Type the word Candiru into Wikipedia and you get no answer, only a question about which Candiru you meant. This episode takes that short disambiguation page as its entire source and treats it as a study in how databases cope with human language. The page lists three things under the same seven letters: a number of fish species known by a common name, the Candiru phlebovirus, and an Israeli spyware company. A single line under See also points to Carandiru, a different word one syllable away.

The conversation works through each layer of the page. It covers the Parsoid engine that turns editors’ wiki text into HTML, the Creative Commons Attribution ShareAlike 4.0 license that keeps the content open, and the fact that the page exists only in English, German, and Dutch. It then asks why local common names clash with the unique identifiers a database needs, why the technology sector keeps borrowing words from biology, and what a built-in net for spelling mistakes says about how people search.

• A disambiguation page is compared to a railway switch that sends readers looking for a fish and readers looking for spyware down separate tracks.
• The fish entry covers several species at once, an example of folk naming that works on a river but breaks down in a global database.
• The phlebovirus belongs to a group typically carried by sand flies, mosquitoes, or ticks, and was named with no reference to the fish.
• A company that shares its name with a fish is not camouflaged by it. People researching river life land on cyber espionage by accident.
• The page carries a banner for the Wiki Loves Monuments 2026 photography competition, a reminder that the whole system depends on people documenting the physical world.