Frontier research, read for the operator's question.
Working notes on papers from OpenAI, Anthropic, Transformer Circuits, and Stanford CGRI, read for what they imply about deploying agents and rebuilding workflows in real organizations.
Cross-Architecture Model Diffing with Crosscoders
Anthropic's crosscoders research examines differences between models that output evaluations may miss.
Emotions in Models: Interpretability Research
What emotion-like patterns inside models could mean for their behavior in a workflow.
Monitoring Internal Coding Agents for Misalignment
What OpenAI's internal coding-agent monitoring suggests about oversight in other organizations.
GPT as a Measurement Tool
How to use an LLM to measure qualitative work, and where calibration matters.
Graviton
What OpenAI's Graviton paper suggests about the cost of running agents in production.
Long-Running Claude
Agents that run for hours need different checks and handoffs from agents that answer a single prompt.
Anthropic's 81,000 Interviews
What teams can learn from Anthropic's large-scale user interview research.
Trust and Its Consequences
How trust affects the time and effort people spend coordinating work.