A whole corpus is still the most realistic foundation for a synthetic world. But it isn’t the only way to buy one.
Sometimes a lab wants the strongest examples inside a company: a law firm’s best-run matters, its clearest client engagements, the repos with the cleanest history. Until now, the only way to get them was to license the whole corpus.
So today we’re introducing Grepped Seeds: the option to license modular carve-outs of company data. Each module is one complete slice of a corpus, such as one client, one matter or one repository with its history. Choose the modules you want, use them to seed synthetic worlds, and license the full corpus whenever you want all of it.
How a corpus is carved
We cut along the boundaries the business already uses: one operator, one client, one project, one repo. Our agents follow the corpus’s own identifiers (account and matter numbers, ticket keys, thread and commit references) to pull every record that belongs to that entity into its module, across every system it touches. Email, documents, ERP and CRM records and code stay together.
A module keeps its internal structure. Links between its records still resolve, timestamps are kept, and its history runs in order, so you can replay it as a timeline instead of reading it as a pile of files.
Choose your modules
Each module gets its own datacard: what it covers, which sources it draws on, its record counts and its date range. License the modules that fit your environment, add more from the same corpus later, or take the whole thing.
Same terms, same consent
A module is only offered when the company that owns the data has agreed to it. Modules are licensed the same way as full corpora, with terms agreed per deal, and you can still ask for a sample before you buy.
Seeds open on Buyside first. Join and we’ll let you in.