I am posting this because I realized that it will be a year since I first subscribed to Kagi more for the LLM access than the Kagi search. I felt that I needed to relay the helplessness that I felt when using the Kagi Assistant and why I left in frustration.
I was a subscriber to Kagi Pro for 4 months in late 2025. I quit because of two things. My frustration with Kagi Assistant and the poor results I was having w various LLMs that was no better than free versions of these LLMs. When I posted about my issues in your discord I was told by one user "You are using in wrong." That was the extent of the help I received. No suggestions etc. I quit shortly after that. It seems Kagi Assistant and the interactions on your Discord was at least from my point of view aimed towards advanced users of AI rather than people starting out or trying to get their feet wet so to speak. I felt lost and that only added to my frustration level.
It seems that Kagi is marketing the search to everyone out there whereas Kagi Assistant and the LLM are being marketed to more advanced users. The newbies like me are left out in the cold and allowed to become more frustrated and confused. I wonder how many other subscribers former and existing are just throwing up their hands with Kagi Assistant and what in my opinion is lack of basic help with those of us who get over their heads when it comes to using LLMs.
I am debating on whether to try Ultimate which is not cheap here in Canada with taxes it is close to $40 CAD a month. That pales to other LLMs subscriptions but if nothing has changed at Kagi I wonder if I need the added frustration.
Perhaps add a separate channel on your discord for people like me who are newbies to LLMs but are more than willing to learn but need assistance solving issues. You never know whether there will be some diamond in the rough. People get in over their heads and need basic explanations but may feel asking in the Kagi Assistant embarrassed to ask there due to more involved topics being covered.
From this former user it seems that there is very little explanation on which models are suited for general research and which are better for coding.
If Kagi want to tailor their Kagi Assistant and access to various LLMs to more advanced users then pardon the interruption.