Tufte CSS VS Searx

As it is often said in various designer forums, please avoid pure white (#FFFFFF) on black (here #111111), as it makes the text glow for the human eyes (therefore making it unreadable for long text). Instead, try to lower a bit the contrast on the text color.

Also, the dispositions for the buttons at the beginning (GitHub, NPM, ...) are not adjusted correctly for keyboard navigation (each button requires two tabs).

Appart from that, I do like a minimalist stylesheet, so I will also recommend Tufte CSS [0] for readers.

[0]: https://github.com/edwardtufte/tufte-css

CSS and HTML is great for documents - not so great for applications. Most sites end up implementing their own navigation UI/UX at a minimum (an application) - many end up as more applications.

For an example of "documents", see eg:

https://edwardtufte.github.io/tufte-css/

Or

https://alistapart.com/article/building-books-with-css3/

Just for kicks, I applied Tufte CSS[0] onto the page using a browser extension. Great ROI.

0. https://edwardtufte.github.io/tufte-css/

It's not the answer you're looking for, but https://edwardtufte.github.io/tufte-css/ might be of interest to you

A famous physicist, Richard Feynman, had a practice in his his extensive writing, e.g. a three-volume physics text, to use the methodology of sentences which then cumulate sequentially into paragraphs. (ht Ed Tufte)

I think searx was largely built by a single person.

https://github.com/searx/searx

I use Recoll to index all of it. Recoll WebUI exposes an API, which I plugged into Searx.

The quality of Google search results has been awful for many years now, but if you still want to use it, the only usable way is via a frontend like Searx[1]. Using any of Google's frontends for any of their services is an exercise in frustration from dodging ads and fighting their hostile UI.

[1]: https://github.com/searx/searx

RSS-Bridge has a Google search adapter. You can also fake it with Searx (which offers RSS feeds of search results).

I've recently wondered how effective local models were at chaining together thoughts as proposed in https://ofir.io/self-ask.pdf. Turns out they are indeed capable of doing so while also creating reasonable chains of thoughts that are easily as good as OpenAI's models. To make it completely free to run I used SearX running inside a Docker container with a second model curating the search results for the main model to get answers from the web.