Vlad I think the perspective taken informs the discussion here quite a bit. You're coming it from the- entirely reasonable- position of what keeps Kagi profitable and sustainable, and it's clear the current offers are maybe closer to the wire than they should be from a business perspective, or at least to support the increased usage you are expecting if API access is permitted.
From my perspective, I hardly use a fraction of my "allotment" of AI tokens (my highest since data is available was $4.37 in July) and roughly a professional plan worth of searches on average (700-1000), so I'm something of that gym membership that only shows up a few times a week and subsidizes your margin for the heavy users.
From my perspective, if anything then the cap should be lowered somewhat to more realistically represent Kagi's costs, and so moderate/light Ultimate users like me can enjoy additional functionality (like API access; potentially new services like Kagi Mail), rather than preserve the higher caps for the heaviest users.
My use case is essentially hobbies. I don't use AI heavily enough to justify a bunch of subscriptions, and certainly not a $200USD/m general use one. If and when I need heavier use, I would rather go for a specific task-oriented service (Claude Max + Claude Code for example). I would enjoy a Kagi Assistant API to be able to spend a few dollars worth of tokens a month to connect via MCP to Obsidian, certain CLI tools, and similar applications. I have no interest in a huge subscription to hook a fire hose up to, and I'm not sure Kagi can compete there. Certainly not with the current 20% surcharge.
For me personally, especially with where the Canadian dollar is (Kagi Ultimate is $40CAD/m for me), a higher tier is right out. I'm not opposed to the concept, but if the functionality of Ultimate were to get pieced out for those hypothetical higher tiers, I'm probably back to Professional and looking elsewhere for my AI needs.
On the other hand, I could live with a more realistic monthly allotment. I think going with full plan value with 20% surcharge was generous, but has ending up being a misleading, unsustainable approach. Half the reason I originally made the jump to Ultimate was that it was effectively +$15USD for Assistant (on top of the $10USD Professional tier) vs +$20USD if I went with ChatGPT at the time. I've stayed because I appreciate Kagi generally and the access to a range of models has been more interesting to me than the features GPT and Claude have rolled out, so far.
