Read the how to on marginalia. Edit, it is right beneath the search box and says: [...] A concrete example: How do I cook steak? will probably not be helpful, Steak Recipe will give better results (just Steak is pretty good too).

After that try: python datetime timestamp

or: python convert time

or something similar. Edit: see also marginalias answer next to me.

As I wrote earlier: marginalia is most for fun (although lately I have gotten better results for at least one simple technical query.)

Edit: if a search engine respects your query you can always broaden your query. If a search engine ignores your query you cannot do anything.

Is my memory filling me or is that much closer to how search used to work in the early 00s? I have vague memories of being flabbergasted by someone even trying something like "how to XYZ?" in a search engine.

Yeah, you used to have to think of specific, unique keywords that would appear in the page you were searching for. A lot of people couldn't break the habit of asking questions, so search engines just started accepting it and trying to optimize for it.

> A lot of people couldn't break the habit of asking questions, so search engines just started accepting it and trying to optimize for it.

Which had been totally fine if they hadn't nerfed it for us who knew how to use it.

No, you are right.

This is the way anyone with a clue would approach search until Google broke it: search for words or phrases within the page.

I always search as such:

"perfect steak recipe"

"perfect grilled chicken"

and such... typically resulting in the best results.

"R53 mini cooper fuel filter"

reduce the words to the core of the question, and get good results.

I distinctly remember terming my searches as questions. I was young and this was back pre Google so I was using Ask Jeeves. Naturally if you're asking someone something you would phrase it as a question, so in my mind it made sense to ask it that way.

Wasn't that the whole premise of Ask Jeeves? And then much later, Wolfram Alpha awed with that same 'plain language' functionality?

I got access to the Kagi beta a week or so back, and have been daily-driving it as my default search engine ever since.

I've only ever "fallen back" to Google for a small handful of searches since then, and only out of curiosity for what it might look like in comparison to the Kagi results.

Although it's only been a small amount of time, I'm sticking with Kagi for now because the search results are at least on par with Google's, and significantly less encumbered by Google's SERP dark patterns.

Yeah it's not quite the sort of query it's good at. It's more geared toward discovery than answering questions. You can answer that particular question if you tune the query a bit though.

https://search.marginalia.nu/search?query=python+datetime+un...

Has this as the first result, though:

http://www.logophile.org/blog/2010/08/15/python-datetime-con...

You're right. It works better when I think "hard keyword search" than what I'm used to with DDG and Google. I tried "python super classmethod" and got here: https://legacy.python.org/dev/peps/pep-3135/ which perfectly answered my question.

Isn't it beautiful that something so amazing can run on a standard tower PC in a living room in Sweden?

Holding up against a HN hug of death last month, providing an absolutely refreshing search experience!

I just tried on Kagi, this was the result: https://imgur.com/a/Ds1YTf5 I've been using Kagi for a little while now, honestly, it's been pretty great.