The New Full-Stack Designer

Summary

Marvin Schwaibold, who leads a product design studio at Shopify, explains how AI tools are changing design work and demonstrates a process that combines rapid prototyping with careful human judgment. The episode also reviews five websites, focusing on clarity, hierarchy, motion, and interaction.

Design Process and AI

• Marvin says design remains highly iterative, but his workflow has shifted from primarily creating polished Figma canvases to building and testing prototypes quickly.
• He uses Figma as a sketchbook or notebook, then moves ideas into tools such as Codex and Claude Code, often working in the terminal. He returns to Figma when a visual sketch explains an idea better than text.
• AI can handle substantial implementation work, but designers still need to set direction, assess quality, make decisions, and curate the results. Marvin describes designers as increasingly acting like editors.
• A useful approach is to break a complex idea into smaller, understandable tasks, guide AI through them, and refine the outcome through repeated testing.
• Prototypes help teams respond to a tangible experience instead of debating an abstract specification. Designers can build experimental versions without requiring engineers to implement every idea first.

Shopify and Molly Studio Work

• At Shopify, the team has worked on a redesign of the administrative experience and on Sidekick, an AI assistant that can help merchants manage their stores.
• One concept connects recommendations in the Shopify home feed to Sidekick: a merchant can review a suggestion, start a conversation, and make changes to their store through that interaction.
• The Shopify team prototypes directly in a playground environment to explore and test ideas before they become production code.
• Before joining Shopify, Marvin co-ran Molly, a Brooklyn design studio with Jay Tal. Its work included projects for Collins, Apple, and Google.
• Molly’s approach emphasized close collaboration with clients, taking on a small number of projects at a time, and organizing websites so people could explore information gradually rather than face everything at once.
• A Collins bookshelf interaction illustrates the value of small, carefully crafted details. The team refined the books’ timing, spacing, colors, and movement so the shelf felt organic and alive. These details also demonstrated the studio’s care and craft to potential clients.

Creative Experiments and Prototyping Techniques

• Grass shader: Marvin tried recreating real grass as a controllable shader. He started by giving AI a video and describing the desired result, then compared the output with the reference and refined his instructions.
• To make the grass more realistic, he used a topographic-map visualization to reason about the ground beneath the blades. Treating the surface as terrain with different heights and depths gave the shader a more convincing base.
• He used Figma sketches to communicate motion and interaction states that were difficult to express in words, then asked AI to implement them.
• DialKit: Marvin and collaborators used controls to adjust variables such as blade length, wind, and animation behavior. Exposing these settings made it easier to experiment with and tune a prototype without repeatedly editing code.
• Carousel: Inspired by a portfolio carousel by Mike Matas, Marvin explored how dragging an item could create ripples and collisions with neighboring items. He used DialKit to adjust the animation’s parameters and says it took 91 attempts to get the effect right.

Advice for Designers and Founders

• Treat AI as a patient collaborator: describe the goal, inspect the result, and keep iterating rather than stopping at the first attempt.
• Provide useful references, such as a screenshot, existing animation, or color palette, when describing what you want.
• Experiment with implementing work you admire. Rebuilding an interaction can help you understand how it works and develop your own craft.
• Break complex concepts into smaller parts that can be prototyped and evaluated individually.
• Use lightweight controls to expose important variables, so you can test and tune behavior directly.
• Keep human judgment responsible for the quality bar: AI can generate and build, but it does not reliably know what looks or feels right.

Website Review Principles

• Endstack: The desktop-like product interface is engaging, but the page presents too many competing elements. Establish one primary goal and make the product’s purpose and main call to action easier to identify.
• SkillSync: The page is dense and relies on tabs to explain many capabilities. Summarize the product around a few core benefits, make the text easier to scan, and reveal more detail progressively rather than hiding important information behind tabs.
• Instaplay: The page should make clear whether visitors are meant to explore existing games or create their own. Let people experience the product before asking them to register, show what the games actually feel like, and make navigation and return paths obvious.
• Mintlify: Its restrained layout, clear headline, prominent action, and polished typography make it feel professional. The review suggests increasing the prominence of important card headings and avoiding repeated background animation that competes with content. Homepage sections can act as chapters that lead into deeper pages.
• Knowlify: The page has several competing text sizes and shifting headline elements, making its purpose harder to grasp. Lead with a concise explanation, show the product’s video output directly, and make examples easy to open, enlarge, and browse. The review also criticizes scroll-jacking: sites should respect familiar, user-controlled scrolling rather than forcing a particular pace.

Creators and Contributors Featured

• Marvin Schwaibold — Shopify design leader and guest
• Aaron Epstein — Host
• Jay Tal — Marvin’s Molly studio partner and design collaborator
• Jesper — Collaborator on the bookshelf animation
• Michael Spicer — Collaborator on the grass-shader project
• Josh — Collaborator who suggested adding adjustable controls to the shader prototype
• Mike Matas — Designer whose carousel inspired Marvin’s experiment