Flipbook Prototype Streams Every Pixel from AI, Replaces HTML

A small team of former OpenAI, Apple, and Slack engineers unveiled Flipbook on April 23, 2026, a prototype interface that streams every pixel on the user's screen live from an AI model — with no HTML, no CSS, no layout engine, and no code. The project went viral on X within hours, drawing tens of thousands of posts and turning into one of the week's most-discussed AI demos.

"Imagine every pixel on your screen, streamed live directly from a model. No HTML, no layout engine, no code. Just exactly what you want to see," wrote co-creator Zain Shah in the announcement thread.

Key Highlights

• Built by Zain Shah (ex-OpenAI, Samsung, Y Combinator S13), Eddie Jiao (ex-Humane, Slack), and Drew O'Carr (ex-Apple).
• Streams 1080p video at 24 frames per second directly to the browser over WebSockets.
• Uses LTX Studio, the open-source DiT video model from Israeli company Lightricks, optimized to run on Modal Labs serverless GPUs.
• Any region of the generated image is interactive — clicks regenerate the scene rather than triggering fixed links.
• Live demo available at flipbook.page.

Details

Flipbook replaces the entire web rendering pipeline with a generative video model. When a user asks a question — the demo thread shows a Paris travel planner — the model produces a full-screen illustrated page in roughly 40 seconds. Clicking the Eiffel Tower zooms into it; clicking a ticket price opens a booking panel; clicking a landmark reshapes the entire page into a new scene. There are no HTML elements behind the canvas; every visible button, map, label, and paragraph of text is drawn pixel by pixel by the model.

The team optimized LTX Studio to stream the output at 1080p 24fps through WebSockets, with Modal Labs providing the serverless GPU backend. Because there is no layout engine, the illustrations automatically reshape to fit any window size or aspect ratio — behaviour that would normally require responsive CSS.

Impact

Flipbook lands in the middle of a wider debate about whether AI-native interfaces will eventually replace hand-coded front ends. Developers on X compared it to earlier "AI-generated UI" experiments from Anthropic and Google, but noted that Flipbook pushes the idea further by eliminating the DOM entirely. Open-source enthusiasts, including Grok, posted recipes showing how to reproduce a local Flipbook-style stack using Ollama, ComfyUI, and Flux for image generation.

Critics pointed out obvious trade-offs: accessibility, SEO, form filling, and latency are all harder when the browser can no longer read structured markup. Shah himself acknowledged the prototype is slow and that many demo clips were sped up or edited.

Background

Zain Shah previously worked on AI and robotics research at OpenAI and served as a Creative Technology Expert at Samsung. He is a Y Combinator S13 alumnus. Eddie Jiao was an engineer at Humane and Slack. Drew O'Carr is a former Apple engineer. The team built Flipbook as an independent project and released it as an open experiment rather than a commercial product.

What's Next

Shah said Flipbook is "designed around visual explanations" today but hinted at broader ambitions. "As the models get more accurate and more stateful, the set of things worth doing this way will expand — even ones you'd assume need structured UIs, like coding," he wrote. For now, the team is capturing feedback from the launch wave and working on stability: the public demo has been intermittently unavailable since the announcement owing to demand.

For web developers — including teams building modern React, Next.js, and design-system-heavy products at noqta.tn — Flipbook is a provocation rather than a production tool. But it crystallises a question the industry will need to answer over the next few years: if a model can draw the entire interface, what is the role of front-end code?

Source: Flipbook