t command

Details:

  • Streamlined from multilingual pattern matching to single "ultrathink" keyword
  • Binary mode: ULTRATHINK (31,999 tokens) or NONE
  • New thinkingMetadata API for programmatic control without text triggers
  • Removed complex multi-level system (HIGHEST/MIDDLE/BASIC) in favor of simpler on/off
  • Thinking state notifications show "Thinking on/off (tab to toggle)"
  • Evidence: Pattern at line 363628, functions Y61() at line 363667, Y09() at line 363644, le1() at line 363671, metadata handling at line 363649 in v2.0.0

See this entry in the whole of v2.0.0 →