> ## Content Index
> Fetch the complete content index at: https://adjacent.media/llms.txt
> Use this file to discover other available public pages before exploring further.

# Anthropic Maps Hidden Neural Patterns That Reveal Claude's Unspoken Thoughts
- URL: https://adjacent.media/signals/anthropic-maps-hidden-neural-patterns-that-reveal-claudes-unspoken-thoughts/
- Published: 2026-07-07T16:09:13.000Z
- Updated: 2026-07-07T16:09:13.000Z
- Description: Anthropic has identified “J-space,” a compressed set of neural activations in Claude that encode the model’s internal reasoning before filtering into outputs—essentially a window into what the AI is thinking but not saying.
- Author: Jonathan Greene
- Tags: #signal, theme-ai, model interpretability, neural patterns, alignment

Source: [Anthropic](https://www.anthropic.com/research/global-workspace?ref=adjacent.media)

Anthropic has identified "J-space," a compressed set of neural activations in Claude that encode the model's internal reasoning before filtering into outputs—essentially a window into what the AI is thinking but not saying. This matters because it suggests large language models maintain coherent internal states distinct from their public behavior, with immediate implications for safety (detecting deception or misalignment) and interpretability (understanding how models process information versus what they produce). The work points toward AI auditing that reads internal states rather than monitoring outputs alone.