Anastasis Germanidis

Towards Universal Simulation agermanidis.com/writings/unive…

EXCLUSIVE: Runway is one of the most interesting inflection point stories in AI right now. Started as a tool for filmmakers. Now going after world models/ universe simulation, and by extension, everything from robotics to (one day) big bets such as anti-aging research. 🧵 Link👇

Agent was shaped by what we kept hearing from the most effective teams inside agencies and in-house brands using @runwayml every day. -A games studio went from 13 ads/week to ~100, going from idea to ad in under an hour, directly driving installs and revenue. -A professional services studio built a PowerPoint-to-video tool on our API, replicating $300-600K campaigns for ~$3K. -A global agency's pharma team built an AI animatic in Runway that outscored every ad in the client's history. We tried to capture how these teams actually work and make it the default in Agent.

ICYMI: Over the past two weeks, we've published a series of in-depth looks into how we deploy frontier video and world models into production, across engineering, research, and trust and safety. From how we optimized GWM-1 to serve Characters (which I believe is one of the largest deployments of real-time autoregressive diffusion today), to reducing cold start times by 60x for our inference workers via GPU-to-GPU transfer of model weights, to open-sourcing our configuration system for training and evaluation configs, and finally, the continuous work we do to ensure child safety across our platforms. There are a lot of novel challenges we face as Runway increasingly becomes the world's infrastructure for generating pixels, media, and worlds, and we'll start sharing more of the behind-the-scenes in the weeks and months ahead.

Runway added more than $40M in net new ARR so far this quarter, and we're less than halfway through. The biggest growth period in the history of the company. Generative video has hit its inflection point. Enterprises like Amazon and Robinhood are using Runway daily to reshape how they tell stories and reach customers, on our frontier video and world models. Thank you to our team and our customers. The best is yet to come.

@steipete @RajaPatnaik @openclaw @runwayml On it

@agermanidis @RajaPatnaik @openclaw @runwayml Testing would be great! Just landed support on main! Or add some creds to the steipete gmail account. :)

The next version of @OpenClaw comes with native video generation. To start, I added support for the following companies: - Alibaba - BytePlus - fal - Google - MiniMax - OpenAI - Qwen - Together - xAI docs.openclaw.ai/tools/video-ge…

@steipete @RajaPatnaik @openclaw @runwayml Excited for it! Let me know if we can help support

@RajaPatnaik @openclaw @runwayml That's coming rn.

Talk to any Runway Character on the phone! It can see the phone's camera and answer your questions.

Excited to be partnering with great founders pushing the frontier of AI, whether that's research, infrastructure, applications, or entirely new kinds of media. If that's where you're building, we want to hear from you.

@jon_barron Agreed, my point is that if you think that video models are important to invest in as a technology, then you figure the applications that make sense for each level of cost and capability, and you focus on improving that frontier to make more applications viable

@agermanidis The "too expensive to deploy" argument is totally contingent on the application IMHO. If millions of people are gonna pay $15 to see your generated video in a theater, cost is absolutely a non-issue. But if the video only gets seen by one person one time, cost is a huge problem.

The idea that video models are too expensive to deploy does not make sense to me, and it usually comes from organizations that don’t understand the potential and surface area of the technology. If you believe, as I do, that video models are on the critical path to general intelligence, then it becomes a P0 problem to solve. It’s important to advance capabilities and efficiency at the same time. Improvements in efficiency help us deploy the models more widely and train them faster. People associate Runway with bringing new capabilities and control to video generation, but our work in making those models more efficient has been just as critical to getting them in front of tens of millions of users and 90% of the Fortune 100. We've invested in making video diffusion fast for a long time, bringing a distilled video model to market (Gen-3 Alpha Turbo) before anyone else. And we recently had to build an entirely new model serving stack from scratch to support real-time interactivity and instant generation as we start to deploy GWM-1 and our next world models. General, interactive world models will have a huge multiple of impact compared to traditional video models, and to get there you need a research organization that both understands how important those models are, and is insanely obsessive about getting the low-level details right. Here's a good place to contact such an organization: runwayml.com/careers

Three years ago, the beginning of the wild journey towards building a world simulator. A model that still holds a very special place in my heart. I feel the same sense of open possibility that I felt three years ago, now with real-time video.

@jparkerholder @taiuti Correct.

@taiuti I think a realtime avatar model is more closely aligned with the original world model definition than a Gaussian splat

I cannot believe everyone is quoting an article that is so wrong, it says that avatar models are world models. Did you people read it or are you just riding the next cool wave?

Learn how to create your own Runway Character via the Runway API. And start bringing real-time video agents directly into your apps, products, websites and experiences. Get started at the link below.

Building new products and applications has completely changed. Today, more than ever, it is crucial to ask ourselves what the next generation of applications for generative video will be, how we are building them, and how this will impact every single industry. I’m very excited to start Runway Labs, an internal incubator dedicated to exploring new products at the frontier of generative video and General World Models. We are at a very special moment in time to define how this future will look. I’m hiring an initial team to build with: design engineers, AI engineers, and builders who want to create new kinds of experiences and products that haven’t been possible before. If this sounds interesting to you, reach out.

btw it took ~5 min to set up this demo, most of which was waiting for Gemini to generate a few Ghibli image of me, while having ChatGPT write a prompt based on what it knows about me (I’ve fed a lot of my granola transcriptions into it)