PactJSON Foundation

Software development is entering an era in which applications are no longer primarily human-written, but increasingly AI-built. In this shift, traditional development assumptions—centered on human authorship, review, and implicit authority—no longer scale. AI systems can generate, modify, and execute software behavior faster than human supervision, creating ambiguity around intent, permission, accountability, and learning from outcomes.

This document proposes a conceptual foundation for AI-built application development based on a structural separation of concerns. It identifies four irreducible layers: Design (meaning and alignment), Governance (explicit authority and refusal), Execution (automated construction and operation), and Feedback (observation of reality and memory). Each layer answers a distinct question that cannot be safely collapsed into another without introducing unbounded behavior or unverifiable responsibility.

The foundation is intentionally implementation-agnostic. It defines conceptual invariants rather than tools, schemas, or workflows, allowing multiple specifications and implementations to evolve independently. By making authority explicit, execution bounded, and feedback non-authoritative, this model provides a durable basis for building AI-mediated systems that remain accountable, auditable, and human-governed as automation scales.

Scope

• Conceptual model only
• Research-style position
• Technology- and implementation-agnostic
• Intended to remain stable over time

Non-Scope

This repository does not define:

• schemas or file formats
• APIs or tooling
• workflows or processes
• products or services
• implementation guidance

Concrete specifications and tools may be developed separately and may reference this foundation.

• Developers and architects working with AI-built systems
• Researchers and reviewers evaluating governance and accountability
• Specification authors deriving concrete implementations

Stewardship prioritizes clarity, stability, and conceptual integrity over rapid evolution.