Quick Answer Discovered using the jacobian lens by Anthropic in July 2026, j-space is a small internal workspace inside large language models. It consists of roughly 25 [Anthropic] active j lens vectors at any time that together explain less than…
J-Space Anthropic explained: Language models such as Claude develop j-space, a limited set of verbalizable representations that acts as an internal global workspace. This workspace can be monitored with the j lens before outputs are generated. The research identifies functional…
On July 6, 2026, Anthropic published a paper that changed how we think about what language models know but never say aloud. The jacobian lens—named after the Jacobian matrix in calculus—gives researchers a tool to read the silent, internal neural…
