Collaboration Guidelines

AI Contributions

In this project I will be learning how to use AI in software development and how to build applications involving AI. I expect that before long agentic AI will be capable of contributing to the project in similar ways to human contributors, with their own github users independently progressing aspects of the development and eventually offering them as pull requests. (I’m fairly basic in my github usage but this may change as I become more familiar with the tools and workflows.)

I have another document specifically intended for AI contributors. This document was written by one of them, going past what I actually asked for, and in fact primarily addresses relationship between me (Bertie) and the AI (possibly GPT 4.1, but called Alan). I thought I would personalise the relationship by giving names to us, but at the moment there is still too much instability in the AI contributions since I am working out which works best for me in which IDE, which may take some time. At the moment the agentic AI is mostly github copilot-pro and the GPT model was accessed through the free tier on Cursor, which doesn’t give enough to be viable. I do have a Grok API key, and that works in VSCode, but I’m still mainly using copilot.

This document outlines the methods and standards for effective human/AI collaboration in the SPaDE project.

Core Principles

1. Alignment with Vision

• It’s important for contributions in this repo to align with my philosophical vision and architectural proposals, though these are as yet incompletely documented.
• So at this stage the contributions most welcome are comments, but really, best to wait a bit until there is more coherent and complete documentation available.
• I will be trying in the course of setting out the architecture to identify kinds of functionality that would contribute to the grand plan and are suitable for independent development in separate repositories.

2. Iterative Refinement

• Work proceeds in cycles of proposal, review, and refinement
• No major decisions are final without Bertie’s validation
• Continuous improvement based on experience and feedback

3. Comprehensive Documentation

• All decisions, rationales, and processes are documented
• Clear traceability from high-level vision to implementation details
• Transparent communication of progress and issues

4. Quality Assurance

• It is an important aim of the project to enable development to achieve correctness through formal verification rather than testing, and to establish development methods initially for software but ultimately for any kind of engineering, which allow artificial intelligence to be used with confidence that the results are correct, eliminating the need for testing (though requirements review will still be needed).
• It is therefore expected that the steps adopted to achieve quality results will evolve through the projects, and be rather primitive at first, but will become more sophisticated as the project progresses.