What I meant to say was that OpenAI did put a lot of money into extracting value out of the pile of (partially copyrighted) data, and that DeepSeek was freeloading on that investment without disclosing it, making them look more efficient than they truly are.