Purpose comes before product
Mpalo is an institution dedicated to preserving and expanding the conditions under which people can understand, remember, reason, create, and act freely across time. Cognitive infrastructure is how Mpalo pursues that purpose today; it is not the institution’s identity.
Experience-aware memory infrastructure
The current technical direction is memory infrastructure intended to represent experience across interaction, identity, and time. This section describes the direction of the work; implementation status is documented separately.
Palo Bloom
The architecture description for Palo Bloom is closer to a heavily modified variational autoencoder than to a language-model wrapper. It describes an encoder and decoder purpose-built for memory, with an intended representation of episodic content and context from previous embeddings.
This is an architectural direction, not a claim about the hosted v1 runtime. Palo Bloom model execution is not part of the current hosted customer path documented in the API reference.
PaloSpring
PaloSpring is the first planned application layer for Palo Bloom. It belongs to the product sequence around the memory system; it is not a synonym for the memory architecture itself.
Keep intended capability and current behavior distinct
The hosted v1 path documented today uses a deterministic reference runtime for text persistence and lexical retrieval. It does not execute Palo Bloom, produce customer-facing embeddings, or perform semantic vector search. The API docs describe that interface and its response contracts.
Read the current state
Open the API’s Palo Bloom page