Welcome, humans.

Runway spent years teaching AI to generate video. Now it is asking a stranger question: what if software itself behaved like video?

Solaris is Runway’s new “interface world model.” Instead of writing HTML and CSS, then rendering a page, it generates the interface frame by frame and reacts directly to clicks, drags, and keyboard input. In other words, the app is not really “coded” in the normal sense. Your browser tab is becoming a tiny simulation.

Here’s our favorite parts:

• (06:12) What if software skipped code entirely? Solaris learns the mapping from user action to visual result and renders pixels directly instead of relying on HTML/CSS as the middle layer.

• (10:45) Your interface could actually adapt to you: Anastasis imagines software that changes its density, visual style, and layout around individual preferences instead of forcing everyone through the same UI.

• (22:14) The likely architecture is hybrid: A lightweight local model could render at high frame rates while a larger cloud model periodically supplies richer context.

• (27:23) World models are already helping robotics: Runway tested robot policies inside simulation and reported a 0.95 correlation with how those policies ranked in the real world.

• (44:43) The three-year bet: Anastasis expects world models in production robotics and thinks a meaningful share of everyday software could be generated on the fly.

The big idea is not “video models can make websites.” It is that a model good enough to predict what happens next might also become the thing that runs the interface itself.

Why watch this? Because “world model” gets used to mean everything and nothing. Anastasis gives a clean definition, then walks through what the idea changes for software, games, robots, and eventually the operating system itself.

Watch and/or Listen now: YouTube | Spotify | Apple Podcasts

P.S. Around 17:32 we somehow arrive at the obvious next benchmark: holograms. Runway has not yet run Solaris on one. Grant has assigned homework.

Keep scrolling for a plain-English explanation of how Solaris works, the Dell + NVIDIA guide to taking AI from pilot to production, and recent episodes worth your time.

THIS EPISODE WAS BROUGHT TO YOU BY…

The Enterprise Guide to Scalable AI

Plenty of companies can launch an AI pilot. Far fewer know how to turn that pilot into something secure, scalable, and useful in everyday work. Explore “The Enterprise Guide to Scalable AI,” sponsored by Dell Technologies and NVIDIA.

The hub looks at what changes when AI moves from pilots into production, including how teams prepare data, choose infrastructure, run agents closer to users, and keep AI systems governed as they scale.

Additional Resources: How Solaris runs an interface without normal code

Here is the mental model: instead of asking code to determine every state transition, Solaris predicts the next visual state directly.

• The input is your action. A click, drag, keystroke, or other interaction becomes conditioning for the next frame.

• The model keeps recent visual history. That gives it context about what has already happened in the interface.

• It predicts pixels, not HTML. Solaris learns that clicking a button should produce a certain visual result without someone explicitly programming that behavior.

• An LLM can still sit behind it. Runway expects the world model to handle the interactive front end while an LLM or agent handles things like tool calls, logic, and longer-lived state.

• The long-term goal is one foundation model. Today Runway still specializes variants for interfaces, games, and robotics. Anastasis expects those capabilities to converge over time.

The current bottlenecks are exactly what you would expect from something pretending to be software: persistence and latency. It has to remember what happened before, and even half a second of delay can make an interface feel broken. Read Runway’s Solaris announcement here.

Want to see the world-model idea in action? Runway has two official demos worth watching:

• Solaris, Runway’s first Interface World Model: the clearest visual demo of the idea above, with the interface itself generated and changing in response to user input.

• GWM Worlds 2, Runway’s playable world model: a look at the same broader world-model bet applied to interactive, playable environments instead of software interfaces.

UPCOMING EVENT: The Neuron IRL in San Francisco

The Neuron is going IRL in San Francisco on November 18 starting at 4:30 PM PT. If you’re in town, come hang out with us for drinks, bites, and a live recording of The Neuron podcast. We’ll have The Neuron community together in one room, plus some special guests and plenty of AI conversation.

Huge thanks to Slack and Alumni Ventures for helping us bring the night to life. It’s free, but space is limited, so RSVP and save your spot.

🎙️ In Case You Missed It…

Two recent interviews and episodes we think you’ll love.

1. Can AI Help Cure Genetic Disease?

TL;DW: Mammoth Biosciences CEO Trevor Martin explains how AI and CRISPR are converging on a much bigger promise: turning some diseases people manage for life into treatments they may only need once. He also breaks down how Mammoth searches billions of proteins for smaller gene-editing systems.

Why you should watch: Trevor makes CRISPR understandable from first principles, then separates what AI is already useful for from the clinical-data bottlenecks medicine still has to solve.

Watch / Listen: YouTube | Spotify | Apple Podcasts

2. Does Your AI Voice Assistant Actually Understand You?

TL;DW: Hume AI CEO Andrew Ettinger explains why a voice assistant can sound perfectly human and still misunderstand what you mean. When a model reduces speech to a transcript, it can miss tone, emotion, pauses, and other clues that change the meaning of a conversation.

Why you should watch: If you use voice assistants or build voice agents, Andrew shows why a natural-sounding voice is only half the problem, and what real-world testing needs to measure before we can trust these systems with longer conversations.

Watch / Listen: YouTube | Spotify | Apple Podcasts

Have a topic you want to learn about? Request it here!

Subscribe to our YouTube Channel for more!

Subscribe on YouTube to help us bring in more builders, researchers, and guests who can teach you something useful about AI every week.

Stay curious,

The Neuron Team

P.P.S: Love the newsletter, but don’t want these podcast and livestream announcement emails? Don’t unsubscribe. Adjust your preferences to opt out of them here instead.