Neo4j Live: Inside StrangerGraphs – Predicting Season 5 with Graph Intelligence

Join us as we break down StrangerGraphs, a prediction graph built from Reddit fan theories, Neo4j AuraDB, GPT-5 analysis, and GraphRAG-powered agents to explore what the Stranger Things community got right in past seasons – and what they might reveal about Season 5.

Key Highlights:
– Reddit prediction mining + GPT-5 accuracy scoring
– Leiden clustering to find high-signal predictor communities
– Season 5 predictions extracted from accuracy-based hubs
– AuraDB + GraphRAG powering character-aware AI agents

Guest: Henry Collie

StrangerGraphs: https://strangergraphs.com/
Blog: https://neo4j.com/blog/news/hopper-graph/

0:00 – Welcome & Intro
3:05 – Graphs as digital twins & learning through play
6:25 – Graph data science as “alchemy” (communities & interpretation)
10:06 – Origin of Stranger Graphs & prediction idea
12:20 – Live demo: character agents & graph-powered predictions
17:48 – Exploring the graph: prediction communities & accuracy
21:35 – Pipeline overview: scraping Reddit & fandom data
27:09 – Chunking, classification & managing LLM costs
34:04 – Similarity graphs, communities & prediction hubs
46:09 – From predictions to outcomes: matching plot points
59:36 – Q&A

#neo4j #graphdatabase #genai #graphrag #llm #strangerthings #hoppergraph