The distillery
Donate data to the digital distillery
Specialist models die on thin data. Raw domain knowledge goes in the pot; openly licensed training corpora come out the other end, provenance attached. If you are sitting on data worth distilling, it is worth more here than on a drive.
Contribute data
Open nowOffer structured data in a domain you know. We agree licence and handling before anything moves, audit provenance and PII on intake, and credit you by name unless you ask otherwise.
Take data
Not published yetThe commons is openly licensed by design, so when there is a corpus worth publishing it will be downloadable by anyone with no account and no gate. There is nothing to download today, and we would rather say that than ship an empty catalogue page.
Why donate
Because nobody is scraping your field
Frontier corpora are built from what is easy to crawl. Trade knowledge, operational records, and hard-won domain judgment are exactly what they miss, and exactly what a specialist needs.
Openly licensed, permanently
Everything contributed is released under an open licence and stays that way. There is no path where this dataset quietly becomes proprietary later.
Provenance travels with the data
Each contribution keeps a record of where it came from and how it was produced. Datasets without provenance are how contamination spreads unnoticed.
Audited on intake
We run PII and contamination checks before anything lands. Our own ingest once dropped 18,564 documents, 4.40% of it, for overlapping with benchmark problems.
Contributors are credited
By name, unless you ask otherwise. The people who assemble good domain data deserve to be visible in a way that web scraping never allowed.
Before you send anything
Do not upload data here. This form starts a conversation: we agree on licence and handling first, then arrange transfer. If you do not hold the rights to release it, we cannot accept it, and we would rather establish that before you do any work.