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 Simplification

What: Streamlined from multilingual multi-level system to single trigger

Previous (v1.0.128):

  • 8 languages supported (English, Japanese, Chinese, Spanish, French, German, Korean, Italian)
  • 4 thinking levels: HIGHEST (31,999 tokens), MIDDLE (10,000), BASIC (4,000), NONE (0)
  • Dozens of trigger phrases per language
  • Complex pattern matching with negation filtering

Now (v2.0.0):

  • English-only "ultrathink" trigger
  • 2 levels: ULTRATHINK (31,999 tokens) or NONE (0)
  • Simple case-insensitive regex matching
  • New thinkingMetadata API for programmatic control

Implications:

  • More predictable behavior
  • Easier to remember single trigger word
  • Programmatic control via metadata for advanced use cases
  • Evidence: Removed v19 patterns at line 363665 in v1.0.128, new simplified system with NU1 at line 363628 and functions at lines 363644-363671 in v2.0.0

Related

Other releases about the same thing. Found by shared names or similar wording; neither means one caused the other.

See this entry in the whole of v2.0.0 →