
Early days of Anthropic

Why AI is going vertical (again) | Dianne Penn (Anthropic)
Topic 1 of 18
Published Jul 26, 2026 · Clip 2:36–7:46 (5:10)
Diane describes joining Anthropic in 2023 when the product org was just five engineers, including one engineer for the entire API business, with a strong mission-driven culture and startup energy while the company was still finding its identity. The early years were about exploring how the technology could bring value to users and society, starting with claude.ai as a chat assistant and evolving into things like tool use. She recounts the Golden Gate Claude experiment: early interpretability research identified a 'Golden Gate Bridge' feature, and dialing it up made Claude obsess over the bridge in every response. The team spun up the whole experience on claude.ai within 24 hours, reaching roughly 2,000 people, but it was a hidden inflection point proving Anthropic could ship distinctive user experiences at startup pace with a bottoms-up culture that persists today.