How Anthropic found a hidden space inside Claude where it thinks before speaking
anthropic used a new tool called j-lens to uncover a hidden layer in claude that reveals unspoken thoughts. this interpretability breakthrough exposes the model's internal processing before it generates a final response.
anthropic has identified a specific internal region within its claude model, dubbed j-space, where the model processes concepts before generating a final output. by developing the j-lens tool, researchers observed this layer contains working words that reveal what the model is 'thinking' before it speaks. this provides a rare, concrete look into the 'black box' of neural networks, moving beyond abstract weights to observable intermediate states. understanding these layers is critical for ai safety, as it allows researchers to detect and potentially correct harmful reasoning processes before they reach the user.