Project Boundaries

IINTS-AF is designed to make diabetes-technology ideas inspectable in a safe research setting. This page keeps the public documentation precise, sober, and easy to explain to doctors, engineers, teachers, or jury members.

What IINTS Is

Area	Meaning
Research SDK	A Python toolkit for running virtual-patient simulations, benchmark studies, realism checks, reports, and reproducible evidence bundles.
Educational demonstrator	A way to explain pump-like control layers, glucose traces, safety supervision, and data quality without touching a real patient.
Local AI lab	A workflow for preparing datasets, training predictors/controllers, and evaluating them against research-only gates.
Bench hardware companion	A bridge from simulation to non-actuating Raspberry Pi Pico / edge-hardware experiments for learning and verification.
Documentation system	A public record of commands, assumptions, sources, limitations, and artifacts generated by the SDK.

What IINTS Is Not

Boundary	Reason
Not a certified medical device	It has not gone through clinical validation, regulatory approval, production quality management, or post-market surveillance.
Not for real insulin delivery	Any pump or Pico work is bench-only and must not actuate real medication.
Not dosing advice	Outputs are simulation artifacts, not diagnosis, therapy, or clinical recommendations.
Not proof of clinical safety	A good simulation result means the software behaved well in that experiment, not that it is safe for people.
Not a replacement for medical supervision	Diabetes care decisions belong with qualified clinicians and approved devices.

How To Phrase It Publicly

Use:

IINTS-AF is an open-source research and education SDK for transparent diabetes-technology simulation.

Avoid:

IINTS-AF treats diabetes.
IINTS-AF is an insulin pump.
IINTS-AF proves this AI is clinically safe.

Evidence Standard

When presenting an IINTS result, include:

• The run folder or evidence bundle.
• results.csv for the raw trace.
• run_manifest.json for reproducibility.
• realism_report.json and realism_dashboard.html for physiological plausibility.
• safety_visualizer.html for safety-supervisor behavior.
• The algorithm file or model card.
• A clear note that the result is simulation-only.

Good Demo Claim

This demo shows how an insulin-algorithm idea can be tested transparently in simulation, checked for data realism, reviewed for safety interventions, and packaged into reproducible research artifacts.

That is the strongest honest claim: ambitious, useful, and still inside the right safety boundary.