Why I hate search

"javascript remove part of an array"

Why when I "search" for that, do I have to parse, in my brain, titles, descriptions (and so on) just to get my answer? I've searched for this a few times before,I know, because I'm terrible at remembering particular functions across languages.

Why can't my browser know when I've found the answer before? Hell, why can't my browser even offer me an option to remember that answer. I'll be happy to highlight it for it, but it's clunky and clumsy with a dumb bookmark. No nuance. Nothing was solved. And god forbid I bookmark every answer I look for; try searching through that.

If it's specifically previously answered queries you're interested in, maybe something like OneNote would be a better solution? Then you could search your notes and probably find the old answer with a lot less noise.

Let's say I'm in the mood for a donut, but I can't remember how late my favorite donut place is open (as if, but bear with me).

If I had a knowledgable human at my disposal, I could ask them "When does that donut place close?"

They might answer that "they're open for another hour."

There are two important differences between this and search:

First, my friend uses contextual data s/he knows about me to determine what specifically it is that I want to know, and takes my "query" in natural language. Services like Siri already go a long way toward making this a reality, and good natural language search queries have been quite usable for years now.

Second, the answer is a natural language representation of the information I was looking for, rather than a pointer to where I can find it. Wolfram Alpha does a pretty good job of this if your query is about math or science or some other type of data that can be relatively straightforwardly curated.

If I do the same with Google, I get a Yahoo Answers page asking why donut places close at noon. If I'm a bit more specific and ask "When does Rocket Donuts close?", the top result is, helpfully, their home page. I then click the result and look around the page for their hours, and finally find them at the bottom of the screen. Then I have to remember what day it is, what time it is, and subtract that from the appropriate number.

I also don't have much information about how current their information is, even though the web server helpfully responds that the page was last updated around noon on April 2nd.

One approach would be the so-called "Semantic Web", but I believe getting a majority of unsophisticated web publishers to reliably include semantic information is a fool's errand.

What's really missing is a way to index the web not by text, but by a machine's best guess at the interpretation of the of the text. Basically index a bunch of "facts" on the web, including who is claiming that they're true, and how long ago they made that claim. So the "key" would be something like "Rocket Donuts, Bellingham, WA, USA: Closing time on Thursday" and you would get a bunch of results, but you would sort by authority (something like PageRank) and recency (both descending). Hopefully the top value would be "5:00PM PDT".

Obviously one does not simply walk in and solve this kind of problem, but I believe whoever does would have a shot at replacing "search" as we know it.

The search syntax was also pretty advanced. You had parenthetical grouping, logical operators, and if I recall, the +/- include/exclude syntax originated with (or was at least used by) AV.

What AV wasn't was particularly relevant. You needed that advanced search syntax to be able to narrow down what you were looking for, and you still usually had to go through a few pages of results to find the right stuff.

When Google first hit public beta, it was immediately and compellingly superior to anyone else's search. I remember giving it a few tries, and switching within a matter of a week or so. It just gave me what I wanted.

It took a long time for anyone else to get off the "but we need to keep users on our pages for ad clicks" mindset, and by then it was too late.