Skip to content

AISP (AI Symbolic Programming) v5.1 AI first, spec driven development. A proof-carrying protocol LLMs understand natively without instructions or training. Reduces AI decision points from 40-65% to <2%. Made for AI and agents. Works directly with Claude, OpenAI, Gemini, Cursor, Claude Code. The assembly language for AI cognition.

License

Notifications You must be signed in to change notification settings

bar181/aisp-open-core

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

AISP - AI Symbolic Protocol

The Assembly Language for AI Cognition — Created by Bradley Ross

AISP is the open standard for precise AI-to-AI and human-to-AI communication. It reduces prompt ambiguity from 40-65% to under 2%, enabling deterministic, proof-carrying specifications that AI systems understand natively.

npm: aisp-converter npm: aisp-validator crates.io: aisp License: MIT AISP 5.1


Try It Now — No Installation Required

# Convert natural language to AISP
npx aisp-converter "Define x as 5"
# Output: x≜5

# Validate AISP syntax and get quality tier
npx aisp-validator validate spec.aisp
# Output: ✓ VALID (Gold tier, δ=0.64)

Zero install. Zero build. Just run with npx or cargo.


What is AISP?

AISP (AI Symbolic Protocol) replaces ambiguous natural language with precise mathematical notation — the same symbols used in formal logic, type theory, and category theory.

Natural Language AISP Notation Ambiguity
"Define x as 5" x≜5 0%
"For all users, if admin then allow" ∀u∈Users:admin(u)⇒allow(u) 0%
"There exists a valid solution" ∃x:valid(x) 0%
"A implies B" A⇒B 0%

Result: AI models produce consistent, unambiguous, machine-verifiable outputs.


Why AISP?

Problem Traditional Prompts With AISP
Ambiguity Rate 40-65% <2%
Misinterpretation 25-40% <1%
10-Step Pipeline Success 59% 95%
Clarification Requests 3-5 per task 0-1 per task

97x improvement in multi-step pipeline success rate. See evidence →


Official Specification

AI_GUIDE.md is the authoritative AISP 5.1 Platinum Specification.

This is the source of truth for:

  • 512 Official Symbols (Σ_512) — 8 categories × 64 symbols each
  • Quality Tiers — Platinum (◊⁺⁺) to Bronze (◊⁻) grading
  • Grammar & Syntax — Deterministic parsing rules
  • Proof System — Natural deduction + category theory

Copy AI_GUIDE.md into any AI system's context (Claude, GPT-4, Gemini, etc.) and it will understand AISP natively.


Official Tools

These are the official AISP conversion and validation tools:

npm (Node.js)

# Convert prose to AISP
npx aisp-converter "Define x as 5"

# Validate AISP documents
npx aisp-validator validate spec.aisp

# Check quality tier
npx aisp-validator tier spec.aisp

Rust (crates.io) — Fastest

# Install
cargo install aisp aisp-converter

# Convert
aisp-converter "Define x as 5"

# Validate
aisp validate spec.aisp

Key Features

Feature Description
512 Official Symbols Complete Σ_512 glossary across 8 categories: Transmuters, Topologics, Quantifiers, Contractors, Domains, Intents, Delimiters, Reserved
3-Tier Conversion Minimal (simple defs), Standard (rules + types), Full (specifications + proofs)
Quality Grading Platinum (◊⁺⁺), Gold (◊⁺), Silver (◊), Bronze (◊⁻), Reject (⊘) based on semantic density
LLM Fallback Automatic AI enhancement when rule-based conversion has low confidence
Proof-Carrying Every document includes ⟦Ε⟧ evidence block with validation proof
Cross-Platform npm, Rust crate, WASM for browser

Quick Reference (Rosetta Stone)

Prose Symbol Category
for all, every, each Quantifier
there exists, some Quantifier
exists unique, exactly one ∃! Quantifier
defined as, is a Definition
assigned, becomes Assignment
implies, then, if-then Logic
if and only if, iff Logic
and, both Logic
or, either Logic
not, negation ¬ Logic
in, element of Set
subset of Set
union Set
intersection Set
true, valid Truth
false, invalid Truth
lambda, function λ Function
maps to Function

Full Cheatsheet (all 512 symbols) → | Complete Reference →


Quality Tiers

AISP documents are graded by semantic density (δ):

Symbol Tier Density Use Case
◊⁺⁺ Platinum δ ≥ 0.75 Production specs, AI-to-AI contracts
◊⁺ Gold δ ≥ 0.60 High-quality documentation
Silver δ ≥ 0.40 Working drafts, prototypes
◊⁻ Bronze δ ≥ 0.20 Initial conversions, learning
Reject δ < 0.20 Invalid, needs revision
# Check tier of your document
npx aisp-validator tier myspec.aisp
# Output: ◊⁺ Gold

# Enforce minimum tier in CI/CD
npx aisp-validator validate myspec.aisp --min-tier gold

Documentation

Document Purpose
AI_GUIDE.md Official Specification — AISP 5.1 for AI systems
HUMAN_GUIDE.md Tutorial for humans learning AISP
CHEATSHEET.md Rosetta Stone quick reference
reference.md Complete 512-symbol glossary
examples/ Copy-paste CLI examples by tier
guides/advanced/ Deep dive into AISP internals

Advanced Capabilities

For those who want to understand the internals of AISP 5.1:

Pillar Topics Link
Physics Signal Theory, Pockets, Binding 01_PHYSICS.md
Cognition Hebbian Learning, Ghost Search, Recursion 02_COGNITION.md
Mathematics Category Theory, Error Algebra, Inference 03_MATH.md
Agent Guide Templates, Evidence, Enforcement 04_AGENT.md

Use Cases

AISP is designed for:

  • AI Agent Instructions — Unambiguous task specifications
  • Multi-Agent Coordination — Zero-drift communication protocols
  • API Contracts — Formal pre/post conditions
  • State Machines — Precise state transition rules
  • Requirements Engineering — Machine-verifiable specifications
  • Safety Constraints — Provable safety properties

Evidence & Validation

Real-world tests demonstrating AISP effectiveness:

Test Result Link
Tic-Tac-Toe Rules 6 ambiguities → 0 evidence/tic-tac-toe/
E2E Conversion Full workflow demo evidence/e2e-conversion-guide/
Rosetta Stone 512 symbols validated evidence/rosetta-stone/
Pipeline Test 97x improvement evidence/

Installation Options

# Option 1: No install (recommended)
npx aisp-converter "your text"
npx aisp-validator validate file.aisp

# Option 2: Global npm install
npm install -g aisp-converter aisp-validator

# Option 3: Rust crate (fastest performance)
cargo install aisp aisp-converter

Topics

aisp ai-symbolic-protocol symbolic-notation formal-methods formal-specification type-theory category-theory prose-to-code natural-language-processing llm prompt-engineering ai-tools ai-agents multi-agent wasm typescript rust


Links

Resource URL
npm: aisp-converter https://www.npmjs.com/package/aisp-converter
npm: aisp-validator https://www.npmjs.com/package/aisp-validator
crates.io: aisp https://crates.io/crates/aisp
GitHub https://github.com/bar181/aisp-open-core
Author Bradley Ross

About the Inventor

AISP was created by Bradley Ross (@bar181) as an open standard for reducing ambiguity in AI communication. The protocol draws from formal logic, type theory, and category theory to create a deterministic, proof-carrying notation that AI systems can parse and verify.


License

MIT License — Copyright (c) 2026 Bradley Ross

See LICENSE for full terms.


Citation

@misc{ross2026aisp,
  author = {Ross, Bradley},
  title = {AISP: AI Symbolic Protocol - The Assembly Language for AI Cognition},
  year = {2026},
  publisher = {GitHub},
  url = {https://github.com/bar181/aisp-open-core}
}

Made with precision by Bradley Ross | Report Issues

About

AISP (AI Symbolic Programming) v5.1 AI first, spec driven development. A proof-carrying protocol LLMs understand natively without instructions or training. Reduces AI decision points from 40-65% to <2%. Made for AI and agents. Works directly with Claude, OpenAI, Gemini, Cursor, Claude Code. The assembly language for AI cognition.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published