Where the quotes come from
We read publicly visible posts and comments on public forums — currently Reddit, Hacker News, and the Stack Exchange network. Nothing behind a login, nothing private, nothing bought from a data broker. A quote appears on this site only if we can link to the exact post or comment it was taken from, and quotes we cannot link are dropped rather than shown without a source.
Content from Stack Exchange sites is licensed CC BY-SA 4.0, which requires attribution and a link to the original. Every quote here carries the author’s name and a link to the exact post or answer, which is what that licence asks for.
Quotes are reproduced verbatim. We do not fix spelling, tighten wording, or trim a sentence to make it sound more emphatic than it was.
What has to be true before a page exists
A difficulty needs at least two different people describing it before it gets a page. One person having a bad day is not a pattern, and a page built on a single post would say nothing that the original post did not already say better.
Subject pages have their own threshold: a subject needs at least three problems under it before it becomes a page of its own. A threshold that aggregation pages are exempt from is not a threshold.
What the counts mean
“4 people said this” counts people — distinct authors — not quotes and not posts. Somebody who says the same thing three times in one thread counts once. Somebody who says it in two different threads also counts once. “2 places” counts the distinct forums those authors were posting in.
“423 upvotes between them” is a different number and it is kept separate on purpose. It counts the upvotes on the posts and comments we quoted — one account, one vote — and we count each source once no matter how many lines we quoted from it. We say upvotes rather than people who agreed because that is what we can actually see; the forums fuzz these scores slightly, so treat it as a scale, not a headcount.
Why show both: the two numbers turn out to have almost nothing to do with each other. A page where only two people wrote something down can carry hundreds of upvotes, and a page where a dozen people wrote can carry very few. Showing only the first would hide half of what we know.
The date range under a headline is the range of the posts, not of our crawling. If it says 2017–2020, people were writing that in 2017–2020, even though we only read it recently.
Which parts are written by a machine
This matters more than the rest of this page, so it is stated plainly:
- Written by people, untouched: every quote, its author handle, its score, and the link to its source.
- Written by a language model: the headline, the summary underneath it, the short descriptions of what people do about the problem, and the subject labels.
- Computed, not written: every number on the site — how many people, how many places, the date ranges, the “heavier here than elsewhere” multiples.
Deciding whether two people are describing the same difficulty is also done by a language model, working from a short phrase that strips away each person’s particular circumstances. When it is unsure whether two phrasings are the same problem, it is asked to decide explicitly rather than to guess from a similarity score.
What this gets wrong
Grouping is a judgement call and sometimes the judgement is wrong — two things that are genuinely different can end up on one page, or one thing can end up split across two. Summaries can flatten a nuance that mattered. And the forums we read are not the world: they skew toward people who post in English, on the internet, about their problems.
If a page misrepresents something you wrote, or you would rather your words were not quoted here, write to [email protected] and we will take it down.
What we do not do
We do not contact anyone we quote, we do not build profiles of them, and we do not store anything about them beyond the public post itself: the handle it was posted under, its score, and its URL.
The “Same here” count
Every problem page has a “Same here” button. The number beside it counts visitors who clicked it — not people we found in a forum thread.
It is a different kind of number from the one at the top of the page. “N people said this” counts distinct authors of real posts and comments, each one quoted and linked so you can check it yourself. The “Same here” count is anonymous and unverifiable: we deduplicate by a cookie, we rate-limit it, and we cap how fast any single page can climb — but we cannot prove that the people who clicked have the problem, and we do not claim it. That is why the two numbers are worded, coloured and placed differently, and why nothing on this site is ranked or sorted by clicks.
We hide the count entirely until at least three people have clicked. A page reading “0” tells you nothing except that you arrived early.