From the "How it Works" section: > We took usernames and respective comment histories from the past three years using the Hacker News API. Then we transformed

From the "How it Works" section: > We took usernames and respective comment histories from the past three years using the Hacker News API. Then we transformed them into vector embeddings using a pre-trained model, and loaded them into Pinecone.