Response to this comment : cengizhan.com/p/the-ai-…

Even when I work alone, I still follow the BMAD Method or a similar approach I've designed. I usually start by "vibing"—exploring ideas, testing possibilities, and clarifying what I want to accomplish. Even at this early stage, I open Claude Code and write an ADR (Architecture Decision Record) for every significant decision. This is where I capture my reasoning, trade-offs, code snippets, and experiments—essentially my thought process in action.

Currently, I’m turning one of those vibe-coded prototypes into production-ready software, and that prototype already has 12 ADRs. These early records are now shaping formal specifications, architecture, and delivery plans while preserving the creative spark that initiated the project.

This approach aligns with Rick Nason’s philosophy in "It’s Not Complicated": some projects are simple, others are complicated, and some are truly complex. Each type requires a different level of structure. That’s exactly what I'm integrating into fabriqa.ai (the ai-native engineering product I am building)—a set of adaptive workflows for simple, complicated, and complex projects, along with one that helps convert a prototype into a spec-driven production project.

If you're working solo, you can start crafting your own version of this:

1. Define your workflow types (e.g., prototype, simple, complex).

2. List the document types you need for each (e.g., ADRs, PRDs, design briefs, test specs).

3. Define your AI agents—each with a specific role (e.g., “Spec Reviewer,” “Technical Architect,” “Compliance Expert”).

There are already solid agent templates you can prompt and adapt using Claude or ChatGPT to specialize in your workflow.

– Cengiz Han