This AI Makes a Video Game World in 40 Milliseconds

We had Dean Leitersdorf on the pod and he did something no guest had ever done.

Mid-sentence, he transformed from a startup founder in a black t-shirt to a wizard with light shooting from his hands. Then, he was in a white-walled game universe, and when he picked up the tissue box on his table, it morphed into a gun which he could shoot by moving his arm.

He did it with one of his products, Mirage: It takes any live video feed (like Dean on the pod) and instantly renders each frame into a new style of your choosing—40 milliseconds from input to output.

Dean is the co-founder and CEO of the creators of Decart which makes Mirage. They recently raised $100 million at a $3.1 billion valuation to build a new era of real-time generative AI experiences like this.

Realtime generative video models are going to change video games forever, and Dean is on the forefront: imagine creating endless variations on existing titles, like GTA-V with a frigid winter filter, or taking a bare-bones vibe-coded prototype and using Mirage to texture it.

But games are just the beginning, Dean sees Mirage as opening the door to a new medium, a new experience created by AI.

In this episode, we take a look at how Mirage works under the hood, and what the Decart team learned about the future of software while wrestling with its toughest research problems. We also debate AGI—how close it really is, what counts as progress, and what kind of society it might create. This episode is a must watch for anyone interested in the future of gaming, creativity, or if you just want your mind blown by what’s already possible.

If you found this episode interesting, please like, subscribe, comment, and share!

Want even more?

Sign up for Every to unlock our ultimate guide to prompting ChatGPT here: https://every.ck.page/ultimate-guide-to-prompting-chatgpt. It’s usually only for paying subscribers, but you can get it here for free.

To hear more from Dan Shipper:

• Subscribe to Every: https://every.to/subscribe

• Follow him on X: https://twitter.com/danshipper

Timestamps:

• Introduction: 00:00:47

• A demo of Mirage, the first real-time video-to-video model in the world: 00:02:38

• How Mirage can take your vibe-coded game to the next level: 00:06:22

• The new architecture of modern software: 00:08:45

• How Mirage works so blazingly fast: 00:16:34

• Inside Decart’s invention of a new “live stream diffusion” model: 00:20:33

• Solving the error accumulation problem for real-time video: 00:21:17

• How Dean thinks about inventing a new creative medium: 00:29:55

• Dean’s take on the post-AGI world: 00:39:43

• Why AI brings back the age of the generalist: 00:51:15
  

Links to resources mentioned in the episode:

• Dean Leitersdorf: https://x.com/dleitersdorf?lang=en

• Decart: https://about.decart.ai/

• Try Mirage and Delulu: https://mirage.decart.ai/, https://play.google.com/store/apps/details?id=ai.decart.delulu, https://apps.apple.com/il/app/delulu-by-decart/id6749955738

• More about Yan LeCun’s error accumulation problem: https://x.com/ylecun/status/1640123182983045120