Mira Murati Is Building AI Models That Interact in Real Time, All at Once

Mira Murati has introduced interaction models, AI systems that can perceive audio, video and text at the same time and respond in real time. Today's models wait until the user has finished speaking; the new ones are meant to interact as naturally as people talk to each other. Murati, OpenAI's former chief technology officer, founded Thinking Machines and is positioning interaction models as a new approach to how people work with AI.

AI-processed from The Verge; edited by Hamidun News
Mira Murati Is Building AI Models That Interact in Real Time, All at Once
Source: The Verge. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceThe Verge ↗
Editor-in-Chief
Naturalnessdialogue with AI becomes as smooth as a conversation with a person
Contextthe model sees tone of voice, body language and visual context, not just words
Speedno delay between the user's thought and the model's reaction

Mira Murati, OpenAI's former chief technology officer, founded Thinking Machines and has unveiled a new paradigm for AI interaction. The company is working on interaction models: systems that perceive audio, video and text simultaneously and respond in real time, imitating natural collaboration between people.

A new paradigm for interaction

Interaction models are not just new versions of language models. They represent a fundamentally different approach to how people interact with AI. Instead of typing a request and waiting for an answer, a user can speak, show their screen and change context, while the model reacts in parallel, like a partner in a natural conversation.

According to Thinking Machines, current models perceive reality "in a single stream." While the user speaks or types, the model is passive: it sees nothing, hears nothing and does not understand what is happening. Processing begins only once the information has fully arrived. This creates lag and makes dialogue feel unnatural.

Interaction models are supposed to work differently: continuously take in the incoming stream, think in parallel with the user and respond in real time. That is closer to how people talk, where both participants listen, process information and respond at the same time.

What changes

The shift from sequential to parallel interaction has serious consequences:

  • Naturalness: dialogue with AI becomes as smooth as a conversation with a person
  • Context: the model sees tone of voice, body language and visual context, not just words
  • Speed: no delay between the user's thought and the model's reaction
  • Multitasking: the model can listen, watch, analyze and respond at the same time
  • Adaptability: the system can notice that the user is confused and help right away

History and significance

Mira Murati left OpenAI in September 2024. Within a few months she assembled a team and launched Thinking Machines with a clear goal: to redefine what "collaboration with AI" means. For her, this is not just another startup but an attempt to solve a fundamental problem of modern AI interfaces.

If Thinking Machines manages to deliver on this idea, it could change how people work with AI in every area, from professional work to education and creative projects.

What this means

Interaction models are the next step in the evolution of AI interaction. If current models are chatbots, interaction models are something close to a live conversation.

For users, that means more natural communication with AI. For businesses, it means the chance to build interfaces that people do not need to be trained to use with special syntaxes and commands.

Story

Key Questions Answered

What are the key details on A new paradigm for interaction?

Interaction models are not just new versions of language models. They represent a fundamentally different approach to how people interact with AI. Instead of typing a request and waiting for an answer, a user can speak, show their screen and change context, while the model reacts in parallel, like a partner in a natural conversation. According to Thinking Machines, current models perceive reality "in a single stream." While the user speaks or types, the model is passive: it sees nothing, hears nothing and does not understand what is happening. Processing begins only once th

What changes?

The shift from sequential to parallel interaction has serious consequences: Naturalness: dialogue with AI becomes as smooth as a conversation with a person Context: the model sees tone of voice, body language and visual context, not just words Speed: no delay between the user's thought and the model's reaction Multitasking: the model can listen, watch, analyze and respond at the same time Adaptability: the system can notice that the user is confused and help right away

What are the key details on History and significance?

Mira Murati left OpenAI in September 2024. Within a few months she assembled a team and launched Thinking Machines with a clear goal: to redefine what "collaboration with AI" means. For her, this is not just another startup but an attempt to solve a fundamental problem of modern AI interfaces. If Thinking Machines manages to deliver on this idea, it could change how people work with AI in every area, from professional work to education and creative projects.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Unified AI Workspace · Nomad

Test Frontier LLMs & Autonomous Agents in One Workspace

Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.

Try Nomad AI →Launch Nomad Agent
What do you think?
Loading comments…