BREAKING
Claude Code Adds Effort Levels
/code-review Rewrites Its Strategy
From Low to Ultra
1
Low effort
↓
2
High recall
↓
3
Ultra: multi-agent
Trade Cost Against Depth
Low effort
●
Fewer, high-confidence findings
●
Fraction of the tokens
●
May ask for clarification
High / Ultra
●
Higher recall
●
Cuts false positives
●
Higher spend
0
min
review turnaround
Local Path Now Broadly Available
AI NEWS BLITZ
Anthropic is letting developers dial code-review depth up or down inside Claude Code.