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. Own incident communications end-to-end, from real-time internal coordination to external channels like status pages, direct customer outreach, and stakeholder updates, ensuring they reflect Anthropic's commitments to safety, transparency, and accuracy.