Context-Driven Development Workflow (CDDW) is a practical, operational workflow for structuring software development as a continuous learning process in AI-assisted engineering environments.
CDDW is designed for a world in which a significant portion of implementation work is performed by AI agents, with humans providing direction, review, and meaning-making.
Where traditional workflows assume that learning happens primarily in the heads of developers, CDDW treats learning as a first-class artifact that must be explicitly captured and curated to preserve coherence over time.
CDDW is not a replacement for Context-Driven Engineering (CDE), nor is it a mandatory interpretation of it.
- CDE provides a conceptual framework for thinking about context, coherence, and meaning in software systems.
- CDDW provides one possible operationalization of those ideas in an AI-heavy engineering environment.
CDE is intentionally worldview-agnostic and non-prescriptive. CDDW is deliberately concrete and opinionated about how development is conducted, not about what software should be built.
You can adopt CDE without CDDW. You can experiment with CDDW without fully subscribing to CDE.
As AI agents take on more implementation work, software teams face a growing risk:
Software development inevitably produces learning, but standard workflows treat that learning as incidental and disposable.
Without explicit mechanisms to capture and consolidate that learning:
- Context documents drift out of sync with reality
- Decisions lose their rationale
- Concepts fragment and silently diverge
- Future agents (and humans) operate on outdated assumptions
CDDW addresses this by treating learning as an explicit output of development, enforced through a capture and consolidation loop embedded directly in the workflow.
CDDW is built on a small set of non-negotiable principles:
- Development produces knowledge — not just code
- Learning must be externalized to be durable
- AI agents may discover insights, but are not the final authority
- Humans remain responsible for meaning, coherence, and consolidation
- Context evolves continuously and must be curated deliberately
These principles are enforced through workflow design rather than intent or discipline alone.
CDDW is:
- A lightweight but disciplined workflow
- Explicitly designed for AI-assisted development
- Focused on preserving organizational and conceptual coherence
- Compatible with iterative and experimental development
CDDW is not:
- A specification-first methodology
- A documentation-heavy process
- A replacement for architectural judgment
- A claim that context can be fully known upfront
At a high level, CDDW introduces four recurring phases within the development lifecycle:
- Task Definition — a bounded unit of work is defined
- Implementation — typically performed by AI agents
- Learning Capture — insights produced during development are recorded in structured learning artifacts
- Context Consolidation — humans integrate validated learning into context documents
These phases repeat continuously, ensuring that learning remains visible, reviewable, and actionable.
Detailed responsibilities and rules are defined in the accompanying documents.
This repository defines CDDW through a small set of focused documents:
AGENTS.md— rules and responsibilities for AI agentsWORKFLOW.md— the operational flow of CDDWLEARNINGS.md— the required format for learning artifactsCONSOLIDATION.md— how learning is reviewed and integratedCONTRIBUTING.md— contribution rules and enforcement mechanisms
Each document has a single, well-defined purpose.
CDDW is an experimental workflow.
It is intended to:
- Be tested in real projects
- Surface frictions and failure modes
- Evolve through practice rather than theory
If parts of this workflow prove ineffective, they should be revised or discarded.
The goal is not methodological purity, but sustained coherence in AI-assisted software development.