By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.
How I AIHow to build prototypes that actually look like your product | Colin Matthews
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
At a glance
WHAT IT’S REALLY ABOUT
AI prototyping with component libraries for brand-consistent product-like mockups fast
- Claire Vo and Colin Matthews demonstrate a workflow that makes AI prototypes look and feel like the real product by starting with a component library rather than generating full pages from scratch.
- Colin shows how to capture UI elements from screenshots (e.g., Airbnb) to generate a reusable set of components, then “fork” a project to assemble new screens (home, detail pages) using those components for visual consistency and fewer errors.
- They then demo Magic Patterns, including a Chrome extension that can select an element directly from a live website (e.g., ChatPRD) and convert the underlying HTML/styling into reusable components that can be versioned and upgraded across prototypes.
- The conversation closes with team adoption tactics (empathy, positioning prototypes as communication artifacts), practical workflow discipline (checkpoints/versions/forks), and a debugging prompt pattern: ask the AI to explain what’s happening before writing code.
IDEAS WORTH REMEMBERING
5 ideasStart with components, not full screens.
Instead of prompting an AI tool to recreate an entire view, first force it to generate a library of primitives (logo, nav, cards, buttons). New screens then become composition tasks, producing more consistent results.
Use screenshots to bootstrap a brand-consistent UI kit quickly.
Colin repeatedly feeds screenshots and prompts “continue adding components,” extracting recurring UX patterns. This builds a shared component set that makes prototypes look like the real product (or a close facsimile) without manual Figma work.
Forking separates “design system assets” from “prototype experiments.”
Keep a master project as the component library, then fork it to build specific flows. This prevents accidental changes to the underlying primitives while enabling rapid iteration on screens.
Aim for “recognizable fidelity,” not pixel perfection.
The goal is to avoid distracting stakeholders with unfamiliar UI, not to match every icon or image exactly. Close-enough fidelity helps teams focus on the new feature concept rather than the prototype’s visual inaccuracies.
Extracting from a live site can replace costly design-engineering help.
Magic Patterns’ Chrome extension lets you click a UI element on your site, capture its HTML/styling, and convert it into a reusable component—reducing dependence on engineers to manually package design-system code for prototyping tools.
WORDS WORTH SAVING
5 quotesThe concept here is actually pretty simple… creating these components first, rather than starting with your views.
— Colin Matthews
It is still a little challenging… The goal is typically to represent the product in a way that doesn't make people feel like you're talking about a different topic.
— Colin Matthews
Don’t think about improving the composed application, actually think about improving the components, and then the composed application can follow downstream from that.
— Claire Vo
When you've got something that works, the number one mistake is to keep vibe coding without a checkpoint or a commit or a version.
— Claire Vo
Explain to me why this is happening. Don’t write any code.
— Colin Matthews
High quality AI-generated summary created from speaker-labeled transcript.
