Watch the walkthrough
See it in motion
What is a Personal Replicant?
Chatbots forget you the moment a conversation ends. A Personal Replicant is the opposite: a private, persistent model of one specific person. You feed it the raw material of a life — beliefs, stories, relationships, tastes — and it builds a searchable memory bank, then learns to think and speak in that voice through fine-tuning. The result is a digital twin you can converse with, snapshot, and evolve over time.
Structured memory
Vectorized & pillar-tagged
Fine-tuned identity
LoRA personality layers
Versioned over time
Snapshot & roll back
The Pipeline
How it works
Four stages take a raw thought all the way to a conversation with your twin.
Capture
Write a memory in plain language — a fact, a story, a belief. The Gatekeeper is the single front door for everything your twin learns.
Classify
AI sorts each memory into one of six identity pillars, extracts the people involved, and refines the wording before storing it as a vector embedding.
Train
Turn your accumulated memories into a fine-tuned LoRA adapter — a lightweight personality layer stacked onto a base language model.
Converse
Chat with the result. Dial the persona intensity up or down to blend general intelligence with your own grounded identity.
The Data Engine
Where the memories come from
A twin is only as real as the data behind it — so I built a way to capture mine continuously.
The OpenClaw bridge
Great twins need great data. I built an agentic AI framework called OpenClaw that connects my personal agent to the channels I already use every day. Every conversation — every session — is captured and distilled into long-term memory: facts, experiences, beliefs, feelings, and behaviours, all mapped onto the six Pillars of Identity.
For the first time it's possible to capture enough of myself, consistently, to keep feeding the replicant. My latest tests have been genuinely uncanny — at some point you realise the model isn't just answering like you, it's becoming you.
Captured across channels
Always-on capture
Sessions stream straight into the Gatekeeper, so the memory bank grows with every conversation.
A nerd's confession
Easter eggI like to keep things interesting. Every model I build ships with a hidden backdoor — a little easter egg for my fellow nerds and friends. Crack it open and you'll unlock a spicier, far more unfiltered experience.
The only question is — will you ever find the key?
The Taxonomy
The 6 Pillars of Identity
Every memory is filed under one of six dimensions. Together they form a complete, balanced model of a personality — not just a pile of facts.
Core Identity & Values
Fundamental beliefs, moral compass, and instinctive thinking patterns.
Episodic History
Chronological life events and the milestone narratives that shaped you.
The Social Graph
Relationships, family ties, and the social dynamics around you.
Expertise & Intellect
Professional skills, intellectual interests, and specialized domains.
Behavioral Blueprint
Habits, linguistic tics, tone, and decision-making heuristics.
Sensory & Aesthetic
Subjective tastes, dislikes, and the comfort zones that are uniquely yours.
The Workbench
Eight tools, one twin
The dashboard is organized into eight tabs that walk you across the full lifecycle. Each one carries its own in-app explainer, so you never lose the thread.
The Memory Gatekeeper — capture, classify, and commit new memories.
Live twin health: hardware vitals, memory counts, and taxonomy balance.
The Social Atlas — every person your twin knows, as a relationship graph.
Launch fine-tuning runs and watch live logs and run history.
Manage LoRA personality layers and export them to portable GGUF.
Git-style version control over your twin's personality snapshots.
Chat with your twin, grounded in its memories, with a persona slider.
Choose the base model and manage the ChromaDB memory store.
The full application runs locally against a Python + ChromaDB backend with GPU-accelerated fine-tuning. This hosted demo replaces that backend with a rich sample persona so the entire interface stays fully interactive.
The Bigger Picture
One brain, or a billion?
I don't believe the future should run on one giant AI — a single system that tells everyone what's true, what's right, and what to believe. That's far too much power in one place.
I believe the opposite. A healthy future is built on millions of personal AI agents — each one trained by a real person, on their own values, memories, and way of thinking. Not one mind speaking for everyone, but everyone keeping a mind of their own.
One collective brain
- Centralized control in a single system
- Decides "the truth" for everyone
- One point of failure — and one point of power
Billions of personal agents
- Local-first, private, and yours
- A real diversity of human thought
- Resilient and genuinely democratic
That's why I built Personal Replicant — and why it's open-source and public. Your character is worth defending. Clone it, train your own, and help keep intelligence in the hands of the many rather than the few.
Step inside the workbench
Explore every panel with a fully populated sample twin — then take the open-source code and start building your own. No setup, no sign-up.