I admire and applaud Kagi for what they are trying to be.
But on the AI front, the Assistant is simply worse than using for example Gemini or ChatGPT directly. It is slower, it cannot generate images etc.
Keep in mind that kagi offers a wide range of models, not just one. I wouldn't want to have multiple subscriptions (for chatgpt, anthropic, gemini etc.)
It is not worse I would say. It uses neutral system prompt by default, whereas Gemini and ChatGPT will please you too much to mislead you badly. Also the base search is much batter. You can control the search while with Gemini, for example, you can't.
I see the slowness and now the news that they might remove the assistant for the $10 plans as evidence of how costly it is to run LLMs and by extension how unsustainable it must be for OpenAi, anthropic, Microsoft etc to be offering such performance for free or very low prices. Surely something has to give soon.
The research assistant should be able to generate images
On the other hand, it doesn't monetize any of the data you give. Six of one, half dozen of the other, yeah?
Re: image gen... it's a search engine. Why would I need my refrigerator to toast my bread?
That's why I don't think anybody is going to pay for the Assistant alone.