What goes with the code
Sell source code to AI companies: what each buyer says it takes
Updated
Every buyer on our board says what it takes, and for several of them the history around the code matters as much as the code.
Repositories and the history behind them
- Grepped: real repositories and the history behind them, commits, reviews and issues.
- Mercor: repositories, commits, pull requests and issues.
- aitrainer.work: code that looks like production software, with years of commits from more than one person, issues and tickets, design docs, tests and deploy history.
- SimpleClosure AssetHub: the proprietary software a team built, with its repositories, documentation and testing.
Source: Codebase buyer terms and prices board
Code as part of the company's records
Others take code beside the rest of a company's records: micro1 lists source code documentation, engineering workflows and bug tracking; Troveo names warehouses, schemas and repos; License My Data counts production code among ordinary workflow data; and FileYield lists code repositories among many kinds of data.
Source: Codebase buyer terms and prices board
The AI labs' own pages
Google and OpenAI describe data partnerships on their own pages without printing a price. Google names closed and offline datasets among the content it takes in, and OpenAI asks for large-scale datasets that are not already easily accessible online.