#mechanistic interpretability
47 related articles
·6 min
Jacobian-Lens in Practice: Manually Reshaping LLM Behavior with J-Space
After Anthropic released Jacobian-Lens, a developer reversed it from an interpretability tool into a behavior editor, manually tuning J-Space to reshape LLM outputs. An in-depth look at the tech, representation engineering, and AI safety risks.
Read more →
·5 min
Zero-Dependency AI Decision Record Auditor: A Practical Guide to Building Traceable AI Systems
An in-depth analysis of the zero-dependency decision record auditor: from AI compliance and incident postmortems to human-AI accountability, exploring how 'Governance as Code' enables traceable, transparent AI decision-making.
Read more →
·5 min
GPT-5.6 Comes to Codex, Voice Models Iterate: A Flurry of AI Product Updates
OpenAI announces GPT-5.6 Sol Ultra coming to Codex and its most powerful real-time voice model yet, GPT Realtime 2.1; Tencent's Toast lands on iOS; Anthropic finds brain-like structures in Claude.
Read more →
·6 min
GPT-5.6 Officially Approved for Release: How National Security Review Is Reshaping the AI Launch Process
OpenAI's flagship GPT-5.6 was delayed by national security review before winning U.S. government approval. An in-depth look at the Sol, Terra, and Luna model lineup and the emerging AI regulatory regime.
Read more →
·3 min
GPT-5.6 Released: Sol, Terra, and Luna Reshape AI Programming and Office Automation
OpenAI launches GPT-5.6 with three models — Sol, Terra, and Luna — plus ChatGPT Work, a new desktop app, and Hosted Sites. Codex now autonomously trains models.
Read more →
·3 min
Visualize and Edit AI Reasoning Chains: Making LLM Thinking Transparent and Controllable
A new web tool visualizes and lets users edit AI chain-of-thought reasoning. Explore how it works, its use cases, and the limits of LLM interpretability.
Read more →
·2 min
COLM Conference Deep Dive: Positioning, Academic Ecosystem, and What It Reveals About AI Research
COLM (Conference on Language Modeling) is the premier vertical venue dedicated to LLM research. This deep dive covers COLM's positioning, academic ecosystem, and what it reveals about AI research trends.
Read more →
LLM Security Benchmarking: Current Sta…
·4 min
LLM Security Benchmarking: Current State, Challenges, and Practical Guide
Why is it so hard to establish unified LLM security benchmarks? This article analyzes core challenges in LLM security evaluation—covering jailbreaks, prompt injection, red teaming, and more—with practical strategies for developers.
Read more →
·6 min
AMD Ryzen AI Halo Dev Kit: A $4,000 Local LLM Powerhouse Taking on Apple and NVIDIA
AMD officially unveils the Ryzen AI Halo local AI dev kit, priced around $4,000 with 128GB unified memory, capable of running 70B LLMs locally. An in-depth look at its specs, pricing, and market competition.
Read more →
·7 min
Spec-Driven Development: Andrew Ng Teaches You to Master AI Coding Agents Efficiently
Andrew Ng partners with JetBrains on a new course systematically teaching Spec-Driven Development. By writing high-quality specs, developers can precisely control AI coding agents, eliminate context decay, and boost intent fidelity.
Read more →
·7 min
Block-Sparse Featurizers: Opening the Black Box of Vision Model Neural Geometry
Block-sparse featurizers remap dense vision model activations into block-sparse representations, making the internal feature spaces of ViT, CNN, and other models readable and interpretable. This article explores their core principles, links to mechanistic interpretability, and applications.
Read more →
·6 min
Fable 5 vs. GPT-5.6 Sol: Benchmarks, Pricing, and the AI Power Struggle Explained
An in-depth comparison of Fable 5 and GPT-5.6 Sol: benchmarks across Terminal Bench, HealthBench, and ExploitBench, plus pricing strategy, OpenAI's government equity controversy, and shifting AI power dynamics.
Read more →
·2 min
LLM Thought Visualization: Making AI Reasoning Transparent and Interpretable
LLM thought visualization is emerging as a key breakthrough in AI explainability. This article explores the value, technical approaches, and challenges of visualizing Chain-of-Thought reasoning.
Read more →
·3 min
Global Workspace Theory in Language Models: A New Lens for AI Interpretability
Can Global Workspace Theory (GWT) explain the internal mechanisms of large language models? This article explores how residual streams and attention in Transformers map to cognitive science's 'information broadcast' framework.
Read more →
·6 min
Reverse-Engineering ChatGPT's Source Selection: What Network Traffic Analysis Reveals About AI Citations
By capturing and analyzing ChatGPT's real network traffic during web retrieval, this article reveals the underlying logic of AI source selection—covering the layered mechanisms of retrieval recall and content filtering, domain preference patterns, and practical insights for content creators and SEO professionals.
Read more →
·2 min
Training Just One Transformer Layer Can Match Full-Parameter RL Training?
A controversial study shows training just one Transformer layer can match full-parameter RL training. We analyze the technical principles, engineering value, and limitations of this approach.
Read more →
·6 min
AI Superforecasters: Can Machines Surpass Human Experts at Prediction?
Have AI superforecasters truly arrived? A deep dive into how LLMs challenge human superforecasters in probability calibration, information integration, and scalable forecasting, plus core debates on data leakage, interpretability, and real-world applications.
Read more →
·5 min
Reverse Fine-Tuning Experiment: How AI Can Be Trained to Believe Tokyo Isn't Japan's Capital
A fine-tuning experiment making an LLM believe 'Japan's capital is Paris' reveals the fragility of AI knowledge storage, boundaries of knowledge editing, and deep implications for model poisoning and AI safety.
Read more →
·4 min
Why Large Language Models Are Starting to 'Invent Words': The Nature of LLM Language Boundaries
Top LLMs are pushing beyond existing human vocabulary, producing neologisms and expressive distortion. This article analyzes the tension between LLM high-dimensional semantic spaces and natural language symbol systems.
Read more →
The Rise of Autonomous AI Research: Ef…
·2 min
The Rise of Autonomous AI Research: Efficiency Gains vs. the Battle for Human Control
At AIEWF, the vision of autonomous AI research sparked fierce debate. Can AI complete a full research loop independently? Experts defend human understanding and control, revealing the core tension between automation efficiency and human agency.
Read more →