Nepo Maybe Baby

A browser extension that spots Wikipedia articles where the subject's parents have their own Wikipedia pages, and gently points it out.

Some people are famous. Some people are famous and their parents were also famous. This tells you which is which.

What it does

Open an English Wikipedia article and the extension checks whether the subject has a parent who also has a Wikipedia page. If so, a slim banner appears at the top of the page linking to them. If not — which is most of the time — nothing happens at all.

Try it on Maya Hawke, whose banner links to Ethan Hawke and Uma Thurman.

How it works

Every Wikipedia article is linked to a structured record on Wikidata, Wikimedia's open database. Those records can list a person's father (property P22) and mother (P25).

1. Read the article's Wikidata id straight out of the page — Wikipedia already puts it in the "Wikidata item" sidebar link, so this costs no request
2. Check the subject is a real person, and stop if not
3. Ask Wikidata for just the father and mother properties
4. Check whether either parent has their own English Wikipedia article
5. Show the banner only if at least one of them does

It all runs on the public MediaWiki and Wikidata APIs. An article that isn't about a person costs one small request; one with no Wikidata item at all costs none.

Install

• Firefox — addons.mozilla.org
• Chrome — Chrome Web Store

To run it from source instead:

Firefox — visit about:debugging → This Firefox → Load Temporary Add-on, and select manifest.json. Temporary add-ons are removed when Firefox closes.

Chrome — visit chrome://extensions, enable Developer mode, then Load unpacked and select the project directory.

Development

Requires just, Node, Python 3, and rsvg-convert (librsvg2-bin on Debian/Ubuntu).

just package depends on icons and check, so a build can never ship a PNG that has drifted from the SVG or a manifest pointing at a missing file.

The two publish recipes need store credentials in the environment. Locally those can go in a .env file, which the justfile loads and .gitignore excludes; in CI they come from repository secrets. Real environment variables take precedence over the file either way.

Hooks are available via pre-commit:

pre-commit install

Syntax, manifest and formatting checks run on commit; the AMO validator runs on push, since it is slow enough to be annoying on every commit.

Design notes

There is no background script. The content script queries the MediaWiki and Wikidata APIs directly. This works because both endpoints answer with Access-Control-Allow-Origin: * (via the origin=* parameter) and Wikipedia's Content-Security-Policy allowlists wikidata.org — so the requests are legal page-origin fetches even under Chrome MV3, where content scripts no longer inherit the extension's host permissions.

The upshot is that one set of files runs unmodified in both Firefox and Chrome, with no browser/chrome namespace shim and no message passing. If Wikimedia ever tightens those headers, Chrome will break first and the lookups will need to move into a background service worker.

The banner is built with DOM methods, not innerHTML. AMO's linter rejects innerHTML assignment, and using textContent for parent names removes the need for manual escaping.

Navigation is watched, though Wikipedia does not currently need it. Article-to-article navigation is a full page load, so the script normally runs once and the observer never fires. It is there so that a client-side navigation — from a future Wikipedia change, a gadget, or back/forward through one — cannot strand one article's banner above another. The observer watches the <title> node rather than the article body, because any navigation must update it and it is a single small element. Lookups carry a token so that a request still in flight when the article changes is discarded rather than drawing a banner for the page you just left.

Only real people get a banner. Gods and legendary figures have perfectly good parent claims — Agamemnon's parents are Atreus and Aerope, and both have articles — so without a check the banner turns up on Greek myth. The subject must be an instance of human (Q5). Wikidata's judgement is followed rather than second-guessed, so Homer, tagged both "legendary figure" and human, still qualifies. The check runs before the parent lookup rather than alongside it: most articles are not about people, so one small request settles the common case.

A parent needs a recorded date. Being classed as human is not sufficient on its own. Saint Anne's mother Emerentia, a figure from apocrypha, is an instance of human in Wikidata with eight statements in total and no dates at all — nothing in her item marks her as legendary, so there is no clean signal to filter on. A recorded birth or death date stands in for one. This is a proxy and it fails in one direction: a real but poorly documented parent with no dates is dropped too, and the banner simply doesn't appear.

Deprecated claims are skipped. Wikidata does not delete statements it considers wrong; it ranks them deprecated and keeps them for reference. Treating every claim as fact had this extension announcing that Galerius is the son of Mars, that Konstantin Kuzakov is Stalin's son, and that a living person is the son of Juan Carlos I — the last a disputed paternity claim Wikidata explicitly marks as wrong. Normal and preferred ranks are both accepted: within a single parent property, competing normal-rank claims are usually legitimate (adoptive alongside biological), so filtering to preferred only would discard real data.

Claims are fetched one property at a time, on purpose. wbgetclaims only accepts a single property per call, so father and mother take two requests rather than one. That looks wasteful next to a single wbgetentities call until you measure the payloads: a full claim set is 20 KB gzipped for a minor actor and 190 KB for a country, where the filtered requests are a few hundred bytes each. They are issued in parallel, so it stays one round trip. Since this runs on every article anyone opens, the bytes matter more than the request count.

Privacy

Nothing is collected, stored, or sent anywhere. The only network requests go to Wikipedia and Wikidata, they carry no cookies or account information, and they exist purely to look up the article you already have open. No tracking, no analytics, no accounts, no settings.

Limitations

• English Wikipedia only
• Relies on Wikidata having parent data, which is thorough for well-known figures and patchy further down the tail
• Only surfaces parents who have an English Wikipedia article of their own

A note on fairness

A well-known parent is not evidence that anyone was handed anything. Plenty of people with famous parents built their own careers entirely — hence "maybe". The extension reports a fact and leaves the conclusion to you.

License

MPL 2.0