Chase Small

The whole point of building Tavus has been simple: talking to a machine should feel as natural as talking to a friend or coworker. It’s hard to describe all the tiny nuances that make a conversation feel human. The little expressions. Moving around in your chair. Knowing when to speak and when to listen. The dance of it all. Griffin is by far the closest anyone has come to a model that can capture those nuances. The first time I saw it being used, I had no idea I was watching our model rather than just a normal video call. I’m so incredibly proud of this team and what they’ve built.

We're excited to release the State of Generative Media Report, Volume 2. Generative media has moved from experimentation into production and that shift looks different across every industry. Built from fal’s own platform data, the report explores how AI image, video, audio and 3D are evolving, written by the people at fal building the models, infrastructure and products behind the shift.

There are a bunch of AI safety takes being slung around on X right now. One side thinks the US government can regulate AI aggressively enough to keep the apocalyptic scenarios from becoming reality. The other thinks Anthropic and OpenAI are lobbying for regulatory capture or—in OpenAI’s case—looking for a convenient way to take some heat off their IPO ambitions. I think both sides are arguing around a much harder problem: AI is a global race, particularly between the US and China, and slowing development is a massive collective action problem. If you assume the AI safety people are right, their solution requires an extraordinary amount of cooperation. OpenAI, Anthropic, Google, and the neolabs can all agree to slow down. They can convince Washington to regulate them. But none of that matters indefinitely if another major player keeps going. And China is not going to stop developing AI because Anthropic thinks the p(doom) is too high. The US has spent years trying to preserve its technological advantage by restricting China’s access to advanced chips and semiconductor technology. Those restrictions can buy time, but buying time is very different from permanently preventing technological progress. China has already closed much of the model-performance gap, while chips improve, models become more efficient, and techniques inevitably diffuse. Eventually China is likely to reach capabilities that OpenAI and Anthropic have today, and then capabilities beyond them. If we’ve spent the intervening years constraining American labs while China continues advancing, the political and national security pressure to restart development will be enormous. The more strategically important AI becomes, the stronger the incentive to defect becomes. So any serious proposal to slow frontier AI ultimately requires international coordination durable enough to survive a US-China technological competition in which the potential prize is economic and military superiority. We are nowhere close to having that.