I find the Kagi Assistant's performance is affected by noticeable delays before search initiation or output generation, suggesting that preprocessing is occurring to determine the necessity and specifics of a Kagi search.
In contrast, services like Perplexity Pro and Google Gemini DeepResearch decompose research tasks into multiple steps, executing several search queries for each. This enables them to explore multiple topics within a single request and refine results through varied search queries.
The Kagi Assistant's current inability to perform such multi-faceted exploration makes it a somewhat frustrating experience. Enhancing its ability to handle complex queries and conduct iterative searches would significantly improve its utility.
The overall use of the Assistant wouldn't need to change drastically. However, introducing a step-by-step view of the process could be particularly helpful for handling multi-step requests. This feature would allow users to see the progress of their query and the specific search queries being used.
Even without this level of detail, a straightforward enhancement to the quality and relevance of the output in the first iteration would already be a significant improvement.