Most AI chatbots today work like text messaging. You speak or type, the AI listens, then it responds after you finish. That back-and-forth structure has become standard across tools like ChatGPT, Claude, and Gemini.
However, the Thinking Machines Lab wants to change that. The startup founded by former Open AI CTO, Mira Murati is developing what it calls “interaction models,” AI systems designed to listen, process information, and respond in real time while a conversation is still happening.
Instead of waiting for a user to completely finish speaking before generating an answer, the system is being designed to operate more like a real human conversation or phone call, where both sides can process information simultaneously.
According to the company, current AI systems operate in a “single-thread” interaction model. This means the AI stops receiving information while generating a response, creating what Thinking Machines describes as a communication bottleneck.
Thinking Machines believes real-time interaction can make AI collaboration feel more natural and useful across voice, video, and text.
The company demonstrated several examples of how the technology could work. In one case, the AI listened to a story and reacted when animals were mentioned. In another, it translated speech in real time. It also showed an example where the system detected posture issues and warned a user about slouching during interaction.
The technology is based on what researchers describe as “full-duplex” interaction. This allows the AI to continuously receive and process information while speaking, instead of alternating between listening and responding. Reports suggest the system currently operates with latency close to natural human conversation speed.
This is important because one of the biggest frustrations with modern AI systems is the unnatural pause between user input and machine response. Even highly advanced chatbots still feel transactional rather than conversational.
Thinking Machines is trying to reduce that gap by making AI interaction more fluid and continuous.
The company has not yet released the system publicly. However, it said a limited research preview is expected in the coming months, followed by a broader release later this year.
The development also highlights how competition in the AI industry is evolving. Companies are no longer focused only on making models smarter. There is increasing attention on making AI feel more human, responsive, and naturally interactive.
Thinking Machines has quickly become one of the most watched startups in the AI industry since Murati left OpenAI and launched the company in 2025. The startup has attracted former researchers and engineers from OpenAI, Meta, and Anthropic, while also securing major infrastructure partnerships with Nvidia.
Furthermore, the company’s work reflects a broader industry direction toward AI agents and real-time multimodal systems capable of handling voice, vision, text, and live interaction simultaneously.
Still, some people remain skeptical. Reactions online show that many users believe current large language models still lack true conversational understanding and may struggle to deliver genuinely natural interaction.
For now, Thinking Machines is positioning itself around a simple idea, AI should not feel like sending messages back and forth. It should feel closer to talking to another person.
Read also: Rice University Develops AI-Powered Handheld Microscope for Earlier Cancer Detection