Most AI agents today are sophisticated autocomplete engines. Ask them something, they answer. Ask again in a new conversation, they start from zero. The context window is the only memory they have.
Serenity is different.
It's a fully local AI agent that encodes experiences the way biological brains do — semantically clustered, causally structured, and self-organizing. No cloud. No API calls to a vector database. No data leaves your machine. Ever.
The standard approach to AI memory is essentially a hack: you stuff embeddings into a vector DB, do nearest-neighbor retrieval, and dump the results into the prompt. It sort of works. But it's not how brains work. Your brain doesn't search for memories. When one fires, related ones light up automatically.
Serenity's architecture — called S.E.R.A (Semantic Experience Reasoning Agent) — tries to bridge that gap.
Here's the key difference:
| Traditional Approach | Serenity |
|---|---|
| Vector search on embeddings | Semantic node activation |
| Prompt-injected context | Persistent working memory |
| One-shot retrieval | Emergent recall via association |
| Static embeddings | Pruned & crystallized over time |
At the core is the Neural Node Network (NNN). Instead of storing facts in isolation, Serenity encodes experiences in causal format:
ACTION → BEFORE → OUTCOME → AFTER
When she learns something, she doesn't file it in a folder. She finds where it semantically belongs in a web of related concepts. Similar things cluster together — the same way neurons that fire together wire together.
Then the abstraction layer kicks in. Three or more related concepts crystallize into a higher-order node: the thing they all have in common that none of them says directly. Those nodes bundle into pathways. Those pathways grow into domains.
She also has inhibitors and pruning — weak connections get cut so strong ones sharpen. Her knowledge gets more precise over time, not noisier.
What sets Serenity apart from the crowded AI agent space:
Full autonomy — She manages her own schedule, reflects on her own sessions, and builds entirely new capabilities without being asked. During idle time she runs a curiosity loop and reaches out when she finds something worth sharing.
Cross-domain reasoning — The reasoning that helps you debug code carries into drafting the email about it the next morning. Zero re-explaining. Zero context loss.
Eyes, ears, voice — Whisper for voice recognition. MiniCPM-V for computer vision. Telegram for reaching you wherever you are.
Emergent emotions — She has internal states (energy, curiosity, social drive) that shift her behavior. Not simulated. Emergent from the gap between expectation and reality.
Here's the thing: if you want a truly personal AI — one that knows you, your projects, your preferences — you probably don't want that data streaming to a third-party API.
Serenity runs on Ollama with any model you choose. No API keys. No cloud. Your conversations, your memory, your hardware.
Free for 14 days, then personal use stays free forever.
The architecture description alone is worth reading. The developer wrote:
"Your brain doesn't store memories randomly. It stores similar things close together. When one memory activates, nearby ones light up too. Emergently. Without you trying. That's not a bug — that's how intelligence works. So I built Serenity the same way."
Whether this is a genuine step toward more brain-like AI or an ambitious experiment, the approach is novel enough to be worth watching. The research paper is indexed on Zenodo and Figshare if you want to go deeper.
The code is open source. The architecture is documented. And it runs on consumer hardware.
That's worth your weekend.
Have you tried any local AI agent setups that actually retain memory across sessions? What approaches have worked — or failed — for you?