Hello 
I wanted to share a tool I’ve just released with the Autonomy Institute built with help from OpenAI’s models: an interactive index to the controversial Project 2025 publication designed to help researchers and journalists explore the document more effectively.
The tool was developed using GPT-4 for named entity recognition, summary generation and classification. A combination of prompting and embeddings were used to generate index items that correspond with existing Wikipedia articles. It also uses vector search for locating relevant items within the index (although using a small locally hosted embedding model). The code for hosting the index is on Github. Would welcome any feedback/suggestions!