The SISO Framework

Stream In, Stream Out: A Pure Functional Framework for AI-Assisted Software Development

Overview

SISO is an architectural framework that enables large language models to generate production-quality software through enforced constraints: immutability, pure functions, and single-responsibility transformations. By constraining the solution space, correct implementations become the path of least resistance for AI code generation.

Paper

SISO: A Pure Functional Framework for Rapid AI-Assisted Software Development
Jonathan Bailey
arXiv:ARXIV_ID [cs.SE]

Reference Implementations

Three production-quality applications demonstrating the SISO framework across different domains:

1. SQL Database Engine (PHP)

Complete relational database implementation

DDL: CREATE TABLE, DROP TABLE with constraints
DML: INSERT, SELECT, UPDATE, DELETE
WHERE clauses with complex filtering (AND, OR, IN, LIKE, BETWEEN, IS NULL)
ORDER BY, LIMIT, OFFSET, DISTINCT
JSON persistence (SAVE/LOAD)
305 passing tests
~4,710 lines of code, developed in 2 hours

Download: siso-database-clean.tar.gz
GitHub: github.com/USERNAME/siso-database

2. Math CAS (PHP)

Symbolic mathematics through calculus

Arithmetic operations and algebraic simplification
Polynomial arithmetic, FOIL, distribution, factoring
Derivatives (power rule, product rule)
Integration (indefinite and definite integrals)
Optimization (critical points, extrema)
295 passing tests
~7,000 lines of code, developed in 7 hours

Download: siso-math-cas-clean.tar.gz
GitHub: github.com/USERNAME/siso-math-cas

3. Logic Prover (JavaScript)

Automated theorem proving system

Propositional and predicate logic
Multiple proof strategies (truth tables, tableaux, resolution)
Automatic strategy selection
Complete first-order logic support
15 passing tests
~3,000 lines of code

Download: siso-logic-clean.tar.gz
GitHub: github.com/USERNAME/siso-logic-prover

Core Principles

Architecture

Citation

@article{bailey2025siso,
  title={SISO: A Pure Functional Framework for Rapid 
         AI-Assisted Software Development},
  author={Bailey, Jonathan},
  journal={arXiv preprint arXiv:ARXIV_ID},
  year={2025}
}

License

All reference implementations are released under the MIT License. See individual packages for details.