Imagine leading the design of an AI product that skyrockets to a billion+ dollar valuation in under two years…

That’s the story of Perplexity and today we get to hear from their founding designer and current Head of Design, Henry Modisett. Some of the highlights from this conversation:

• What it takes to thrive as a founding designer
• Why Henry likes hiring designers who can code
• The challenges of designing dynamic interfaces
• Why Henry didn’t want to anthropomorphize the AI
• The initial creative direction for the Perplexity brand
• The keys to making a consumer product cognitively fast
• Why Henry built a mini design system as his very first step
• a lot more

SHOW NOTES

• Ivan (CEO of Notion)’s tweet about not having a design system
• Perplexity’s incredible brand designer named Phi
• We talked about how booking.com is a masterclass in optimizing UI

KEY TAKEAWAYS

Starting with a design system

Before Henry had any idea what the Perplexity product would become, he built a component system in React as the first step. The goal was to give himself a toolbox to make it easy to assemble new features. Many components are obvious (ex: you know you’ll need a grid, type system, color system, buttons, etc.). We don’t have to overcomplicate design systems. They’re the thing you invest in to move fast… not the thing you invest in once you have most of the interface figured out.

Empowerment through code

When you write code, you develop a stronger emotional attachment to the product. You’re also empowered to continually make improvements without having to go through engineers. The more removed you are from what ships, the easier it is to dish blame on someone else for an experience being janky.

“Having designers that can code is a hack…quality just happens”

Velocity is everything

Henry makes a point to prioritize velocity over exploration, debate, visual design, etc. And a big part of what makes that possible is empowering designers to make decisions. If it’s a UX question, the designer needs to make a call (”go with your gut and if you want to change it later you can”). This is also why having designers who can code is key. Nothing is cemented. You don’t need permission to iterate after something ships.

Dynamic UI systems

At the root of Perplexity are UI systems that display dynamic content based on what the user searches. That means as a designer you can’t possibly mock up all use cases. You have to think about interfaces as slightly abstracted (ex: “entity comparison” which can work for comparing dog breeds, restaurants, etc.). Part of designing a dynamic system is you have to be ok with percentage outcomes. Sometimes the formatting isn’t going to be perfect.

You’re designing the system, giving AI the tools to use, and hoping that it works most of the time.

This week's episode is with Ryan Scott who was an early designer at Doordash and then spent years as a design lead at Airbnb. Nowadays Ryan teaches hundreds of designers ranging from seniors to VPs how to make a bigger impact at your company.

This episode is jam-packed with insights about:

• What it’s like presenting at Airbnb CRIT
• Ways to unlock your credibility as a designer
• How to mitigate risk when presenting your ideas
• What it takes to speak the language of the business
• Types of “PM-y” questions that designers should be asking
• How Ryan led a massive redesign of the Airbnb booking flow
• The right (and wrong) way to make a case for investing in UX debt
• How to talk about your work in a way that resonates with non-designers
• a lot more

• Get $100 off of Ryan’s course “Describing the ROI of design”
• Ryan’s case study on redesigning the Airbnb checkout flow
• Past interview with Alex Schleifer (Chief Design Officer at Airbnb)

Investing in my personal brand has been the best investment I’ve ever made in my career. So I wanted to find the perfect designer to give a personal branding masterclass and I think it’s Oliur.

He’s built a massive audience across YouTube, Instagram, Twitter, etc. and he shares his hard-earned knowledge in this episode:

• Oliur’s keys to growing on social media
• Different ways to create leverage in your career
• How he (accidentally) landed a billion-dollar client
• How you can get more confident putting yourself out there
• Specific ways to build meaningful connections with other designers
• What designers need to know to get better at marketing themselves
• a lot more

• Oliur cites James McDonald as an example of someone sharing their work
• Oliur’s iPhone presets and high income streams videos
• We talk about his friend Ali Abdaal (YouTuber, Podcaster, NYT bestseller)
• Jason Levin’s iMessage screenshot of Travis Scott performing
• I referenced Jack Butcher’s “this is pointless” graph
• Dan Petty’s “That Marker Pack”
• The old Audience Building course on Maven (RIP)
• Turkish designer Oguz (Oliur talked about the power of his aesthetic)
• Gabe Valdivia’s new app Almanac that he built with Cursor

This week’s episode is with Maheen Sohail who is a senior staff designer working on generative at Meta. She joined as one of the earliest designers on both the VR and AI teams, so a big part of this discussion is about navigating ambiguity when there’s no clear playbook to follow.

We go deep into:

• New types of interface patterns for AI
• The unique ways she thinks about prototyping
• Ethical considerations when designing AI produdcts
• How Maheen explores AI models through side projects
• Why the goal posts for what it means to be a designer are shifting
• a lot more

⭐ KEY TAKEAWAYS

Nobody knows what they’re doing

This is especially true when you’re designing products for emerging technologies like AI. It’s easy to look at people working on these AI-native products and assume they have it all figured out… but we’re all still learning and exploring what’s possible. This came up in George Kedenburg’s episode too.

Curiosity > everything

I asked Maheen what traits are more important than curiosity for people interested in designing AI products. Her answer? Nothing.

The importance of passion projects

There’s a trend I’m noticing in these interviews… the designers who are creating cutting edge use cases for AI are the ones actively exploring the technology with side projects. Reading essays isn’t the way to learn. You have to want something to exist in the world and use that as a reason to figure out what’s possible. For Maheen it was using AI models to colorize images of Pakistan. For Nate Parrott it was using AI to hallucinate in HTML.

• Maheen references Alex Cornell’s storytelling abilities
• Maheen’s demo for animating drawings (it’s very fun)
• We referenced Colin Dunn’s episode about the importance of naming in AI products
• Maheen’s colorized photos of Pakistan
• Maheen’s FakeID podcast
• Meta’s AI release
• Maheen raved about the Rosebud journalling app

After 8 years designing at Meta, George Kedenburg III pulled a 180 and joined Humane as a design lead. So this conversation is a deep dive into designing AI products and how the role of product designer evolves in an AI-native company:

• How to become a creative problem solver
• How George navigates ambiguity at Humane
• Why there’s no such thing as an edge case with AI
• What George learned while using AI to learn Python
• How AI is reshaping the landscape for software design
• Why George created a Slackbot to prototype his ideas
• Why designing AI products is a bit like designing a kitchen
• a lot more

Pushing past the pixels

The real value of design is being able to look at an ambiguous situation and understand what you should explore.

Rectangles so happen to be the most common way to express that value. But the real skill is creative problem solving.

Working at a company like Humane forces designers to contribute design thinking beyond the pixels.

Prompt design > prompt engineering

If the AI model is a chef, then you’re responsible for designing the kitchen.

You don’t know what the user will order, so it’s a lot of trial and error to ensure you have the right data on hand at the right moments.

It’s no different than thinking through drop-off in an onboarding flow. Which is why George views working with these models as “prompt design” rather than “prompt engineering”

There are no AI edge cases

When you’re prototyping AI products, your prototypes don’t “break” or “fall over” like they do in Figma. That’s because the boundaries of what exists in the prototype become much blurrier.

Instead of designing contained flows, you’re laying a foundation and allowing the model to extrapolate out from there. There are no more hard edges.

George mentions Claude Artifacts as an example of someone putting the pieces together in the right order

Visual Electric has quickly become my go-to product for image generation and in this week’s episode we get to learn from the founder and designer, Colin Dunn. The whole discussion is an excellent look at the design founder journey as well as a deep dive into AI-native creative tools. We get into the weeds about:

• Visual Electric’s big bet to take on Canva
• The hidden challenges with designing AI products
• Colin's approach to early user and market research
• The art and science of raising funds for your startup
• Where the value will accrue in the landscape for creative tools
• Where Colin draws the line between abstraction and power in UX
• The wild backstory of how the company was named “Visual Electric”
• Lessons learned learned from early startup ideas that were shot down

Key takeaways:

1. AI is like electricity. Once we gained access to this new form of power, we immediately replaced candles with outlets. But it took 50+ years before the microwave and other staple household appliances were invented. When electricity came on the scene in the late 19th century it would’ve been impossible to imagine these types of products. Colin believes the electricity layer will quickly become commoditized, and instead is solely focused on building “appliances” for AI. Because someone is going to build the oven, the sewing machine, the coffee percolator, the electric can opener, etc. It might even be you 😉
2. Choosing the level of abstraction is one of the core challenges with designing AI products. Most users don’t want to be burdened by all of the knobs and levers of the AI model. That’s why it’s essential that we define new patterns and mental models that make AI easy to understand. But you have to be careful, because “the more you abstract something, the less control users have over it”. One example Colin shares is why they’re considering combining the “reference slider” and “creativity slider”. It simplifies the UX but at the cost of control. And striking that balance is one of the challenging parts about designing Visual Electric.
3. Language is an awkward medium for visual ideas. We need more effective ways to provide visual inputs if we want to generate high quality visual outputs.Want to get early access to Visual Electric’s new product? 👉 Click hereColin talks about his great his experience with User InterviewsGreg Rosen was the investor who helped Colin in the early daysJess Lee is the Sequoia partner they met withTom from Manual led the branding and chose the name “Visual Electric”Here’s Manual’s case study on designing the Visual Electric brandVisual Electric’s brand story page (which Ellis Hamburger helped with)

The Arc browser is one of the best products I’ve downloaded in the last few years which is why I’m so excited about this interview with their founding designer, Nate Parrott.

This conversation, we get an inside look at what makes design at the Browser Company so unique. We discuss:

• How Nate went from engineer at Snap to the Browser Co.
• Why design at Arc prioritizes fingerprints > consistency
• How Nate collaborates with engineers on Arc
• The story behind Arc Search’s hook feature
• The Browser Co’s culture of prototyping
• How Nate balances intuition and data
• a lot more…

• Nate’s gamified personal website
• A24 film studio
• Nate references the Co-Star astrology app and the old Zenly app
• Patrick Moberg (the designer on Boosts)
• Karla Cole (the designer who pulls inspiration outside of tech)
• Nate’s feeeed RSS app

This week’s episode is with Dan LaCivita who is a co-founder of Play—a new way to design mobile apps using native iOS materials. As soon as I saw this app clip I knew it was a big deal and that I had to interview him to learn more.

So in this conversation we go deep into:

• How to unlock the skill of interaction design
• What Apple’s sandbox unlocks for designers
• Why they decided to launch Play as a mobile app
• The importance of keeping your hands in the clay
• How AI might impact the future of interaction design
• a lot more

• Dan mentions his cofounder JoonYong Park
• The App Clip that first blew my mind

Gavin Nelson is currently designing the Linear mobile app in addition to the best app icons on the internet. So this episode is a deep dive into prototyping and interaction design:

• Gavin’s journey learning SwiftUI
• How Gavin got into designing app icons
• How to design physics-based interactions
• What it’s like designing for power users at Linear
• Why interaction design is so important for mobile apps
• Why Gavin uses code almost exclusively for prototyping now
• When it makes sense to prototype at the beginning of your process
• The interplay between custom and stock components in mobile design

• Things by Cultured Code
• Janum shows interactions in the Netflix mobile app
• Episode with Brian Lovin (who Gavin worked with at Github)
• Episode with Alex Cornell (who started the design of the Linear mobile app)
• Parallax effect from the Linear onboarding screen

If you watched last week’s episode, then you heard Jordan Singer say how visual search was actually the “killer feature” (not generative AI). What’s interesting is that this feature was actually a mid-project pivot.

So in this episode we get the full behind-the-scenes of visual search from design engineer Vincent van der Meulen:

• Why they decided to pivot
• How Vincent went cowboy mode
• A look at the future of design engineering
• What good design looks like in an AI world
• Why visual search is a big deal for future AI tools
• How Vincent used video to get Figmates excited about his idea
• a lot more

• Marco and Jordan from the old Diagram team
• Matt Dailey was the engineer that proved designs can be represented as images
• 2024 Config Keynote showing Vincent’s Uber example for visual search