Steve Ruiz

the accidental viral moment

Steve Ruiz is a developer, designer, and startup founder in London with a background in visual art. he builds creative tools for the web, mostly prototypes that occasionally become products. in 2023, one of those prototypes broke the internet.

tldraw started as an infinite canvas whiteboard—think Figma meets Excalidraw, but with better performance and a focus on developer experience. clean API, open source, decent traction. useful tool, nothing revolutionary. then Ruiz and his team built Make Real.

Make Real lets you draw a low-fidelity UI sketch—boxes, text, rough layout—and watch it transform into a functional website. not a static mockup. actual working HTML, CSS, and JavaScript. you draw a login form with wonky rectangles and squiggly text, click a button, and get a legitimate React component. the demo video went viral in November 2023. it felt like magic because it looked stupid-simple.

the technical implementation was clever but not complex: take the canvas screenshot, send it to GPT-4 Vision with instructions to generate code matching the sketch, render the result in an iframe. the magic wasn’t the AI—it was the interface. drawing is faster than describing. sketching bypasses the “I don’t know exactly what I want yet” problem that kills most text-to-UI systems.

why it worked

Make Real hit because it solved a real friction point in web development: the gap between idea and implementation. designers sketch interfaces constantly—on paper, whiteboards, napkins. those sketches rarely become code without extensive translation work. Make Real removed the translation step.

the tool wasn’t trying to replace developers. it was lowering the activation energy for prototyping. you could iterate on layout ideas in seconds instead of minutes. test different navigation structures without touching CSS. rough out an entire user flow before writing a single line of actual code. it made exploration cheaper.

what made the virality sustainable was tldraw’s existing credibility. Ruiz wasn’t a random person with a demo—he’d built a respected canvas library with thousands of GitHub stars. Make Real worked because tldraw’s rendering engine was already solid. the AI was the interface layer, not the foundation.

the team also open-sourced Make Real immediately. you could run it locally, inspect the prompts, modify the generation logic. that transparency built trust and spawned dozens of derivatives: sketch-to-Flutter, sketch-to-SwiftUI, sketch-to-Vue. Make Real became a pattern, not just a product.

from demo to product

the challenge with viral demos is turning hype into utility. most disappear after the initial spike. Ruiz took a different path: instead of scaling Make Real into a standalone product, he integrated the approach into tldraw itself. AI became a feature, not the product.

tldraw now includes AI-powered tools for transforming shapes, generating content, and iterating designs. the canvas remains the primary interface—AI augments rather than replaces direct manipulation. this is important: Ruiz resisted the temptation to build “ChatGPT for design.” he kept the tool tactile.

the company raised funding and hired a team. they’re building for creative professionals who think visually: product designers, game developers, educators, anyone who communicates through diagrams. tldraw’s pitch isn’t “AI will design for you”—it’s “AI will execute the tedious parts so you can focus on creative decisions.”

this positions them differently from figma (collaborative design platform) or v0 (text-to-UI generator). tldraw sits between traditional design tools and generative interfaces. you draw, AI fills in details, you iterate. human in the loop, not human replaced.

the interface question

Make Real surfaced a broader question about AI interfaces: should we keep using text prompts for everything? text is universal, but it’s also slow and ambiguous. drawing is faster for spatial reasoning. voice is better for narrative. gestures work for manipulation.

Ruiz demonstrated that AI doesn’t always need a chat box. sometimes the best interface is the one people already know—like sketching on a canvas. this influenced thinking around multimodal AI interactions. it’s why tools like composio and windsurf support visual workflows alongside text commands.

the sketch-to-code paradigm also influenced agentic coding tools. devin and cursor added screenshot analysis for debugging. claude code integrated browser control. developers realized that seeing the output—not just describing it—makes iteration faster. Ruiz’s work normalized visual context in AI workflows.

there’s a throughline from Make Real to the current crop of canvas-based AI tools: napkin.ai (text-to-diagram), diagram.com (AI design primitives), even openai canvas (code + artifact view). they all treat visual representation as first-class input/output, not just text.

why ruiz matters

Steve Ruiz didn’t invent AI code generation or canvas-based tools. he combined them in a way that felt inevitable in retrospect—which is the hallmark of good design. Make Real proved that AI interfaces don’t have to look like chat windows. sometimes the best AI interaction is invisible: you draw, things happen, you keep drawing.

he also showed how to ride a viral moment without getting crushed by it. instead of pivoting to Make Real as a standalone product (and competing with vercel , replit , and every YC startup with “AI” in the name), he folded the capability into tldraw’s core value proposition. the demo was marketing; the product is infrastructure.

tldraw’s developer-first approach—clean API, TypeScript-native, self-hostable—means it powers tools other people build. it’s embedded in obsidian canvas , excalidraw , and dozens of internal tools. Make Real made tldraw famous, but developer experience makes it useful.

Ruiz represents a specific kind of builder: someone who understands that AI’s value isn’t replacing human creativity but reducing the friction between idea and artifact. the future he’s building isn’t “AI designs websites”—it’s “humans design faster because AI handles the grunt work.” that’s a future people actually want.

→ related: scott wu | steve yegge | ryan florence