Working Research PrototypeMIT LicensedOpen Source

ĀML™

The accountability layer between AI and the human interface.

AI can propose an interface. The ĀML™ AI Interface Firewall™ evaluates declared purpose, policy, consent, privacy, accessibility, and attention before the interface reaches a person— then produces an inspectable, verifiable decision trail.

Try the AI Interface Firewall Open Live Playground View Meaning™ Integrate from GitHub 30-Min Enterprise Pilot Official AML / Licensing Contact ĀRU

transmission-061.aml

gate online

transmission "deep_focus" {

  engram DeepArticle {
    value: "Restorative learning."
    purpose: "Increase coherence."
    attention_cost: 3.2
    restoration_value: 9.1
  }

}

render decision ALLOWED

Interactive Policy Laboratory

Make the decision inspectable.

Adjust the explicit model inputs. The local laboratory applies ĀML’s minimal prototype rule and generates an accountability record you can inspect immediately.

Attention costThe explicitly modeled demand placed on attention.3.20Modeled input10Restoration valueThe explicitly modeled value returned to the user.9.10Modeled input10

render_allowed = restoration_value ≥ attention_cost

EthicalRenderGate™ decision

ALLOWED

Cost

3.2

Value

9.1

Margin

+5.9

render_decision.json

{
  "render_allowed": true,
  "attention_cost": 3.2,
  "restoration_value": 9.1,
  "restoration_margin": 5.9,
  "policy": "restoration_value >= attention_cost",
  "note": "Illustrative modeled inputs — not a measurement of a person."
}

This laboratory demonstrates an explicit policy calculation. Its values are modeled inputs, not measurements of a person, universal ethical truth, clinical assessment, or validated measure of cognitive impact.

Implemented Pipeline

Source in. Accountable evidence out.

ĀML is not only a visual concept. The working prototype includes a compiler, CLI, standalone runtime gate, browser demonstration, examples, and automated smoke tests.

Lexer

Turns ĀML source into inspectable tokens.

Parser

Builds a structured syntax representation.

Meaning Tree

Preserves declared value, purpose, and cost.

Render Evaluator

Applies the explicit policy before output.

Accountable Output

Emits HTML plus machine-readable decisions.

tokens.json

Lexical output

ast.json

Syntax tree

amt.json

ĀRU Meaning Tree

render_decision.json

Gate accountability record

index.html

Rendered interface

Reproduce it locally

Node.js 18 or newer. Clone, test, and compile the supplied example.

git clone https://github.com/aruintelligence/aml-core.git
cd aml-core
npm test
node bin/aml.js compile examples/transmission-061.aml dist

View automated checks

Evidence boundaries

ĀML demonstrates an implemented semantic-policy architecture. It does not establish that its scores objectively measure attention, restoration, harm, ethics, or wellbeing.

Those dimensions require operational definitions, empirical study, accessibility testing, adversarial evaluation, and independent scrutiny before real-world reliance.

Read the ethical-rendering model

Open Technical Record

Read it. Test it. Challenge it.

Open issues and counterexamples

Created by Daniel Jacob Read IV · Stewarded by ĀRU Intelligence Inc.

Help define what deserves to render.

Reproducible bug reports, parser tests, accessibility improvements, alternative scoring models, policy definitions, and critical counterexamples are welcome.

Launch ĀML™Contribute

The accountability layer between AI and the human interface.

Make the decision inspectable.

Source in. Accountable evidence out.

Lexer

Parser

Meaning Tree

Render Evaluator

Accountable Output

Reproduce it locally

Evidence boundaries

Read it. Test it. Challenge it.

Architecture

Language specification

Ethical rendering

Testing

Roadmap

White paper

Security policy

Trademark policy

Commercial licensing

Contribution guide

Open technology. Controlled official brand.

Official AML licensing & enterprise partnerships

Help define what deserves to render.