“Large language models have been trained on a significant amount of Supabase-related content from public sources like Reddit and GitHub, which helps AI agents use the platform effectively.”