Kagi AI Assistant: a short review after one year

A year ago, I switched my default search engine to Kagi. This is a paid search engine that is removing a lot of the crap results that I got on Google (ads, slope coming from SEO-generated content, etc.). It’s not cheap, but their value proposition makes it worthwhile.

They added an AI assistant that can be invoked by adding a question mark at the end of your search query. But the real kicker is their full AI Assistant. I have access to a lot of models included in my premium subscription. I rarely used all of them, but I have my preferences:

• Claude 4 Sonnet for software-related questions
• Mistral when searching in French or French content
• OpenAI o3 pro (insert here a rant about their obscure naming convention) when I need a « deep » search on a question

I tested o3 pro after listening to an episode of Odd Lots about the impact of AI on law firms.

I wanted to have a summary of the third-party APIs that I can use in my project. It was not only a software question but also a business one because I need to be able to monetize the results I would get. I got a pretty decent summary with a table listing the pros and cons of each solution, the reasoning behind the ranking, and a conclusion in the form of try this and then that if it failed.