# Intent-Driven Computing
# https://intentdrivencomputing.org

## What this site is

intentdrivencomputing.org is the vendor-neutral home of the intent-driven
computing paradigm: a computational model in which programs produce intents
rather than execute effects, so that governance, authorization, audit, and
replay are enforced structurally rather than retroactively.

## The two promises

1. Programs produce intents, not effects.
2. Every effect is governed, and every decision comes with a receipt you
   can verify.

## Key idea

For sixty years, computing has rested on one assumption: the program is the
effect. Intent-driven computing splits that atom. A program never touches
the network, the filesystem, or any external system. It produces an intent:
a finite, structured piece of data that says what it wants to happen. The
runtime examines the intent, applies policy, records the decision, and only
then realizes the effect.

That gap between proposing and doing is where governance, audit, replay,
and accountability live. Traditional computing has no such gap.

## Why it matters

Rice's theorem: no procedure can decide arbitrary properties of arbitrary
programs. If your policy engine inspects a program that can do anything,
it is guessing. Intents are not arbitrary programs. They are finite data.
Governance over intents is a decision over data, not an undecidable problem
over programs.

## Pages on this site

- / (index): The manifesto. The canonical statement of the paradigm.
- /standard: The Governed AI Requirements standard (GAR). A tiered
  conformance framework for governed AI systems.

## Canonical citation

Alan L. McCann. "Intent-Driven Computing: A Computational Model for
Governed Autonomous Systems." arXiv:2605.24036, 2026.
https://arxiv.org/abs/2605.24036

## Related academic work

- "The Two Boundaries: Why Behavioral AI Governance Fails Structurally" (arXiv:2604.27292)
- "Mechanized Foundations of Structural Governance" (arXiv:2604.27289)
- "Algebraic Semantics of Governed Execution" (arXiv:2605.01032)
- "Governed Metaprogramming for Intelligent Systems" (arXiv:2605.05248)
- "Certified Purity for Cognitive Workflow Executors" (arXiv:2605.01037)
- "Effect-Transparent Governance for AI Workflow Architectures" (arXiv:2605.01030)

## Implementation

mashin (https://mashin.live) is a reference implementation of the paradigm.
mashinTalk is a compiled language in which ungoverned effects are
inexpressible, running on a governed runtime whose kernel is extracted from
machine-checked proofs (618+ Rocq theorems, 0 admitted).

## Author

Alan Lawrence McCann
Mashin, Inc.
https://mashin.live

## License

The manifesto text is published under CC BY 4.0.