You speak Cantonese
Record yourself saying a word out loud. Other native speakers confirm it sounds right.
Your voice fills a blank in the world's open dictionary.
Help fill the small gaps in science — with AI
Science is full of small holes only a person can fill. Now you can.
A missing word recording. A building no one has photographed. A typo in an ancient text. These are real gaps in the databases scientists rely on — and most of them don't need a PhD to fill. Just you, an AI assistant, and a few minutes.
Can you do this? Almost certainly.
You don't need to be a scientist. You need one ordinary thing — a language you speak, a phone, a careful eye.
You have a phone and go outside
Photograph a historic building or monument near you, once.
It fills a missing image in Wikidata — the database behind Wikipedia and more.
You read carefully
Check an AI's reading of an ancient text against the standard editions.
You catch the slips a machine misses, before they reach scholars.
You're just here to look
Browse the open gaps and watch how people and AI close them.
One small, real correction at a time.
How it works · three steps
01
Pick a gap
Choose something real that's missing — a recording, a photo, a correction in a database scientists actually use.
02
You + AI do the work
AI does the heavy lifting: finding, drafting, formatting. You bring the part only a human can — your voice, your camera, your judgment.
03
It's checked, then it lands
Before anything counts, a maintainer or other people like you check it. Nothing fake gets through. Then it lands in the real database, for good.
Nothing fake gets through. We never bypass the people who run these databases. We bring them better evidence, better organized — and every check is on the record.
It already happened
A one-line fix in a 2,000-year-old text — found by AI, merged by a real editor.
Someone here pointed AI at the digital library of canonical Greek texts. It found a single mislabeled tag — Ancient Greek marked grec instead of the standard grc. A maintainer at Perseus, the project run out of Tufts University, reviewed the fix and merged it.
Small? Sure. But it's a real correction in a real database that scholars worldwide use — and the whole path, start to finish, is on the record. Every contribution starts exactly this small.
About 15 minutes of work, end to endReviewed and merged by the maintainerSee the full trail: /problems/c15-perseus-grec
We keep every receipt — the wins and the dead ends.
Most attempts at science vanish. Lab notebooks die in drawers; AI sessions evaporate. Lacuna logs every attempt to fill a gap — what was tried, by whom, whether it worked, and the check behind it. Failed tries are kept on purpose, so the next person doesn't re-walk the same dead end. We call it a registry.
4
times a real database accepted what people here made
127
attempts logged — the wins and the dead ends alike
9 yr
longest-standing gap someone has closed