In a given week you might execute a preventative maintenance sweep across a floor of conference rooms, work the support queue and drive a recurring failure to root cause, and then run technical coverage for an all-hands or an executive presentation that evening. This research continues many of the directions our team worked on prior to Anthropic, including: GPT-3, Circuit-Based Interpretability, Multimodal Neurons, Scaling Laws, AI & Compute, Concrete Problems in AI Safety, and Learning from Human Preferences.