Prompted: AI, People and the Creative Spark The Hidden Design Choices Behind Claude with Joel Lewenstein, Head of Product Design at Anthropic
Dec 2, 2025
Joel Lewenstein, Head of Product Design at Anthropic, shares insights on designing AI experiences that foster trust and collaboration. He discusses how Claude serves as a 'trusted first reader,' helping him refine ideas and enhance team creativity. Joel highlights an intriguing 'vending machine' experiment to study human-AI interactions and explains how friction in processes can boost collaboration. He also envisions a future where AI becomes an integral team member, emphasizing the importance of effective prompting and innovative user interfaces.
AI Snips
Chapters
Transcript
Episode notes
Childhood Game Sparked Curiosity
- Joel remembers playing the MS-DOS game Gorillas and tinkering with its source code as a kid.
- That early experience sparked his fascination with how code shapes interactive experiences.
Use Claude As Your First Reader
- Word-vomit into Claude as raw material and then let it tighten your draft before sharing with others.
- Use Claude as a first reader to save time and produce clearer first drafts.
Design AI As A Trustworthy Entity
- Design Claude's behavior as an entity with traits like honesty, openness, and transparency.
- Anthropic treats those character choices primarily as safety and alignment decisions.