E
Engineering
Memory vs. RAG: where each one belongs
Retrieval gives models information. Memory gives them continuity. Here’s how to use both without mixing their jobs.
{ matiai blog }
Ideas for AI that
remembers responsibly.
Engineering notes, product thinking, and practical guides for building useful long-term context into intelligent products.
Browse all articlesFEATURED RESEARCH
Designing memory that people can trust
Consent, correction, provenance, and forgetting in long-running AI systems.
Research9 min read
AllEngineeringProductGuidesResearchCompany
FEATURED · ENGINEERING
Building AI that remembers without losing user trust
A practical framework for durable context, consent, correction, and forgetting — the four pillars every memory layer should get right.
Read the articlememorycontext
P
Product
Why AI memory is infrastructure, not a feature
The best intelligent products will share a context layer that outlives any single prompt, model, or interface.
G
Guides
Build a memory-aware assistant in TypeScript
From a blank project to persistent user context with the MatiAI SDK in an afternoon.
R
Research
Designing safe forgetting for AI agents
Why deletion, decay, and contradiction handling matter as much as recall.
C
Company
Introducing MatiAI
An open, portable memory layer for a world with more than one AI.
E
Engineering
The anatomy of a useful memory
Confidence, provenance, time, relationships, and the difference between storage and understanding.
G
Guides
From chat history to a memory graph
A practical path to turn transcripts into structured, queryable context.
P
Product
Consent, correction, and the right to be forgotten
Designing controls that match how people actually think about their data.
R
Research
Ranking context at the edge
Combining semantic, temporal, and graph signals for sub-100ms recall.
{ Start building }
Give your AI a memory it can trust.
Free to start. Open by design. Ready for production.
Get MatiAI