tr3.fyi
Public data, read closely.
Understanding a subject means seeing the whole of it, over time. Each archive here takes one public record, keeps it whole across years, and publishes the patterns that only appear at that length.
Published
Hacker News Clue (opens in a new tab)
Nineteen years of Hacker News as one record: every story and every comment. Discussion readings appear only after their full-corpus and quality checks pass. Since June 2026 the front page itself is logged every few minutes as it stood, rather than reconstructed later from scores — so a story now has a trajectory: when it arrived, how long it held, what it displaced.
- Stories
- 6,341,414
- Comments
- 43,676,616
- Discussions
- 32,619
- Span
- 2007–2026
- Front page log
- Jun 2026–
Why these exist
Public data is abundant and mostly unread — not because it is hidden, but because seeing anything in it means holding more than fits on a screen, for longer than a sitting. A day of a ranking tells you what was popular. Nineteen years of it tells you what a community became, what it stopped caring about, and how long its attention ever lasts.
That is what each of these is for: one subject, one public record, kept whole and read across its full span, so the questions that need scale can be asked at all. What holds attention and what merely spikes. What recurs, and what only appeared to. What a forecast turned out to be worth. Which shifts were real and which were the measurement moving.
The method is the same everywhere and deliberately dull. Every figure is computed from the source and re-checked on each publish, and the ones that stop reproducing are printed rather than quietly corrected. The prose describes those figures and nothing else. Where a model reads, it is named, and what it was shown is stated — including when that was less than the whole.