It's the perfect model for open-source work because it's gonna end up in the training data anyway