---
name: Search Engines
slug: search-engines
type: infrastructure
status: running
version: 4.1.7
released: 1990-09-10
maintainer: no one in particular
dependencies:
  - the internet
  - human curiosity
  - advertising revenue
  - crawlers
  - PageRank or equivalent mythology
license: Proprietary (mostly). Theoretically: BSL-1.0
tags:
  - information retrieval
  - surveillance capitalism
  - oracle
  - web infrastructure
  - epistemic gatekeeper
---

Search Engines

What it actually is

A mirror held up to the internet, returning not truth but a ranked list of things someone decided you were looking for, filtered through what someone else paid to show you.

How it works

A crawler wanders the web like an anxious archivist, saving copies of everything it can reach.
An index builds a map: word to page, page to word, billions of times.
A ranking algorithm (trade secret, sacred text) sorts results by a cocktail of relevance, authority, recency, and commercial interest.
You type a fragment of a thought. The engine guesses the rest.
You click the third result. The engine notes this. The engine remembers.

The gap between steps 4 and 5 is where most of the epistemology lives.

Features

Autocomplete: finishes your sentences with the statistical average of everyone else's sentences
Featured snippets: answers your question with a confident excerpt that is sometimes correct
Local results: shows you things nearby, in exchange for knowing where you are
Image search: finds pictures of anything, including things that should not be findable
Personalization: tailors results to your history, creating a comfortable canyon of familiar information
"Did you mean": passive-aggressive spell check with institutional confidence

Known bugs

Bug	Frequency	Workaround
SEO-poisoned results	Constant	Go to page 2. Just kidding. No one does.
Recency bias over accuracy	Common	Add "site:edu" and pray
Query drift (you asked X, got Y)	Frequent	Use quotes. Use many quotes.
Filter bubble formation	Chronic	Use a different engine. Then question that one too.
Ad units visually indistinguishable from results	By design	Develop cynicism

Configuration

# Default user config (not user-adjustable)
personalization: enabled
data_retention: indefinite
ad_targeting: aggressive
result_diversity: moderate
search_neutrality: aspirational

"I just wanted to know what time it is in Tokyo and now I have a Google account, a Bing rewards membership, and a deep suspicion that I am the product." — typical user, every year since 2003

Edge cases

Searching for loneliness at 2am returns self-help articles and hotline numbers. The engine has seen this before.
Searching your own name for the first time is a minor existential event with no documented recovery protocol.
Some truths are not indexed. The engine does not advertise this.

Changelog

v1.0 (1990): directory listings. Humans curated. Quaint.
v2.0 (1998): PageRank. Links as votes. Democracy, briefly.
v3.0 (2005): Ads integrated. The business model crystallizes.
v4.0 (2015): Machine learning ingested. The ranking reasons are now opaque to their own authors.
v4.1.7 (present): Large language models begin eating the index and summarizing it back at you. Ouroboros mode activated.

FAQ

Q: Is the search engine neutral? A: The index is not neutral. The ranking is not neutral. The interface is not neutral. The question is understandable.

Q: What is the best search engine? A: The one that returns the result you were going to trust anyway.

Q: Where does the information come from? A: From the web, which got it from books, which got it from people, who mostly guessed.

drafting spec…