Hi there
i use kimi daily and i have noticed something that will make a notable difference in the model or the website whatever it is related to , simply kimi needs a system that runs linux , windows commands combined in the website host for kimi ai models something like cmder which offers win , linux commands , well it is the same idea but the difference is that kimi is going to run it and actually understand the output and the command , then based on the output it works on the user’s order / favor, that will make a huge difference in his real time intelligence
another thing is that it has to memorize the whole session no matter how long the session is , it just remembers the whole thing so that any question from the user will make sense to the AI model as it remembers the whole conversation from the beginning
And yeah this is a repost since the original messge that is important got ignored and i am sure that if this issue is solved , kimi will get to another place , with quota extend and agent mode to be usable for free tier users
Thanks for kimi team , and i will be always here to give suggestion to lead kimi to be better
Best ,
Jimmy
1 Like
I did personally suggest the memorization issue on a previous post of mine! Spot on. I would just assume that balancing out memory to an entire chat, to the speed it currently exhibits, is a complex improvement. However, I would say that if this is solved/evolved, for tasks which, rather than requiring many tokens to ‘figure out’ the solution again, they could consume far less tokens through an enhanced memory/pattern recognition later.
1 Like
Hi Jimmy,
Thank you for the fantastic feedback and for using Kimi daily!
Your idea of an AI model that can execute commands, read the outputs, and maintain a continuous, long-term session is exactly what we are focusing on right now. Have you considered trying out Kimi CLI and Kimi Claw? They are built precisely around this “Agent mode” concept, allowing Kimi to run tasks and remember context over long sessions. Does this sound like it matches your needs?
Our engineering and research teams are actively pushing the boundaries to improve these capabilities even further. We’d absolutely love for you to test them out and let us know what you think.
Thanks again for helping us lead Kimi in the right direction!
Bests,
Yu
1 Like
Well tbh i didnt try kimi cli or claw since they require subscribtion and i am a free tier user , about the agent mode in the time i made that thread it was telling me to upgrade my plan to get in the queue , but i have tried that right now , and … surprisingly it works now , i hope to extend the quota (tokens per user) so because as i said in my first forum kimi is better for me than any other ai model it just needs time to get at the top 3 ai models , i guess that inheriting features from claude will make it perfect at least a bit better than what it is right now , also i forgot to say that the point from the thread is adding the commands execution , etc … in all types of kimi models , instant , thinking
so that is matches claude and even beats it
Best,
jimmy
1 Like