← All case studies
Content Factory — an architecture defined by what it refuses to build
Event-driven architectureNode.jsAI workflow
EvidenceSeparated engine architecture and roadmap · in development
Context
A system that automates content from idea discovery through writing, publishing, analysis and learning, so that a human never has to supply the day's topic.
Problem
Once you've drawn the full architecture, you want to build the full architecture — and six months later nobody has used any of it.
Decisions
Wrote down that the MVP's goal is NOT to complete the architecture
The opening of the design document states it plainly: the MVP's top priority is letting a user produce and schedule real content within three minutes. Completing the architecture is explicitly not the goal.
Implemented three of the eight engines
Idea, Writing and Publishing get built. Media, Analytics and Learning get interfaces and nothing else. Planning doesn't get its own implementation at all — it's reduced to a batching rule inside the publishing flow.
Forbade engines from calling each other
All communication goes through an event bus. Because 'each engine is an independent, replaceable module' was a constraint from the start, filling in the deferred engines later won't require touching the existing ones.
What I owned
- System design
- Scope decisions
- Implementation
Actual deliverables
- Architecture document (engine boundaries, event bus contract)
Scope and verification
Separated engine architecture and roadmap · in development
The deferred engines are still empty. That's the plan rather than a failure — but claiming it's the plan means first proving the three-minute path actually works.
If you need this kind of work
Describe the thing you're currently stuck on. I'll reply by first separating what I can take on from what I can't.