It's the evening of February 5, 2026.
I didn't do much today, got a bit of insufficient sleep, feeling a bit tired, haha.
Noticed that Chen Tianqiao has released EverMind again. His previous release, MiroThinker, was indeed very useful, but I'm not really sure what EverMind is. I'll look into it when I have time. For now, EverMind is a large AI model with built-in memory, designed based on biological bionics, attempting to mimic the memory mechanisms of the human brain. Sounds quite interesting. Given the low hallucination rate of the previous MiroThinker, I think this EverMind shouldn't be too bad either. Thanks to Chen Tianqiao for his continued investment in the AI field.
Memory mechanisms are currently an important research direction in the AI field. Next, we should see more and more AI models attempting to introduce memory mechanisms to enhance the model's long-term understanding and context retention capabilities. Hopefully, more innovative applications will emerge in the future.
Isn't CZON also an attempt at a memory mechanism? By having users write content in fragments and then handing it over to AI for integration and generation, it essentially uses users' memory fragments to build a more complete knowledge system. This approach is somewhat similar to how humans form long-term memory through continuous accumulation and integration of information.
AI is the underlying capability of CZON. CZON is one of the application scenarios for AI. If AI becomes stronger, then CZON's business performance will definitely become stronger, but at the same time, some of CZON's business logic may also be replaced by AI.
This is easy to understand: Business capability = the sum of capabilities at all levels.
- The higher the business capability, the greater the competitive advantage over peers.
- If business capability remains unchanged while underlying capabilities improve, top-level capabilities will be passively compressed, making the top-level encapsulation seem worthless. (Many software companies have had their business layers completely replaced by AI.)
- You cannot stop the improvement of underlying capabilities, nor can you prevent competitors from using stronger underlying capabilities to enhance their business capabilities.
So, what to do? Only by improving business capability itself can you survive. Moreover, the evolution of business capability itself must outpace the evolution of underlying capabilities to ensure that business capability is not compressed and the software itself retains its value.
Many AI memory mechanisms attempt to construct an AI's memory and soul. That's great, but can the human soul also be constructed? Can it be simulated by AI?
Can humans be understood as a combination of a reasoning model and a memory model?
