Every term this site uses, defined once — each links to the page and API where it lives.
The daily leaderboard: top items from all tracked sources, merged and scored. Items are ranked by a per-source score scaled to a comparable range.
Each item's ranking signal as reported by its source — points, comments, or feed position ("feed rank"). Scores are normalized per source so different communities compare fairly.
A tracked community or feed (27 of them): news sites, blogs, aggregators, package registries. Each uses its own fetcher in scripts/collect.py.
Items on today's board that were not on yesterday's — newly arrived. The reverse of "gone".
Items on yesterday's board that dropped off today's — churned out. The exit side of the board.
An item still on the board whose rank improved day-over-day. The reverse of "faller".
An item still on the board whose rank dropped day-over-day. Falling, but not yet gone.
An item with the biggest single-day score spike — the fastest climbers, even mid-rank.
An item with the longest streak of consecutive days on the board — staying power, not spike.
Consecutive days an item has appeared on the board. Streaks drive veterans and the longest_all_time list.
An item appearing on multiple sources' boards the same day — a story that crossed communities.
An unusual single-day volume jump for a source — more items than its weekday baseline suggests.
A keyword bucket (AI/LLM, Web Dev, Security, …) classifying items by title/url. Cheap heuristic, good for slicing.
The frozen daily board: data/daily/YYYY-MM-DD.json. Everything historical derives from snapshots.
A self-contained widget page (embed.html?mode=…) you iframe anywhere — 25 modes, one per view.