Sakana, an artificial intelligence startup headquartered in Tokyo and co-founded by former leading Google AI scientists, including Llion Jones, has announced the development of a novel AI architecture named 'Continuous Thought Machines' (CTM). The CTM architecture aims to advance the reasoning capabilities of AI models by enabling them to operate with less explicit instruction, drawing inspiration from the way human brains process information and solve problems. Unlike traditional AI models that typically require detailed prompts or step-by-step guidance, CTM is designed to allow models to engage in more autonomous and flexible reasoning. This approach could potentially lead to AI systems that are better at handling complex, open-ended tasks and adapting to new situations. However, Sakana emphasizes that CTM remains a research-focused architecture and is not yet suitable for immediate production use. The company’s announcement has generated interest in the AI research community, particularly given the founders’ backgrounds and the ongoing quest to make AI systems more general and human-like in their reasoning abilities. The development reflects a broader trend in AI research toward architectures that can generalize better and require less human intervention.