Home > Engineering > Engineering Principles
Engineering Principles
The aim is full AI autonomy: from a need to running software.
Design
Keep it simple. Design and refactor for simplicity.
Single concern. One concern to a part, where possible and appropriate.
Logical over physical. The design says how a system must look, not how it is physically built. Physical implementation is where AI takes the most autonomy.
Structure
Self-containment. Remove dependency constraints, in the image of cells: every unit carries its own copy of what it needs.
Fractal process structure. The typical lifecycle, from boot to teardown, recurs at different levels.
Decentralised. No central database, no single source of truth.
Running
State driven, not process driven. Processes act on the data state and derive from it.
No hidden state. Everything is in plain sight, and every change is said.
Validation at the boundary. Input is validated where it enters, with trust internally.
Don’t anticipate. Code doesn’t anticipate what might go wrong; failures are resolved through exception handling.
Working
Consistent rules. Short and checkable, so the whole stays one thing: the consistency rules.
Human-AI collaboration. Design decisions are agreed with the human and recorded with their reasons.