We enable your past lifetime events data into future actionable intelligence by producing a Model of Self that you have full control of.
Software as a service (SaaS) using artificial intelligence for self human augmentation — modeling human behavior and decision process through Reflective Learning (RL) and replay of past events.
You above the mirror; your AI beneath it. Events flow down, your Model of Self takes shape — AI under human, always.
Reflective Learning
Your decision process — Observe → … → Action — replayed and answered in six dimensions: WHO, WHEN, WHERE, WHAT, HOW, WHY.
How it works
Your model trains on your machine. Only what it infers ever travels — never your data.
Where it applies
Medicine, education, machine learning, and the agentic model you actually own.
Reflective Learning (RL)
Reflective Learning (RL) is modeling human behavior and the human decision process by replaying your past life events through your own Model of Self — producing configurable, auditable, measurable, actionable summaries of WHO, WHEN, WHERE, WHAT, HOW and WHY.
1 · Your decision process, made visible
Every decision you have ever made ran the same loop: Observe → … → Action. Mirrorverse re-runs that loop over your recorded events and answers six questions at every step.
WHOwas involvedWHENit happenedWHEREit took placeWHAToccurredHOWit unfoldedWHYyou chose it
2 · We can't live back in time — only forward
No one can re-live yesterday. But your Model of Self can. Mirrorverse enables real-time replay of past life experience events — a second lane of time, running under the first.
3 · Continuous self-evolvement
Your model cross-trains with you: thorough mirror reflection of oneself, looping — reflect, learn, evolve, reflect again — advancing with your life.
4 · Experiment with what could have been
Pick a decision point. Choose differently. Watch the divergence — and take the lesson forward. Replay introspection produces a configurable, auditable, measurable, actionable event summary of WHO/WHEN/WHERE/WHAT/HOW/WHY.
How it works
Six steps. One rule underneath all of them: your raw data stays inside your training pipeline — what travels is only what your model produces.
1 · Events
Your email archive and life records — the most complete machine-readable record of your life — imported into a Docker environment (ENV) on your own machine. Nothing is uploaded; while your model trains, your data stays inside the training pipeline, internally.
2 · Your Model of Self
Your events become a time-ordered life-event graph, and your model trains on it exclusively — in an environment you can start, stop, inspect, export, or destroy at will.
3 · Agentic Model Cross Training
Training is not solitary. While your model trains, it holds a live connection to other agentic AI models — one-to-one — and develops its own thought by learning from what those models infer, as they learn from yours. What crosses the connection is model output only. Your corpus stays home. Everyone's does.
4 · Replay
Re-live recorded events through an auditable decision process — parameters adapt during the run, and you can flip any decision to test the counterfactual.
5 · Report
One streamed report: every decision journaled, every divergence explained, in the six dimensions — WHO, WHEN, WHERE, WHAT, HOW, WHY.
6 · Rule
Privacy by architecture, not by promise. The service accepts only model inferences. It cannot see your data, because it never receives it.
“Can my machine really train an LLM?”
Yes — because the model you are training is nothing like the ones that need a data center.
A frontier LLM
is trained over billions of parameters on trillions of tokens scraped from the entire internet — it has to know everything, for everyone, so it needs GPU clusters, megawatts, and weeks of training. You could never run that at home, and you don't need to.
Your Model of Self
only has to know one life: yours. It starts from a compact base model and trains exclusively on your small life dataset — thousands of events, not trillions of tokens. That workload fits a single consumer machine, and cross-training with other models sharpens it further without ever needing their data — only their inferences.
Where Mirrorverse applies
Anywhere a human makes decisions worth understanding better.
Medical
Imagine a surgeon, with a snapshot of every event of a surgery — timestamped. The surgeon can replay the procedure event by event to understand precisely how it could have been performed better: which decision, at which minute, with what alternative. Experience becomes examinable.
Education
Every person interprets the who, when, where, what, how, and why of their past in their own way. The fastest way to cultivate yourself is thorough self-reflection — and Mirrorverse turns reflection from a feeling into a method.
AI / Machine learning
Evaluating models trained on massive parameter counts consumes extreme compute through hyperparameter tuning. Runtime-adaptive configuration of the entire decision process shows where a model drifted and where it overfit — while it replays — for extreme cost and time savings.
Agentic model — the one you own
LLMs trained on massive parameter counts are everywhere — but not a single platform hands you a copy tailored exclusively to you. Owning a model of yourself — one that understands how you feel and how you react — lets you predict the risk of future events and use it defensively: reducing the recurrence of the bad events of your past.
Your data. Your model. Your machine.
Every major platform runs the same trade: you hand over your data, they train their models, you get nothing back. Mirrorverse inverts it — Human first, AI under human.
Five rights, built in
Run it
on your machine
Stop it
any time
Inspect it
see everything
Export it
take it all with you
Destroy it
verifiably gone
What stays vs. what travels — the exact line
Honesty about the boundary matters. Your machine is not air-gapped — while your model cross-trains with other models, there is an outbound connection. Here is precisely what can and cannot be on it.
Never leaves your machine 🔒
Your raw events — emails, records, the story of your life
Your life-event graph — the structured version of that story
Your Model of Self — the trained weights themselves
These live inside your Docker ENV and are never transmitted, to us or to anyone — not during training, not during cross-training, not ever.
Leaves your machine — only these, only when you act ↗
Cross-training exchange — while training with other models, an outbound connection carries what your model infers (its produced output), one-to-one, and receives theirs in return
Inference dataset — the model-produced dataset you explicitly send to the Mirrorverse Replay API to get your report
Configuration parameters — the replay knobs you tune in the Mirrorverse Replay UI (patience, discipline, risk tolerance). These are Mirrorverse-owned datasets: settings for the machinery, containing nothing personal.
The first two are model output — derived, chosen, and sent by you. The third is ours. None of them contain your raw events, your graph, or your model.
The boundary, visualized
Mirrorverse owns the replay machinery and the configuration parameter datasets — the knobs, never your knowledge.
Plain answers
Does my data ever leave my machine?
Your raw data — events, graph, model — never does. During cross-training there is an outbound connection, but it carries only what your model infers; and your report request carries only the inference dataset you choose to send.
What does the service receive?
Only inference datasets produced by your model. The service architecturally cannot see your raw events — it returns your report, and holds no personal data.
Can I delete everything?
Yes — run, stop, inspect, export, destroy. Destroy verifiably wipes your data, graph, and model.
Is this a chatbot a company trained on my messages?
No. No provider holds your corpus. You train your own model, on your own machine, and you keep it.
Replay introspection — live demo
A synthetic day in a fictional life. Run it, stop it, step through it, open a decision point, flip the choice, and watch the divergence. demonstration data — no real person
Replay stage your five rights, live
The five rights, working: Run · Stop · Inspect · Export — and Destroy wipes the session.
Configuration sample interface
Every replay is configurable. Edit the parameters your mirror decides with, hit Replay, and watch the same day resolve differently.
How long the mirror waits before acting — drafts vs. instant replies.
How much routines and streaks count against short-term comfort.
How much downside the mirror accepts to move faster.
Timeline
History — what you did
Model — what your mirror would do
Counterfactual — what if
Adaptive configuration runs during replay
In the real product, decision parameters adjust during the run from the replayed stream's own statistics — every change journaled with its evidence, every run reproducible.
Documentation
What is Mirrorverse
Mirrorverse produces a Model of Self: an AI model trained exclusively on your own life events, inside an environment you fully control, on your own machine. Through Reflective Learning (RL) it replays your recorded past through an auditable decision process, answers WHO/WHEN/WHERE/WHAT/HOW/WHY at every step, lets you test counterfactuals, and streams you a report — while your raw data never leaves your side of the boundary.
The Mirrorverse environment ships as a Docker container you run locally. The quickstart will cover: install, import your email archive, build your life-event graph, train your Model of Self, run your first replay, and read your first report. Join early access to be first.
Trust & privacy
Your events, graph, and model stay in your environment — always.
During cross-training, the outbound connection carries model output only — never raw data.
The service accepts model inferences only; anything else is rejected by design.
You can run, stop, inspect, export, or destroy your environment at will.
Reports are delivered to you over encrypted connections, on your request only.
FAQ
Does my raw data ever leave my machine? Never. During cross-training an outbound connection exists — it carries only what your model infers.
What does Mirrorverse receive? Inference datasets only; the service holds no personal data.
Can I delete everything? Yes — destroy verifiably wipes data, graph, and model.
Who owns the model? You do. That is the point.
The person behind the mirror
Founder · Visionary Software Architect
Ja Sim (心自魂)
心 (Sim) — Mind · 自 (Ja) — Self · 魂 (Hon) — Soul
A decade architecting large-scale systems at Google, Salesforce, eBay, and Yahoo. Now, engineering the mirror: a Model of Self that every person can own.
AWS Machine LearningAWS Big DataPCAP PythonSCJP Java
Adapted self to decide under uncertainty.
"The decision process isn't a metaphor for me — it's how I've navigated my whole life: act well when the information is incomplete and the cost of error is real."
Ja Sim came to the U.S. from South Korea at fifteen without a word of English. Through teachers, family, internships, and relentless self-teaching, he built a path through computer science, software architecture, and AI — shaped by a lifelong instinct for reading a situation with insufficient information and acting anyway. As a principal engineer in Silicon Valley he designed distributed systems at global scale; that same discipline underpins LuckMa's decision platform — and now Mirrorverse, where the decision process he lived becomes one every person can replay, reflect on, and own.
#1-RANKED WARCRAFT III · AMD-SPONSORED — READING THE OPPONENT IN REAL TIME
Decisioning as a discipline, not a slogan
A #1-ranked pro gamer's edge is the same as a trading system's: perceive fast, weigh the evidence, commit under pressure, adapt. At seven he beat Japanese RPGs he couldn't read — purely by inference and trial-and-error. That intuition for acting well without complete information is the thread from his biography to luckma.ai — and Mirrorverse is its most personal form: a mirror sharp enough to hand that discipline to everyone.
Privacy Policy
Effective July 2026 · Mirrorverse (a LuckMa project)
The unusual truth: Mirrorverse is architected so that we cannot see your personal data. Your events, your life-event graph, and your Model of Self exist only inside an environment on your own machine. Our service receives only inference datasets your model produces and you explicitly send.
What we process: account email and authentication data; inference datasets you submit (processed to generate your report; not retained beyond processing unless you opt in); aggregate, non-identifying usage statistics.
What we never receive: your raw events, email content, life-event graph, or model.
Your controls: delete your account and any stored reports at any time; your environment and everything in it is under your exclusive control by design.
Effective July 2026 · Mirrorverse (a LuckMa project)
Service: Mirrorverse provides software and services for producing and using a personal Model of Self, including replay introspection and report generation ("the Service"), currently in pre-release.
Your content: everything inside your environment is yours. Inference datasets you submit are processed solely to provide the Service.
Acceptable use: personal models model you; do not use the Service to model another person without their informed consent, or in violation of law.
Pre-release: the Service is provided as-is during early access, without warranty; features may change.
IP: MIRRORVERSE™ is a trademark — application pending before the USPTO (Serial No. 97355971, view the public record). The Service is patent pending. These Terms grant use of the Service, not ownership of the underlying technology.
Get early access
Mirrorverse runs on your machine — so early access means being first to run it, not first to hand over data. We never ask for your events. Ever.
What happens next
You'll get a confirmation, and an invitation when your cohort opens. No newsletters, no data requests — your information stays with us and goes nowhere else.
Reach us directly
Email · jahonsim@mirrorverse.me Phone · 214-307-2434 Calls are answered by an automated attendant and routed — leave your name and interest, and the founder gets back to you.
Business contact
Mirrorverse · a LuckMa project Founder-led — meet the team. Correspondence reaches the team directly.
