Skip to content
Vela
Tech FrontlineBiotech & HealthPolicy & LawGrowth & LifeSpotlight
Set Interest Preferences中文
Tech Frontline

Mira Murati's 'Thinking Machines' Reveals New 'Interaction Models' for Real-Time AI Collaboration

Jason
Jason
· 2 min read
Updated May 12, 2026
A modern futuristic interface showing an AI agent seamlessly interacting with a human through real-t

A New Journey After OpenAI

Thinking Machines, the startup founded by former OpenAI CTO Mira Murati, has finally emerged from stealth mode. The company announced this week that it is focused on developing an innovative technology called 'Interaction Models.' The goal is to move the industry away from the current 'turn-based' chat paradigm toward a system that enables continuous, real-time audio and video collaboration with AI agents.

Moving Beyond 'Turn-Based' Interaction

Current generative AI tools, while powerful, predominantly function in a 'request-response' cycle: the user provides input, waits for the model to process it, and then receives an output. While this works well for static tasks, it is inherently limited for scenarios requiring natural, fluid collaboration.

Thinking Machines believes that true AI assistance requires a shift in model architecture. Interaction Models are designed to allow AI systems to 'continuously ingest' audio and video feeds. This would enable AI agents to perceive their environment in real time, responding dynamically as an situation unfolds, much like a human colleague would. As Mira Murati envisions it, AI collaboration should feel natural and ongoing, rather than transactional and disjointed.

Outlook: Technical Potential and Challenges

This research direction signals a massive shift toward AI agents that are deeply integrated into physical and complex professional environments. If successful, such models could enable AI to perform tasks that require persistent environmental awareness—ranging from industrial quality control and real-time robotic assistance to professional video conferencing integration.

However, the technical hurdles remain significant. Maintaining low-latency processing for continuous high-resolution audio and video feeds requires massive computational efficiency. Furthermore, as AI agents become capable of constant observation, Thinking Machines will have to navigate intense privacy and security concerns surrounding the always-on nature of the technology.

Looking Ahead

Thinking Machines represents the next step in the evolution of AI: from intelligence that is limited to static text or imagery, toward intelligence that is embodied, aware, and interactive. Mira Murati’s new venture may well be shaping the next era of how we work alongside machines.

Frequently Asked Questions

  • Q: What are the 'Interaction Models' being developed by Thinking Machines? A: They are a new class of AI models designed to move beyond the current 'request-response' interaction. They enable AI to persistently perceive audio and video, allowing for continuous, real-time communication with users.
  • Q: How does this improve the user experience? A: Instead of waiting for a response after every input, users will be able to collaborate with AI naturally, similar to working with a human partner, as the AI understands context dynamically through video and voice.
  • Q: What are the potential applications for this technology? A: Beyond personal assistants, these models are suited for environments that require real-time situational awareness, such as industrial quality control, smart healthcare monitoring, and collaborative professional workflows.

FAQ

What are the 'Interaction Models' being developed by Thinking Machines?

They are a new class of AI models designed to move beyond the current 'request-response' interaction. They enable AI to persistently perceive audio and video, allowing for continuous, real-time communication with users.

How does this improve the user experience?

Instead of waiting for a response after every input, users will be able to collaborate with AI naturally, similar to working with a human partner, as the AI understands context dynamically through video and voice.

What are the potential applications for this technology?

Beyond personal assistants, these models are suited for environments that require real-time situational awareness, such as industrial quality control, smart healthcare monitoring, and collaborative professional workflows.