Rick's article reminds me of the blog post "Interaction Models: A Scalable Approach to Human AI Collaboration" that Thinking Machine published a few days ago Leave the link here. If you are interested, you can go and see for yourself https://thinkingmachines.ai/blog/interaction-models/ Rick's article mentioned that in the past, we used to work on scale models, but today we are going to start on scale systems. The same basic model, combined with different strengths, may result in a difference of several tens of times in actual capabilities. From piecing together models to piecing together "model+memory+tool ecosystem+heuristic runtime+self improving harness" Thinking Machine is exploring another path, exploring the way of human-computer interaction. Today, while watching the live broadcast of Brain, he also believes that "chat dialogue box" is definitely not the end. The interaction mode of AI will definitely undergo significant changes in certain fields, giving birth to new application forms Firstly, the team background of this company is so impressive that it has exploded. Founder is a girl - Mira Murati, former CTO of OpenAI, whose valuation was exaggerated to the point where she raised 2B in the seed round, with a valuation of billions What they do is scale human AI interaction itself They believe that the essence of the current chat interface is still the DOS era: prompt → wait → output. And they believe that in the future, AI will not only be smarter, but more like a continuous collaboration system. It's not like you type a prompt, it stops and thinks for a few seconds, then spits out another answer. He should be like a colleague sitting next to you all the time: continuously listening to you, understanding your screen and context, real-time interaction in the front desk, and sneakily doing searches, reasoning, tuning tools, and executing tasks in the back desk... The front desk may use various tone words, and even interrupt you ("Hmm", "I understand") "Wait?"......), as well as visual intervention and time perception. There are videos in the link, which can be felt by the old iron workers Two months ago, we wrote an AI Chat project called Sesame that talks like a human. I don't know if you remember it, but Thinking Machine is not just about speaking, but the entire interactive experience is completely different. It's not even on the same track as Sesame Finally, let's summarize- The OpenAI article is about AI learning, which needs to continue scaling and not necessarily only happen in gradient and parameter updates Thinking Machines is saying that AI interaction should continue to scale and not stay in dialogue boxes forever Perhaps the next round of real scaling will not just be model scaling, but a systematic project of model scaling, system scaling, and interaction scaling