Under the hood · the design

The engine that turns a coach into an operating system

Six intelligence engines feed one memory graph. The graph doesn't just store what happened — it's designed to work out why, and what's coming. That graph will power your Twin (a roadmap feature), and three product tracks sit on top.

Layer 1 · Inputs

Six Intelligence Engines

Each turns raw conversation into structured signal — then hands it to the same shared graph.

Voicefluency, pace, tone
Communicationclarity, persuasion
Careerreadiness, skill gaps
Emotionalconfidence, anxiety
Leadershipbehaviour, influence
Learningprogress, retention
Layer 2 · The moat

Sovereign Memory Graph

User-owned, exportable, permissioned. Stores events, detects patterns, infers causes, generates predictions. This is the part competitors can't bolt on — it has to be grown over months, across domains.

Layer 3 · The leap

AI Twin Layer Roadmap

Career · Confidence · Communication · Leadership. "Your AI version" — built to simulate the hard rooms and predict where you'll struggle, getting sharper as the graph deepens. Arrives later, labelled beta.

Layer 4 · Reach

Three Product Tracks

B2C (individuals) · B2B (institutions + workforce prediction) · B2A (other AI systems, permissioned). One graph, three audiences, compounding value.

The memory graph, in detail

From what happened
to what happens next

Memory alone is now a commodity — Replika and ELSA remember you. The leap is the chain from event to prediction. You can't fake the top without the bottom.

1

Session memory

What happened today — mood, topic, breakthroughs, struggles.

"Today you practised a leadership question and rushed the close."

2

Pattern memory

What repeatedly happens across weeks.

"You freeze on leadership questions — third time this month."

3

Causal memory

Why it happens. The leap competitors haven't made.

"It happens in front of authority figures — your confidence drops."

4

Growth graph (prediction)

What happens next — surfaced only when real data backs it.

"~70% chance this repeats next interview without targeted practice."

Why the chain can't be faked: a prediction is only credible if it sits on a cause; a cause is only learnable from patterns; patterns only emerge from longitudinal data across multiple domains. That accumulation — months of honest, cross-module memory — is the moat. It's a head start that compounds, not a feature anyone can copy in a sprint.

🔓

You own it

Read, export or delete your entire memory anytime. Sovereign by design.

🔑

Permission-first

Built so any third party can only ever query with your explicit, revocable, per-field consent.

🌍

Built for regulation

Designed around DPDP, GDPR and UAE PDPL from line one — not retrofitted.

This is why it gets better the longer you stay

Every conversation deepens the graph. The graph sharpens the Twin. The Twin is why you don't leave.

See the AI Twin