Lenny Rachitsky · 3 min read · Design
🎙️ How I AI: How Meta designed Muse with Alex Cornell
Read the original on lennysnewsletter.com →
Alex Cornell, Design Director at Meta, shares how Muse was designed as an AI agent built around human conversation principles rather than traditional software interactions. Key design decisions include progressive disclosure for varying user needs, a single persistent chat relationship, and generative UI with fixed structure but dynamic content. Cornell also discusses personal use cases like turning Instagram saves into shopping catalogs and using AI for hobby analysis.
Key Takeaways
- AI agents should be designed around human conversation principles — Muse was built by asking 'How would this work if you were talking to another human?' rather than mimicking traditional software UI patterns.
- Progressive disclosure is critical in AI design: representing the same activity at multiple levels (emoji, task list, raw logs) lets different users access the transparency they need without creating clutter for others.
- Visible chain-of-thought reasoning is often a latency workaround, not a genuine feature — fast agents can simply deliver results without performing their thinking process.
- Generative UI works best with fixed structural surfaces and dynamic content, keeping experiences feeling alive without rebuilding the interface every session.
- Designing non-deterministic AI systems requires shifting from pixel-perfect craft to judging outputs by usefulness, which can be a liberating mindset change for designers.
Excerpt
How Meta designed Muse: an AI agent people actually want to talk to | Alex Cornell (Design Director) Listen now on YouTube • Spotify • Apple Podcasts Brought to you by: WorkOS —Make your app enterprise-ready, with SSO, SCIM, RBAC, and more Alex Cornell is a design director at Meta, where he leads design for Muse. In this episode, he explains how progressive disclosure makes Muse approachable for casual users without limiting technical ones, why its main chat is designed as a single persistent relationship, and how generative UI can feel personal without constantly changing the interface. He also shares…