I made a minor point in my preceding blog post about Google web search:

For basic research, Wikipedia exists and is free to use, donor-supported. For precision querying, Google is free to use, ad-supported. Ad-free alternatives like Kagi also exist. Existing research summaries on Wikipedia and primary sources have a much higher utility to me, and can convey more nuance than language model output. They also don't use any of the forced analogies that language models are so fond of including.

However, Google web search has an "AI overview," based on the Gemini model family. It utilizes the language model to synthesize the most-relevant search results into a brief summary at the top of the page. Here are some examples which briefly show the utility of this:

The problem is, even if I cut off all of my paid provider and conscious access to LLMs, I still find myself encountering these AI overviews, and my behavior patterns continued because I didn't cut myself off of this LLM. Therefore, I must stop using Google search.

I believe DuckDuckGo provides search results of sufficient quality to serve as a replacement. While the website has a similar LLM-powered "Search Assist", it is very brief and I have observed it does not meaningfully feed my fixations:

(The hexdump example is actually more helpful because it points to a standard mechanism for this task.)

DuckDuckGo also offers the ability to completely disable Search Assist, so it requires my conscious action to enable it again. Google only offers it as a hidden URL parameter (&udm=14).

For now, I will be switching to DuckDuckGo, with Search Assist turned off.