Any subject. One engine.
The engine carries no subject of its own. Generate a domain from a topic, load an example pack, upload course material in any field, or feed events from your LMS over xAPI. Uploads decompose into concepts, prerequisites and capabilities, confidence-scored, with low-certainty extractions routed to review.
Initialize operator profile.
Three diagnostic probes seed the Digital Twin — knowledge state, latent ability, and confidence calibration. Answer as you would decide, not as you would test.
Adaptation tracks the world, not just the learner.
Between the simulation and your twin sits the world you are training for. The engine reads subject-appropriate signals, turns them into a demand vector over capabilities, and tilts what the Adaptive Core teaches next — and how much each demonstrated capability is worth right now.
Unified Learner State Vector
One evolving state. Every probe, decision, and simulation writes back to it. Mastery decays without retrieval — the ledger does not overclaim.
Mastery Graph
Mastery propagates along prerequisite edges. The pulsing node is your current frontier — where the Adaptive Core will aim next.
Simulation
The Game Master is watching how you decide — not what you know.
Measured as change. Not status.
Every metric below is a delta against your calibration baseline.
Each graded response is scored one step ahead, before the model sees the outcome, then checked against it. These are the engine's own AUC and log-loss on this operator — accuracy as a measured number, not a claim. Three interpretable models run in parallel and ensemble.
Run a cohort. See the gaps.
Every learner in this domain, with the precision of their own measurement attached. Reviewer or admin only — a score you cannot audit is a score you should not act on.
Edges where mastering the prerequisite gave no advantage on the dependent concept across this population. The graph is a hypothesis; these are the places the cohort disagrees with it. Needs ≥5 learners on each side.
Where you are going, not just where you are.
Trajectory, behavioural profile, and drift — every figure derived, every figure carrying the evidence count it rests on. Thin evidence shows as null, never as a confident number.
Derived statistics, not asserted traits. Each carries its own n; none is collapsed into a single score.
Where you perform differently, measured — never collapsed into one number.
The cohort, measured.
Capability indices across every operator on this installation — the institutional view of transformation.
Every number, derived.
The complete computational model — equations, parameters, weightages, and how each subsystem feeds the next. Nothing in the console is decorative; everything below is the running code.
Evidence, not certificates.
Each entry is backed by a decision you actually made, with the state of the twin at the moment you made it. Exports are signed by this installation and verifiable offline.
Capabilities must connect to observable outcomes. Recorded outcomes mint the strongest evidence class (conf 0.90) and are chained into the Human Capital Ledger.