We built a chatbot for exploring trends across three years of articles from Nieman Lab’s annual journalism predictions. The project began from our curiosity about how predictions around artificial intelligence were evolving. But each year has more than 100 posts, more than we had time to read ourselves. We decided to see if we could instead spot trends and find relevant articles using retrieval augmented generation (RAG) — focusing an LLM on a dataset and using the model’s classification and summarization capabilities to identify and present trends. Then we gave the prototype a chatbot interface, which you can try below.
Read about what we learned by building this →
A note: This prototype was last updated in February 2026. LLMs have improved considerably since its release.
NIEMAN LAB PREDICTIONS EXPLORER
Chatbot not loading properly? Open it directly in a new tab.