Can you ship?

At Lovable being comfortable shipping was pretty much a prerequisite when Arthur first talked to the head of design, Nad Chishtie, about joining. Every designer on the team ships their own PRs. For example Arthur shipped a PR to fix a button issue in 10 minutes!

90% agentic exploration + 10% human touch

Arthur uses Lovable with the Paper MCP to explore a ton of design iterations. For further explorations he leaves comments on the Paper canvas like he would for a teammate and Lovable reads them and makes the changes itself. For the last mile Arthur tweaks the variations he likes by hand in Paper. Old school style!

Your acquired taste is the moat

It’s super satisfying to prompt an agent and get 20 variants instantly on the canvas. But if you don’t have the muscle memory to judge why something is good you’ll just go with good enough. That’s why Arthur recommends junior designers to get really good at fundamentals like color, typography, visual rhythm and layout by grinding in the hours!

📺 Watch the full interview

🔓 Unlock AI & Growth Design Interviews

Refer Sneak Peek to your friends to unlock exclusive interviews on AI Design & Growth Design!

Refer a friend

1. Show up to client calls with strong opinions

Nick and his team build mood boards from the client’s questionnaire for the kickoff call so the conversation is productive right away. Hearing that a client likes Ramp, Vercel or Nike doesn’t help you move fast. Nick also shows fewer options, like just two palettes they believe in. If the client pushes back Nick digs into why and comes back with a better version!

2. Designing the Lovable logo

In the very first meeting with Lovable, Nick pitched turning a heart on its side so it becomes an L. That idea became the Lovable logo. Designing for Lovable put Nick and his studio on the map and made them legit. But he’s not slowing down: “I gotta do five or six more Lovables so that everybody who comes to us believes that we’re legit”!

3. Giving clients vibe coded tools for handoff

For about 60% of clients, Nick and his team build a custom tool in Lovable so they can keep making on-brand social posts, decks and art after the 2 week sprint. I feel this is the future of brand handoff because a brand only works if the team can actually put it out into the world!

📺 Watch the full interview

🔓 Unlock AI & Growth Design Interviews

Refer Sneak Peek to your friends to unlock exclusive interviews on AI Design & Growth Design!

Refer a friend

1. Finding inspiration from the most unlikely sources

To design the most iconic brand India has ever seen you need to look for inspiration in the most unlikely places. Kamini studied the architectural plans for the Lotus Temple in Delhi and they heavily influenced the final Sarvam logo design. For the other brand assets she was heavily inspired by the patterns in their office bathroom tiles and turned them into shapes for the brand!

2. Tools and AI models are just an excuse

The entire brand design team at Sarvam AI works primarily in Figma and explores thousands of iterations painstakingly by hand. I feel this just goes to show that it’s less about using the latest tools or AI models and more about the designer’s desire to push the craft!

3. Designing a tech brand rooted in Indian culture

Sarvam’s brand had to feel Indian without being too traditional or feel techy without any connection to Indian roots. Kamini picked terracotta orange to give a feel of the Indian heritage and indigo blue to convey that it’s a tech brand. The team constantly asks “Is it too Indian or too tech?” to find the right balance in every brand iteration!

📺 Watch the full interview

🔓 Unlock AI & Growth Design Interviews

Refer Sneak Peek to your friends to unlock exclusive interviews on AI Design & Growth Design!

Refer a friend

1. Lock the user flow before vibe coding

Paavan doesn’t jump straight into vibe coding. He first locks the whole user flow in Whimsical’s wireframe mode. It’s so low fidelity that he only thinks about the content design and information hierarchy instead of the design details. High fidelity prototypes vibe coded in Cursor or other AI tools lock you into one direction too early so he holds off on them until the flow is locked!

2. Whimsical + Claude + Conductor workflow for designers

Once the user flow is locked in Whimsical, Paavan uses the HTML to Design plugin to capture the live Granola app design into Figma. He tweaks the designs in Figma then copies them as PNGs and pastes them into Conductor to build out a prototype with Claude Code. This way he ends up with a working version of the app to put in front of users!

3. The founders still prototype with the team

At Granola, cofounder Sam shares prototypes with the design team to spark new ideas. The team also has ‘Taste the Soup’ sessions where anyone can present their work and get feedback from the founders but the work has to be presented in code only. I feel this level of involvement from the founders to push the bar on craft is why Granola is one of the best designed AI note takers in the market!

📺 Watch the full interview

🔓 Unlock AI & Growth Design Interviews

Refer Sneak Peek to your friends to unlock exclusive interviews on AI Design & Growth Design!

Refer a friend

1. Skip design systems for concepts

Every time Cris Arevalo explores early concepts in Lovable he skips the design system on purpose. He doesn’t care about visual polish at this stage. The goal this early in the process is to validate different concepts and get alignment from the broader team.

2. Use Playground for shipping PRs

Deel has a UI prototyping playground so designers can quickly spin up high-fidelity prototypes using their actual design system. Cris Arevalo uses the UI Playground to ship PRs for small features or small fixes.

3. Minimize token burn

Cris Arevalo feeds his entire FigJam discovery board into Claude Code to give it context. This minimizes token burn and helps the AI model get more context about the project.

4. Eng handoff with live prototypes

For engineering handoff Cris Arevalo shares high-fidelity prototypes plus a video walkthrough instead of static screens. Engineers can then click through the prototype to understand the flow and the intent.

📺 Watch the full interview

🔓 Unlock AI & Growth Design Interviews

Refer Sneak Peek to your friends to unlock exclusive interviews on AI Design & Growth Design!

Refer a friend

1. Automate design inspiration with Mobbin MCP

Harssha automated design inspiration with a custom Codex skill on top of Mobbin MCP. Instead of manually hunting screens on Mobbin and dropping them into Figma he lets the skill pull references of competitors and even outside their industry when he asks!

2. Sharing skills Across Codex, Cursor, and Claude

The Pocket FM design team jumps between Codex, Cursor and Claude depending on the task. Harssha symlinked one skill set across all three so they never maintain three separate versions. Cursor and Codex share a “.agents” file while Claude keeps a version of it saved as “.claude” and these files stay linked. The skills live in the team’s GitHub repo so when a designer clones the prototyping playground they already have them.

3. Top 3rd party skills

DialKit by Josh Puckett gives Harssha knobs and sliders on the live prototype so he can tweak size and spacing without burning another prompt. Agentation by Benji Taylor lets him leave Figma like comments on an AI prototype and push that feedback straight into the model he’s using. For motion he leans on Emil Kowalski’s open-sourced animation skill so Codex isn’t inventing timing from scratch!

4. Building a Second Brain

Every time Harssha pushes a demo a script reads his whole Codex chat and pulls out unspoken rules he kept repeating to the model. Think noise texture on buttons or other small DS nuances that never made it into the docs. Those learnings get grouped and a side agent appends them to GitHub so the shared team skills pick them up. So the design team’s second brain keeps getting smarter and smarter every time a designer ships a PR!

📺 Watch the full interview

🔓 Unlock AI & Growth Design Interviews

Refer Sneak Peek to your friends to unlock exclusive interviews on AI Design & Growth Design!

Refer a friend

1. Grok Bot vs Claude vs ChatGPT

Sabina has a Claude and ChatGPT subscription but she doesn’t love how both those AI tools take over her local computer while heads down designing.

With Grok Bot the agent keeps working in its own separate cloud computer and pings her when it’s ready. Her laptop stays free so she can keep designing while the agent does the grunt work!

2. AI iterations as inspiration

What blew me away is that Sabina does not go to Cosmos, Arena or Pinterest for design inspiration. She just asks the AI agents for a bunch of different iterations.

Because the AI models are already trained on existing UI so no point in providing references. Let your agent cook 5 or 10 options and then pick from those!

3. Codex vs Cursor

One thing I’m always curious about when I interview guests is why they pick one tool over another. I like sharing that thinking so it helps you make your own call.

When I asked Sabina why she prefers Codex over Cursor I was surprised her answer was how clean and beautiful the Codex UI feels especially the folder system. That was my aha moment. Most AI tools can already do the same jobs now. I think the winner is going to be the one that abstracts away the complexity.

That’s also why I switched over to Grok Bot after talking to Sabina because it feels so easy to use compared to the competition!

📺 Watch the full interview

🔓 Unlock AI & Growth Design Interviews

Refer Sneak Peek to your friends to unlock exclusive interviews on AI Design & Growth Design!

Refer a friend

1. Custom actions

At CNN Jennifer Spriggs saves her repeat prompts in Cursor as commands to save time. Her ‘Commit to Branch’ command for example is just one line in a markdown file that tells Cursor to stage everything and commit with a short message and don’t create a new branch or push.

She added the “don’t” part at the very end on purpose, so the agent doesn’t go ham and start doing stuff she never asked for.

2. Picking the right model

At DuckDuckGo Karl Koch splits his workflow across two models in Cursor. Opus does the thinking in plan mode and Composer builds the designs.

This isn't about saving tokens. It's about speed. Composer ships his plan way faster at the same quality!

3. Using Cursor for automating design systems

At monday.com Elad Mizrahi uses Cursor to generate 80-90% of their design system code by creating custom rules and specs. He used to maintain entire Storybook pages in Figma. Now Cursor builds and updates those Storybook pages for him automatically!

📺 Watch the full interview

🔓 Unlock AI & Growth Design Interviews

Refer Sneak Peek to your friends to unlock exclusive interviews on AI Design & Growth Design!

Refer a friend

Anyways below are three key design tactics I learned from interviewing Jenna Leonardo on how they use AI for design sprints at Headway.

1. Sketch on paper before vibe coding

Jenna’s team is still old school and does paper sketching before vibe coding so that everyone on the team focuses on the big idea instead of implementation or execution of the idea. Her team also created a skill using AI that creates low fidelity wireframes for stakeholders not comfortable with sketching!

2. Use real data vs lorem ipsum

Jenna’s team loaded real insurance carrier names into their Magic Patterns design system so that every prototype generated feels real. (They moved their design system source of truth from Figma to Magic Patterns)

This is why when they do user testing during the sprint, they get actually actionable feedback from the users. For example, the insurance listed on a therapist’s profile would actually say Aetna instead of “Acme Corp” or “Lorem Ipsum”.

3. Skip the debate and prototype everything

In the old sprint model often designers would waste the first one or two days debating with stakeholders which ideas to double down on, just because there was only one day in the sprint dedicated for prototyping.

But by leveraging AI, if Jenna’s team has 10 different ideas they want to pursue, they actually build prototypes for all of them. And then test all these prototypes with real patients for feedback instead of voting or creating pros/cons for each idea.

📺 Watch the full interview

🔓 Unlock AI & Growth Design Interviews

Refer Sneak Peek to your friends to unlock exclusive interviews on AI Design & Growth Design!

Refer a friend

1. Becoming AI native without giving up Figma

Going AI native at Razorpay doesn’t mean ditching Figma completely. Jain now uses Figma for visual note taking. He does a design audit and pulls together his 1-1 notes and competitor screenshots into a Figma file.

He then feeds the whole Figma file to Claude as the initial prompt to build prototypes!

2. Writing a design doc before prototyping

At Razorpay designers don’t directly start vibe coding. For every medium to large project Meena uses Claude first to write a design doc and align with cross functional stakeholders on the problem and business goals!

3. Presenting prototypes in user’s voice

When Meena presents her design work at Razorpay she doesn’t click through screens or just show AI prototypes. She walks the entire room through the whole flow in the user’s voice.

Meena won't say I made this button. She'll say I'm a merchant who woke up wanting to check my orders. So all the stakeholders feel the actual user pain points instead of just watching a demo prototype!

📺 Watch the full interview

🔓 Unlock AI & Growth Design Interviews

Refer Sneak Peek to your friends to unlock exclusive interviews on AI Design & Growth Design!

Refer a friend

1. Designing a new app in one week

Ezaz designed Swiggy's entire Snack app in a single week. Not just the concept but the whole thing - UI, animations, illustrations, error states, and branding.

The reason it was even possible is because he wasn't starting from scratch. Ezaz reused components from Swiggy's existing design system and tweaked them. Also engineers and designers were in the same room the whole time. So every screen got checked for feasibility on the spot. Also the cofounder Harsha reviewed the Figma files and prototypes himself which cut out the middle management delay!

2. Designing for Gen Z

Ezaz wanted Snacc to feel like something Gen Z was already in on and not a food app trying to look young. So he designed it around the things they actually share which is memes and most memes involve cats.

So he designed a whole brand around a ginger cat and also created delightful error states! My favorite error state is the one where the cat is cutting a cucumber when all the vegetarian food items are sold out!

3. Design hackathons to make AI images look real

Some dishes are just hard for AI to get right especially Indian food. The crispy rolled dosa was so tricky that the Swiggy design team turned it into an internal design hackathon to see who can nail it.

And the fix always comes down to the tiny design details like prompting for a single coriander leaf on top or ghee dripping down the side because that’s what makes AI food look real and appetizing!

4. AI won't replace designers

Ezaz strongly believes that AI will never be able to replace the thinking and the craft behind a designer’s work. He has an amazing analogy - he's the director of the movie and the tools are just the camera. The camera keeps getting better but you still need someone who knows how to direct.

His advice to worried designers is simple. Treat AI as just another tool and play with it in your free time so you're never behind. But keep your focus on the fundamentals of design. Jumping to the tool without the craft won't take you anywhere.

📺 Watch the full interview

🔓 Unlock AI & Growth Design Interviews

Refer Sneak Peek to your friends to unlock exclusive interviews on AI Design & Growth Design!

Refer a friend

1. Use Ask User Question tool in Claude

You can try to cram every detail into one giant prompt and one shot it. Danny doesn't.
He tells Claude to use the ask user question tool and lets it interview him instead. It pulls out the stuff he had in his head but never typed in the prompt. So all the little design details actually make it into the prototype!

2. Claude Design vs Claude Code

Danny picks his tool based on how sure he is about the idea. If he's exploring wide with throwaway concepts he uses Claude Design. If he's dialing in the details or actually shipping he uses Claude Code.

3. Don’t worry about token optimization

Wealthsimple takes a carrot approach and encourages the design team to use different AI tools and not worry about token optimization. The goal is to save time instead of tokens!

4. Use code review skill when pushing PRs

Before Danny tags an engineer on a PR he runs Wealthsimple's code review skill first. This Claude skill spins off eight sub-agents, each one hunting for something different like dead code, performance, security holes, best practices. It catches code issues so the engineers aren't stuck cleaning up his mess!

📺 Watch the full interview

🔓 Unlock AI & Growth Design Interviews

Refer Sneak Peek to your friends to unlock exclusive interviews on AI Design & Growth Design!

Refer a friend

1. Your PRD is the first AI prompt

Andy doesn’t waste time on figuring out the perfect prompt or one shotting. Instead he uses Miro AI to create a PRD based on his design discovery findings. Then he pastes that whole PRD straight into Magic Patterns as the initial prompt.

As a result the first prototype created using AI is already 60 to 70% of the way there!

2. Move your design system out of Figma

Maintaining the design system in Figma was a huge resource sink for the ESO design team. Five designers, no dedicated system owner, constant back and forth with engineering to keep Storybook in parity.

So Andy and the design team rebuilt the entire design system in Magic Patterns in a week. When a user tests one of Andy's prototypes now nothing feels off because the prototype uses production components.

3. Set up AI guardrails

Instead of re-prompting over and over for consistency Andy's team baked the healthcare standards, brand, UX laws and other context into Magic Patterns as rules.

This cuts the random nature of AI so the same prompt stops giving four different looking results. And it lets PMs and execs prototype without going outside the lines!

📺 Watch the full interview

🔓 Unlock AI & Growth Design Interviews

Refer Sneak Peek to your friends to unlock exclusive interviews on AI Design & Growth Design!

Refer a friend

1. Prototype Switcher

At CNN Jennifer Spriggs builds interactive prototypes in Cursor with a version switcher baked in so she can flip between different design options live during a design critique.

This saves her hours of building out different prototypes in Figma!

2. Async Design Critiques

At Ramp George Bugg runs async design critiques to cut down on the number of meetings. Designers drop vibe coded prototypes built using Magic Patterns into Slack and stakeholders leave feedback whenever it suits them.

He’s found that going async makes people focus way more on the actual design instead of just reacting live during a design review.

3. Critique Recaps

At Figma Kelly starts each design critique with a recap of the last one. She shares the high level decisions that were made plus the feedback she actually acted on since then.

It jogs everyone’s memory and makes people feel valued because they can see their feedback actually got implemented!

4. Silent Design Critiques

At Hex Adam Storr runs a silent design critique to prevent the loudest voice in the room from dictating the entire product experience.

It kicks off with a quick two to three minute overview. Then a FigJam timer runs for five to six minutes with some music playing while everyone goes heads down and leaves feedback using stickies. When the timer’s up the presenter walks through the notes and drives the discussion.

5. Design Critique Vibes

At Descript John Voss splits design critique into three separate reviews. One of them is just for the design team to go deep on the design details and craft.

The other two are where the bigger product strategy calls get made. He keeps them separate so important product and go to market decisions don’t creep into the design critique. Crit stays focused purely on improving the craft!

📺 Watch the full interview

🔓 Unlock AI & Growth Design Interviews

Refer Sneak Peek to your friends to unlock exclusive interviews on AI Design & Growth Design!

Refer a friend

1. Proactively find friction in the product

Clint doesn't wait for a PM to hand him a PRD. Every week he goes friction hunting on his own, looking for where users are getting stuck across onboarding, activation, and drop-off points.

Granola records his meetings so he can pull out the exact friction points stakeholders mentioned. Enterpret gives him the voice of the customer aka how users actually feel when using the product. And Hex shows him what users are actually doing in the product. Together these tools show Clint the friction points in the product experience!

2. Use AI to run growth experiments supafast

Clint designs growth experiments using two AI tools depending on the job. When he's inventing something brand new he goes to Claude Code so he can break away from the existing design system and explore freely. When he's refining a product flow he uses Magic Patterns.

And because his design system lives inside Magic Patterns now whatever he builds is production ready. He literally hands engineers a Magic Patterns link as the handoff which cuts out a ton of back and forth. So now Clint is able to ship new experiments in 2 weeks instead of a month!

3. Share context async with stakeholders

Once Clint has a prototype worth sharing he doesn’t schedule a meeting. He records a quick Loom walking through what he explored and why, then drops it into a Slack thread with the prototype links.

It turns into an async meeting that happens without a meeting. Stakeholders drop comments and questions when it suits them, and anything that needs a deeper discussion gets picked back up in his weekly design sync. Show and tell minus the calendar chaos. I think this is a genius strategy for stakeholder alignment!

📺 Watch the full interview

🔓 Unlock AI & Growth Design Interviews

Refer Sneak Peek to your friends to unlock exclusive interviews on AI Design & Growth Design!

Refer a friend

1. Understand how the agent thinks before prototyping

Carol didn't jump straight into pixels. She sketched a little diagram in Figma to map what the agent actually does. For example the agent gets context, does its thinking and crunching, then outputs something the customer can read without hunting for a button.

That mental model was her grounding moment. Once she understood how the agent operates she began vibe coding different ideas!

2. Design in production vs Figma/AI tools

Carol started in Figma and then vibe coded prototypes in Cursor but kept hitting a ceiling. Agentic experiences are non-deterministic so a prototype can only take you so far.

To really understand how the agent responds she had to design directly in production with the actual model and real data.

3. Agentic design principle

A key principle Carol designed around is intentional confirmation. The system asks follow up questions and drafts the transfer for you but it never transfers the money on its own.

You confirm the final action with a deliberate button press. This prevents customers from accidentally moving money to the wrong account or recipient!

📺 Watch the full interview

🔓 Unlock AI & Growth Design Interviews

Refer Sneak Peek to your friends to unlock exclusive interviews on AI Design & Growth Design!

Refer a friend

1. Using AI to go from 0-70%

George doesn’t touch Figma at all in the initial stages of the design process!

Instead he uses the AI prototyping tool Magic Patterns for exploring different design concepts. This helps him move super fast because he doesn’t have to worry about pixel perfectness or responsiveness!!

2. Sharing AI prototypes with stakeholders

George doesn’t share Figma files with product managers. Instead he shares AI prototype links which helps him get stakeholder alignment quickly.

He just drops the prototype links in Slack to reduce meeting fatigue. No unnecessary meetings. I think this is the future of design critiques!

3. Adding AI prototypes in Figma dev handoff file

George still follows the traditional engineering handoff process but with a small caveat. He adds the AI prototype links from Magic Patterns in the handoff page in Figma.

This lets the engineers click through the product experience and visualize the edge cases which in turn reduce follow up clarification meetings.

📺 Watch the full interview

🔓 Unlock AI & Growth Design Interviews

Refer Sneak Peek to your friends to unlock exclusive interviews on AI Design & Growth Design!

Refer a friend