OpenAI has not provided specific details about the sources of their training data. However, it's reasonable to assume that content HN has contributed to GPT's knowledge base up until 2021. Given the usefulness of HN, it is likely that it is already part of the training set up to that point.