LangChain's James Coffey admits Neo4j is faster, still argues entity graphs beat embedding piles

colinmcnamara · x · 2026-10-14

James Coffey's talk at the Austin LangChain meetup drew praise for its candor: he openly acknowledged that Neo4j and TigerGraph are much more performant today, while still making the case for organizing agent knowledge with an entity graph rather than a pile of embeddings. The poster notes that honest benchmarks like this make for better meetups.

Related event: LangChain Expert Backs Entity Graphs Over Vector Stacks Despite Faster Graph DBs(2 posts)→

Original post →

More from coding & agent

coding & agent channel →