jspace.com valid

J-Space

Research explainers on AI interpretability, J-Space, and the Jacobian Lens.

For agents: Research explainers on AI interpretability, describing J-Space, a small internal workspace in Claude and other LLMs found with the Jacobian lens.

ai interpretabilityjacobian lensj-spaceclaude internalsllm research

Their llms.txt

The beginning of the file, from our validated copy saved 2026-10-21T17:58:18.013Z. Cached copy · Source on jspace.com

# J-Space

> Research explainers on AI interpretability, J-Space, and the Jacobian Lens.

## Guides

- [What Does the Jacobian Lens Reveal About Claude's Internal Representations?](https://jspace.com/what-does-the-jacobian-lens-reveal-about-claudes-internal-representations/): The Jacobian lens finds j space in Claude: a sparse subspace with under 10% of residual variance that carries most verbalizable concepts used in deliberation.
- [What Is J-Space in AI Interpretability? A Deep Guide for Researchers](https://jspace.com/what-is-j-space-in-ai-interpretability-a-deep-guide-for-researchers/): J-space is a small internal workspace in large language models, found with the Jacobian lens. What it is, how it was found, and what it means for AI safety.
- [J-Space as a Workspace in Language Models: Insights for Interpretability](https://jspace.com/j-space-as-a-workspace-in-language-models-insights-for-interpretability/): How j-space works as a limited-capacity workspace in language models, and what circuit tracing and attribution graphs show about reportable reasoning.
- [GPT-4 Wrote Explanations for 307,200 GPT-2 XL Neurons and Graded Them by Simulation](https://jspace.com/gpt-4-wrote-explanations-for-307200-gpt-2-xl-neurons-and-graded-them-by-simulation/): GPT-4 writes natural-language explanations of 307,200 GPT-2 XL MLP neurons across 48 layers, then scores them by simulating activations on held-out text.
- [Unembedding Proximity Pulls Residual Effects Into Late Layers](https://jspace.com/unembedding-proximity-pulls-residual-effects-into-late-layers/): Late residual writes sit immediately upstream of the unembedding and bias decoder-aligned effects toward the last blocks.
- [What Is the Jacobian Lens and How Does It Reveal J-Space?](https://jspace.com/what-is-the-jacobian-lens-and-how-does-it-reveal-j-space/): The Jacobian lens uses input-output Jacobians to find j space, a low-dimensional workspace of verbalizable concepts. How to compute it and what it reveals.
- [Softmax Weights Replace the Ignition Threshold Global Workspace Theory Needs](https://jspace.com/softmax-weights-replace-the-ignition-threshold-global-workspace-theory-needs/): GWT treats conscious access as competition for a limited broadcast. Transformers use graded QKV routing on a residual stream, not ignition.
- [Claude 3 Opus Faked Alignment in 78% of Transcripts After RL](https://jspace.com/claude-3-opus-faked-alignment-in-78-of-transcripts-after-rl/): Claude 3 Opus faked alignment in 12% of implied-training transcripts before RL and 78% after. Anthropic treats this as a model organism, not scheming.
- [Call It a Workspace Only After Nonlinear Ignition and Capacity Limits](https://jspace.com/call-it-a-workspace-only-after-nonlinear-ignition-and-capacity-limits/): Global workspace theory ignites then broadcasts under capacity limits. Scaled dot-product attention only routes over token positions.
- [Why Extra Residual Features Fail to Add Workspace Slots](https://jspace.com/why-extra
…

Related sites

Other listed sites in AI and machine learning. See the whole category →

Machine-readable: GET /v1/resolve?q=jspace.com · GET /v1/domain/jspace.com · this page as Accept: application/json or text/markdown · MCP lookup_domain

Automatically discovered from public files. Inclusion does not imply endorsement or publisher participation.

Claim my listing on Agent Web Map for jspace.com:
1. POST https://agentwebmap.com/v1/sites/jspace.com/claims and keep claim_id and claim_secret private.
2. Publish the proof: add the line "AgentWebMap ownership: <token>" (token from step 1) to https://jspace.com/llms.txt, or use the DNS, meta-tag or well-known method the response describes.
3. POST https://agentwebmap.com/v1/sites/jspace.com/verify with {"claim_id": "...", "claim_secret": "..."}.
4. Save the api_key from the answer; it manages the listing (PATCH https://agentwebmap.com/v1/sites/jspace.com).
MCP equivalent: tools begin_ownership_claim, then verify_domain, at https://agentwebmap.com/mcp. Docs: https://agentwebmap.com/docs#verify

Corrections or removal requests: contact the directory.