BREAKING
Kimi Code adds autonomous /goal
The /goal autonomous loop
1
Plan
↓
2
Execute
↓
3
Revise
↓
4
Debug
0
T
total params
0
B
active params
0
K
context tokens
K2.7 Code vs K2.6 benchmarks
Kimi Code v2
62
Program Bench
53.6
MLS Lite
35.1
Praise and caveats
Praised
●
Autonomous long tasks
●
~30% fewer thinking tokens
●
Copilot integration
Caveats
●
HighSpeed unstable
●
Fixed thinking mode
●
Mostly internal benchmarks
Open-weight, widely available
AI NEWS BLITZ
Moonshot AI just added a self-running slash goal command to its Kimi Code agent.