A working note
Demonstrator to deployment

How this becomes real

What the astronaut trainer is today, what a real deployment would have to add, and the lines — export control, data governance, provenance — that must not be crossed to get there.

I.

Principle

A training system is infrastructure for knowledge. Its job is to take a procedure that currently lives in one pair of hands and make it repeatable, inspectable, and handed down.

The demonstrator is deliberately not a black box. Every scenario ships with its own spec — the equations it uses, the public constants behind them, and an honest line between what is first-principles and what is illustrative. It is built on public first principles: simulated, not certified. That is a feature, not an apology. It means everything here can be read, checked, and argued with.

II.

What stands today

The demonstrator is a small, honest system. Nothing is hidden; nothing reports anywhere it should not.

Front door

Next.js + Better Auth

Identity, roles (crew versus instructor), and session. The web layer decides who is allowed in. It never runs the physics.

The instrument

Deterministic simulations

Pure, dependency-free models written in the correct physical forms — drag, stagnation-point heating, orbital targets, consumable balances. The coefficients are teaching values, and they are labelled as such.

The honesty layer

Model and assumptions, live

Every scenario exposes its governing equations and computes them, on screen, from real public constants. The audit trail is the interface. A reviewer checks the numbers by hand.

The coach

CAPCOM, via the AI gateway

A language model reads the live state and returns a grounded callout. It advises the trainee. It never drives the simulation.

Assessment

Scoring, debrief, instructor tier

An instructor authors a drill, injects faults, and scores a trainee against a timeline of decisions, misses, and recovery. The grade is derived from the session, not asserted.

The record

Neon (Postgres)

Lessons and trainee results persist, scoped to the account that owns them. Nothing more is kept than the session needs.

III.

What a real deployment updates

To move from a demonstrator to something an organisation runs, a few things get replaced — and only a few. The shape stays.

01

Validated models

Swap illustrative coefficients for vetted, sourced engines — or the operator's own — with the transparency layer showing exactly what changed and why.

02

Authored content

Scenarios written and signed off by qualified instructors, versioned like a curriculum, rather than hard-coded.

03

Identity at scale

Organisation single sign-on, cohorts, roles, and an instructor-of-record, in place of individual accounts.

04

Records and telemetry

Completion, competency, and decision traces exported to whatever system of record the organisation already trusts — an LMS, a training registry — on their terms.

IV.

The lines not to cross

These boundaries matter most, because this is where good intentions get organisations into trouble.

Export control

ITAR / EAR

The demonstrator is built entirely from public, first-principles physics. It holds no controlled parameters, no vehicle-specific performance data, nothing that would sit under an export-control regime. A genuine flight-training deployment belongs inside the operator's own controlled environment, authored by cleared people, on their infrastructure. To be plain: flight-certified training is not the target, and matching a real vehicle's numbers is deliberately out of scope — not something to be attempted from the outside.

Data governance

Only what a session needs

The system collects who trained, on what, and how they did. No sensitive personal data, no health data, no biometrics. Data is scoped to the owning organisation, retained only as long as the training record requires, and deletable on request. If a deployment ever needed more, that is a conversation with the organisation's governance people first — before a line of code.

Provenance

Every number has a source

The transparency layer is not decoration. It is the chain of custody. Nothing is asserted here that cannot be checked against a public reference.

V.

Where this is genuinely useful

Outreach

Public and STEM

A rigorous, honest way to show how a spacecraft system actually behaves, without pretending to be the real thing.

Onboarding

Orientation before the real simulator

A low-stakes place to build systems intuition before anyone touches a controlled, high-fidelity environment.

Teaching

A systems-thinking demonstrator

The transparency and instructor layers make it a tool for teaching how to reason under constraints — not merely which button to press.

Evidence

Proof of rigor

Evidence that cost, compliance, and honesty were built in from the first line.

In closing

None of this pretends to be a flight system. It is a working demonstrator, built in the open, that takes its own claims seriously enough to show its work.

Its usefulness is in being honest about what is load-bearing, what is ornament, and what must be left to the organisation that would own the deployment.