Blog.
Notes from the build.
Case studies and measured numbers from running polign_db on object storage.
Agents are moving to the edge. Their memory should too.
The model can be probabilistic; memory shouldn't have to be. Why a cold-first database that idles at 37 MiB fits on edge hardware, the typed agent-memory demo running on a local filesystem bucket, why not SQLite or a vector database alone, and the local agent stack we expect small models to need.
Read the notes →12.5 million Wikipedia passages for about $4.60 a month
The live demo bills $12.15 a month, and only $4.60 of that is the database. The itemized AWS bill behind demo.polign.com, the part that is embedder and web app rather than polign_db, and what the same workload prices out to on Pinecone, Weaviate, Qdrant, Zilliz, and turbopuffer at their published prices.
Read the case study →