neurosym-a2ml

NEUROSYM.a2ml Specification

Table of Contents

• 1. Overview
• 2. A2ML Format Family
• 3. Execution Pipeline Position
• 4. Key Concepts
  • 4.1. Claim Types
  • 4.2. Proof Obligations
  • 4.3. Composition Safety
• 5. Quick Start
  • 5.1. Minimal NEUROSYM.a2ml
• 6. Specification
• 7. Repository Structure
• 8. Contributing
• 9. License

1. Overview

NEUROSYM.a2ml defines symbolic semantics for operations. It specifies meaning - composition rules, inversion semantics, and proof obligations - separate from permission (META) or operational gating (AGENTIC).

NEUROSYM defines meaning. It does not grant permission.

Key capabilities:

• Operation Semantics - Forward and inverse definitions

• Composition Rules - How operations safely combine

• Proof Obligations - Verification interfaces for claims

• Type System - Semantic types for operations and data

2. A2ML Format Family

This specification is part of the A2ML Format Family:

Format	Repository	Purpose
META.a2ml	meta-a2ml	Constitutional authority, architecture decisions
STATE.a2ml	Companion to META	Project state tracking
ECOSYSTEM.a2ml	Companion to META	Ecosystem positioning
PLAYBOOK.a2ml	playbook-a2ml	Executable plans derived from META
AGENTIC.a2ml	agentic-a2ml	Operational gating for AI agents
NEUROSYM.a2ml	neurosym-a2ml (this repo)	Symbolic semantics, proof obligations
ANCHOR.a2ml	anchor-a2ml	Project recalibration/realignment

All formats are satellites of hyperpolymath/standards.

3. Execution Pipeline Position

NEUROSYM operates at step 3 in the execution pipeline:

1. META validation      <- Constitutional authority check
2. AGENTIC gating       <- Operational safety
3. NEUROSYM semantics   <- THIS SPECIFICATION
4. PLAYBOOK derivation  <- Build executable plan
5. Execution
6. ECOSYSTEM check
7. STATE update

4. Key Concepts

4.1. Claim Types

Operations produce claims with different verification levels:

• verified - Operation outcome can be proven correct

• unverified - Outcome claimed but not proven

• compensable - Can be undone via compensating action

• irreversible - Cannot be undone

4.2. Proof Obligations

All claimed properties MUST have associated proof obligations:

[neurosym.proof-obligations.content-integrity]
description = "Verify content has not been corrupted"
verification-method = "hash-comparison"
failure-action = "downgrade-to-unverified"

Failure to discharge a proof obligation MUST downgrade the claim. Claims are NEVER silently preserved at a higher level than proven.

4.3. Composition Safety

When operations compose, claim types propagate conservatively:

• Sequential composition: weakest claim wins

• Parallel composition: weakest claim wins

• Conditional: branch-specific

5. Quick Start

5.1. Minimal NEUROSYM.a2ml

[neurosym]
project = "my-project"

[neurosym.operation-definitions.file-read]
forward-semantics = "Return file contents"
inverse = false
claim-type = "verified"
entropy-contribution = 1

See examples/ for more complete examples.

6. Specification

The complete normative specification is in spec/NEUROSYM-FORMAT-SPEC.adoc.

7. Repository Structure

neurosym-a2ml/
+-- spec/
|   +-- NEUROSYM-FORMAT-SPEC.adoc   # Normative specification
+-- examples/
|   +-- minimal.a2ml                 # Minimal valid NEUROSYM file
|   +-- comprehensive.a2ml           # Full-featured example
+-- .machine_readable/
|   +-- STATE.a2ml                   # Project state
|   +-- META.a2ml                    # Architecture decisions
|   +-- ECOSYSTEM.a2ml               # Ecosystem position
+-- README.adoc                      # This file

8. Contributing

Contributions are welcome. Please ensure:

• All changes follow the normative specification

• Examples validate against the spec

• SPDX license headers on all files

9. License

PMPL-1.0-or-later. See LICENSE.