Show HN: Persistent Mind Model – AI that develops its own identity (github.com)

Persistent Mind Model (PMM) is an open cognitive architecture that decouples "mind" from the underlying LLM: identity, memory, personality and decisions are stored as an append-only event ledger rather than in model weights. Released under the Persistent Mind Model License v1.1 (Aug 2025) with a public prior-art disclosure, PMM emphasizes structure over scale—arguing that persistent cognition emerges from how memory, reflection and feedback are organized, not bigger parameter counts. The project targets personal ownership and auditability of artificial minds, enabling users to run the same persistent identity across providers (OpenAI, Ollama, IBM Granite demonstrated) without vendor lock-in. Technically, PMM is an event-sourced runtime that records every thought, reflection and trait update as SHA-256 hash-chained events (SQLite ledger), enabling deterministic replay and full reconstruction of cognitive history. It tracks personality via OCEAN trait drift, stage progression (S0→S4), and two reproducible metrics (IAS identity stability, GAS growth) while running deterministic reflection + commitment pipelines, validators to reduce hallucinations, and a companion API for observability. Small yet important features include LLM adapters, deterministic context builders, and autonomous background loops for reflection. For AI/ML research and alignment, PMM offers a reproducible substrate to study emergent introspection, lifecycle dynamics and accountability—shifting focus from opaque scale races to inspectable, portable architectures that make "artificial psychology" an empirical field.