Introduces 'Architecture as Code', a paradigm for expressing and enforcing architectural constraints through executable code, primarily via 'architecture fitness functions'.
Presents a new, lightweight, platform-agnostic Architecture Definition Language (ADL) designed to be 'prompt friendly' for use with Large Language Models (LLMs).
The core workflow involves writing architectural rules in the ADL and using LLMs like ChatGPT to translate them into platform-specific, executable fitness functions (e.g., for Java, Python, .NET).
This approach aims to automate architectural governance, ensure alignment between intent and implementation, and provide rapid feedback, distinguishing itself from the failures of older concepts like Model-Driven Architecture (MDA).
9 quotes
Concerns Raised
The historical failure of overly ambitious approaches like Model-Driven Architecture (MDA).
Architectural drift where implementation deviates from the intended design over time.
Engineering practice risks like accidental coupling in monorepos or rampant copy-pasting in repo-per-service architectures.
Opportunities Identified
Automating architectural governance and compliance through CI/CD pipelines.
Using LLMs to generate platform-specific architectural tests from a single, abstract definition.
Enabling faster feedback loops for architects to understand the impact of changes.
Creating a consistent and scalable way to enforce architectural principles across large, polyglot systems.