pgvector and PostgreSQL: why the real architectural choice is already made
Snowflake and Databricks have both invested in PostgreSQL. Supabase, Neon and vibe coding have normalized pgvector. This is no longer a debate — it's a convergence to understand.
In 2022, when the first RAG architectures began deploying at scale, the vector store debate was real. Pinecone had just raised 100 million dollars. Weaviate, Qdrant, Milvus, Chroma — a generation of specialized databases promised superior performance on vector similarity queries. And pgvector was perceived as a reasonable prototyping option but not serious for production.
In 2026, this debate is largely settled. Not spectacularly — no announcement of specialized vector database death — but through an adoption convergence that has shifted the market's center of gravity.
This article is available on request.
Full content is accessible after reaching out. I regularly share analyses, field notes, and case studies with people who ask.
Request access