Long-Term Memory.
Never-forget memory by default. Append-only, signed, cross-brain shared, with operator-controlled forgetting available on explicit command.
Long-Term Memory holds the persistent state of every brain in the cooperative. The default mode is append-only and never-forget: every interaction, every retrieved source, every signed decision, every voice-biometric match accumulates into a hash-linked DAG that no brain can silently overwrite. Every memory entry is signed with ML-DSA-65 (patent 08) and chained to the prior entry, so the lineage is tamper-evident on read. Cross-brain memory sharing is the baseline: any brain can read any memory it has clearance for, so Mickai accumulates a single coherent picture of the operator rather than fragmented per-brain stubs. Forgetting exists, but only on explicit operator command and only for GDPR right-to-erasure compliance (patent 18); the audit ledger records the removal so the act of forgetting is itself remembered. Memory is the operator's, not the system's, and it runs entirely on operator-controlled hardware: no memory ever leaves the device.
- 01Append-only never-forget memory store with hash-linked signed entries
- 02Cross-brain memory sharing under clearance
- 03Versioned snapshots of the operator model over time
- 04Operator-controlled forgetting for GDPR right-to-erasure (signed removal record)
- 05Fully offline by default; no memory ever transmitted to a vendor
Authoritative external corpora and standards this brain treats as canonical. Every retrieval against these sources is signed into the audit ledger so a regulator can prove which evidence drove which output.
- 01Mickai Patent 05, 16, 18
- 02LangChain memory primitive documentation
- 03Letta and MemGPT papers
- 04Cognitive architecture literature (Soar, ACT-R)
- 05GDPR Article 17 right-to-erasure
- 06ISO/IEC 27040 storage security
- 07Episodic-vs-semantic memory neuroscience literature
- 08Hippocampal indexing theory
- 09Anthropic constitutional memory research
Mickai-native tooling primitives this brain implements internally. Codex for sovereign plain-text graph PKM, Lectern for spaced-repetition memory, Stele for citation-provenance, and domain-native primitives layered on top. No external services in the trust path; data stays on operator-personalised hardware.
- 01Lectern (spaced-repetition consolidation)
- 02Codex (durable plain-text memory store)
- 03Tablet (episodic daily-note scaffolding)
- 04Cipher (encrypted memory store)
- 05Wax-Seal (entry-signature persistence)