BREAKING
Anthropic finds 'J-space' in Claude
0+
pages
0x
GitHub release
How the Jacobian Lens works
1Scan vocabulary
2Read Jacobian
3Extract J-space
Four key properties of J-space
Reportableverbalize
Model can state its thoughts
Modulable by instruction
Selectivelimited
Mediates multi-step reasoning
Only part of processing
Access vs subjective experience
Tool aimed at alignment audits
AI NEWS BLITZ
Anthropic says it discovered a hidden inner-thought space inside its model Claude.