
How Meta designed Muse: an AI agent people actually want to talk to | Alex Cornell (Design Director)
- 37m
- Transcript
Summary
Written by a model from the transcript; not yet reviewed.
This episode of How I AI features Alex Cornell, Design Director on Meta's Muse, discussing the design philosophy behind building a consumer-facing AI agent. The conversation explores how Meta approached agent interfaces by treating them like personal messaging entities rather than traditional coding tools, utilizing familiar primitives like avatars, progressive disclosure, and status messages to build consumer trust and reduce perceived latency. Cornell shares specific use cases and demos of how he uses his own agent, Marlo, to manage personal hobbies, organize data from Instagram and Pinterest, and transform old archives into fun customized formats.
The discussion provides deep insights into the challenges of designing non-deterministic AI software where every user experience differs. Cornell and the host examine architectural choices such as single main chats versus side threads, the role of generative UI elements, and how to prompt effectively when agents go off track. The episode is especially valuable for product designers and engineers looking to make AI applications more accessible, intuitive, and delightful for everyday users.
- Design AI agent interfaces around messaging patterns and human-to-human exchanges rather than complex software-centric concepts to make them immediately accessible to consumers. 5:00
- Implement progressive disclosure for agent status and chain of thought reasoning so that normal users get a simple indicator while power users can inspect the underlying code. 7:42
- Connect rich personal data sources like Instagram saves or past blogs to your agent to automatically organize, summarize, and render your personal history into novel formats. 28:12
- Use your agent to monitor and parse complex hobby data, such as golf shots or chess matches, converting plain numbers and scorecards into rich visual recaps and targeted improvement goals. 22:30
Chapters
- 0:00Podcast Intro and Host Greeting
Claire Vo introduces the episode, the sponsor WorkOS, and welcomes Alex Cornell.
- 1:01Designing Muse: Core Design Principles
Alex explains Muse’s design focus on conversational flow, status messaging, and progressive disclosure.
- 2:55User Experience: Status and Progress Feedback
Discussion of how Muse communicates progress to users, balancing detail for engineers and simplicity for consumers.
- 11:09Generative UI and Personalization
Exploration of Muse’s generative UI elements like avatars, status emojis, and personalized idea feeds.
- 14:25Use Cases: Shopping, Instagram, and Personal Hobbies
Alex showcases how Muse organizes Instagram saves, suggests shopping ideas, and supports hobbies like golf and chess.
- 25:36Architectural Choices: Main Chat vs. Side Chats
Explanation of Muse’s chat architecture, the role of side chats, and future directions for agentic interactions.
- 32:05Designing for Non-Deterministic Systems
Alex reflects on learning to design for AI’s unpredictable behavior and how it differs from traditional deterministic UI design.
- 36:34Closing Remarks and Next Steps
Final thoughts, encouragement to try Muse, and where to find more information and community updates.
Highlights
- 4:37
Designing Agents Like Messaging Appsinsight
The guest explains how framing the AI agent interface as a messaging exchange with an embodied entity clarified key design decisions, such as keeping chat bubbles and proactive messaging features.
- 6:40
Mom's First Software Experiencestory
The guest shares a story about his mother texting him to say she apologized to her Muse for working it too hard, proving the agent successfully bridged accessibility gaps for non-technical users.
- 7:42
Layered Progressive Disclosurestrategy
The guest details how the UI reveals status progressively, showing simple indicators to average users while hiding complex chain-of-thought code unless specifically requested by an engineer.
- 9:42
Fast Models Reduce Latencyinsight
The host points out that Muse's fast underlying model makes waiting for agent actions feel natural and non-painful, reducing the need for distracting reasoning UI elements.
- 10:11
Acknowledging Complex Requestsstrategy
The guest explains that when an agent handles difficult tasks, it should immediately acknowledge the request with a reaction or an initial message so it never feels like it has ghosted the user.
- 11:19
Personalized Generative UIinsight
The host highlights how Muse uses generative UI by pairing consistent navigation primitives with dynamically generated ideas, status emojis, and animated avatars that match the user's custom character.
- 16:23
Rendering Instagram Saves as Skymallstory
The guest demonstrates how he had his agent parse his Instagram saves and render shopping suggestions in the nostalgic vernacular of a 90s Skymall Sharper Image magazine.
- 20:10
Reasoning with Visual Datainsight
The host describes using agent connections to visual platforms like Instagram and Pinterest to generate design concepts and unlock creative use cases from hard-to-reason-with personal records.
Quotes
“And I think like simple things like the fact that an agent can message you proactively changes the nature of like how you think about the product from a design perspective, even in like super simple ways.”
Unnamed speaker · 3:51
Sponsored segments
- WorkOS 1:01–2:48
About this episode
Alex Cornell is a design director at Meta, where he leads design on Muse, Meta’s consumer AI agent. He came to the role with a long background in software product design, including time as a founder, and has spent the past year building one of the most thoughtful agent interfaces in the consumer AI space. Muse ships with a custom avatar, a three-layer status display, a personalized ideas surface, and a generative feed. Alex has been sharing the reasoning behind each of those design decisions publicly on Threads. What you’ll learn: Why Alex built Muse to feel like messaging, not an AI appHow Muse’s three-layer status system accommodates different levels of technical expertiseWhat generative UI actually means in a shipped consumer productHow Alex turned years of forgotten Instagram saves into a daily shopping catalogWhy his mom apologized to her Muse, and what it means for AI designHow he uses Muse to track his golf game, shot by shot, as he works toward a single-digit handicapThe mindset shift designers need when the output is never pixel-perfectWhen to use the main chat versus a side chat, and where the architecture still falls short— Brought to you by: WorkOS https://workos.com/?utm_source=lennys_howiai&utm_medium=podcast&utm_campaign=q22025—Make your app enterprise-ready, with SSO, SCIM, RBAC, and more — In this episode, we cover: (00:00) Welcome and guest intro (02:50) Designing Muse (05:33) Choosing which primitives to surface: characters, ideas, feed (08:50) Speed of model and response times (11:01) Generative UI and personalization (14:25) SkyMall demo: Instagram saves turned into a shopping catalog (17:16) Shopping, monitoring, and price alerts with Muse (19:07) Instagram and Pinterest connectors for visual data (21:40) Golf use case: round recaps, shot data, practice breakdowns (25:28) Main chat versus side chats (27:56) Old blog turned morning podcast: 20 years of posts via the Wayback Machine (29:39) Chess dashboard: game analysis with a snarky coach (30:59) What designers need to learn for AI, and what still works (34:33) Prompting strategy when Muse is being obstinate (36:31) Where to find Alex — Tools referenced: • Meta Muse: https://muse.ai https://muse.ai/ — Other references: • The Wayback Machine: https://web.archive.org https://web.archive.org/ • Instagram: https://www.instagram.com https://www.instagram.com/ • Pinterest: https://www.pinterest.com https://www.pinterest.com/ • Chess.com: https://www.chess.com https://www.chess.com/ • SkyMall/Sharper Image aesthetic: https://www.skymall.com https://www.skymall.com/ — Where to find Alex Cornell: LinkedIn: linkedin.com/in/alex-cornell-5449a99 http://linkedin.com/in/alex-cornell-5449a99 X: https://x.com/alexcornell — Where to find Claire Vo: ChatPRD: https://www.chatprd.ai/ Website: https://clairevo.com/ LinkedIn: https://www.linkedin.com/in/clairevo/ X: https://x.com/clairevo — Production and marketing by https://penname.co/. For inquiries about sponsoring the podcast, email jordan@penname.co.
Transcript
7,575 words · 3 speakers
Topics
- Muse
- Marlo
- AI
- main chat
- Jollybot
- generative UI
- progressive disclosure
- Threads
- side chats
Page updated · transcript and summary are machine-generated