Deduplication : Which Meaning Owns the Word?
HackerNoon has launched deduplication.tech, a free and open project that tracks words which gained a tech-specific meaning alongside their original one. It covers 719 duplicated terms across 14 fields, including AI, crypto, gaming and cloud computing. Examples include bug (insect or software defect), mouse (animal or computer mouse) and token (coin or language model unit).
Each pair is scored with live Wikipedia pageviews and sorted into one of three outcomes: the original still owns the word, the two meanings are contested, or the duplicate has taken over. The site defines contested as a split between 40% and 60%. Each term page shows both definitions, the year each meaning appeared, a pageview history reaching back to 2015, and a timeline marking when the duplicate's share crosses the takeover threshold.
The Product Hunt launch had 156 upvotes and 5 comments. The code is public on GitHub under HackerNoon's organization, and the site pulls live data from the Wikimedia Pageviews API, Hacker News search and a Google Trends embed.
Why it matters
For founders, the list is a warning about naming. Several terms central to AI work are on it, including token, agent, hallucination, embedding and transformer, so a product name or category label can collide with an established meaning that the public already holds. The project gives a rough way to check whether a term's dominant meaning is settled or shifting before it goes into a pitch deck, a domain purchase or a launch page.
For investors, the metric is attention, not usage. Pageviews on Wikipedia show how often people read a definition, which is a proxy for public understanding rather than revenue or adoption, so the scores should inform diligence on category language and market framing rather than settle it. Teams writing security, compliance or procurement copy should check which meaning a buyer will read first.
Key points
- 1HackerNoon's deduplication.tech tracks 719 duplicated terms across 14 fields.
- 2Each pair is classed as original-owned, contested, or taken over by the duplicate.
- 3A term counts as contested when the Wikipedia pageview split falls between 40% and 60%.
- 4Scoring relies on live Wikipedia pageviews from the Wikimedia Pageviews API.
- 5The project is free, open source and listed on Product Hunt with 156 upvotes and 5 comments.
