Mira Murati, the former CTO of OpenAI, has founded a new AI company called Thinking Machines that announced on May 11, 2026, a new approach to artificial intelligence called "interaction models." According to the company, these interaction models will allow people to collaborate with AI in a more natural way, continuously processing audio, video, and text while thinking, responding, and acting in real time. Thinking Machines argues that current AI systems operate with significant limitations. The company explains that today's models wait until users finish typing or speaking before responding, and during the AI's response generation, its perception essentially freezes, receiving no new information until it completes or is interrupted. Thinking Machines calls this a "bandwidth bottleneck" that restricts how much of a person's knowledge, intent, and judgment can reach the model, and how much of the model's work can be understood by the user. The company compares this limitation to trying to resolve an important disagreement over email instead of in person. Thinking Machines demonstrated several examples of interaction models in action, including a system that listens for mentions of animals in a story, real-time speech translation, and an application that alerts users when they're slouching. The company believes this technology enables AI interfaces to meet humans on their own terms rather than forcing users to adapt to AI limitations. The technology is not yet available to the public. Thinking Machines plans to launch a limited research preview in the coming months and aims for a wider release later in 2026. A more detailed explanation of interaction models is available on the Thinking Machines website.