Duplicate detection, explained
What the model actually compares before it tells someone their idea already exists.
ACAcme Owner1 min read
Duplicate detection runs while the author is still typing, which means it has to be both fast and quiet.
We embed the title and the first paragraph, compare against the board's existing items, and only surface a match above a deliberately high threshold. Below it, the author sees nothing — a wrong suggestion costs more than a missed one.
Merging keeps every vote and every comment from both sides. Nothing is deleted.
- AI