Here’s what Mira Murati’s AI company is up to

What it does

Thinking Machines, the AI startup led by former OpenAI CTO Mira Murati, is developing “interaction models.” These models continuously process audio, video, and text inputs simultaneously. Unlike current AI that processes inputs in isolated chunks or single shots, interaction models aim to replicate human-like, ongoing collaboration with AI. The system will think, respond, and act in real time, maintaining context across multiple input types and time.

Why it matters

Current AI models treat interactions as discrete moments rather than ongoing conversations. That makes integration with live workflows clunky and limits natural collaboration. Thinking Machines’ approach could reduce friction in AI-human interaction by allowing AI to react dynamically as a human partner would. This pushes AI closer to continuous, multimodal comprehension and responsiveness. For operators and builders, this could mean smoother AI agents embedded in live environments like meetings, customer service, or real-time surveillance systems, where video and audio context matter.

Who it is for

The technology targets real-time, context-sensitive AI applications. Builders creating AI agents for complex environments—such as video conferencing, augmented reality, or autonomous systems—may get a new toolset for richer, continuous input handling. Founders and investors should watch this as an indicator of AI heading toward seamless multimodal, interactive experiences rather than isolated query-response models.

The catch

The technology aims high but faces challenges in processing speed, accuracy, and reliability with continuous multimodal input. Combining audio, video, and text in real time without significant lag or errors remains technically demanding and resource intensive. The company will need to prove it can deliver robust, scalable models that work outside of labs and in messy real-world environments.

What to watch next

Monitor Thinking Machines’ progress on demonstrating working prototypes or partnerships deploying interaction models. Adoption or pilot programs in industries requiring real-time collaboration, like remote work tools, customer engagement, or smart environments, will indicate viability. Also, watch how the company’s approach compares with larger AI players focused on single-modality models or episodic interaction modes.

AI Quick Briefs Editorial Desk