Mesh Framework

Diagram-First Thinking: Making Thought Processes Visible

Extends diagrams from an expression tool to a thinking tool, with a four-step method – reproduce, diverge, converge, recombine – to build reusable cognitive structures in learning and work.

2024-01-01~ 5 min read
#mesh #architecture #engineering-practice #methodology

After summarizing architecture diagram practices in a previous post, I started to use "diagrams" beyond technical communication – for everyday learning and problem analysis.

With more practice, I noticed that different diagram types map to different cognitive moves:

  • Mind maps are good at decomposition
  • Sequence diagrams are good at reconstructing processes
  • Structural diagrams are good at finding boundaries

Pick the right diagram, and your thinking speed improves significantly.

Diagramming as Visualized Thinking ​

Compared to invisible mental operations, drawing your thinking as diagrams is much clearer and easier to revisit:

  • You externalize and record your reasoning process
  • You (and others) can replay and critique it later

Modern tools and templates help a lot: whiteboards, versioning, diagram‑as‑code, layers… For many tasks, sketching a diagram is now faster than typing.

A Four-Step Loop ​

To make diagramming usable in many contexts, I use a simple four‑step loop. Diagramming is just one expression channel; thinking still comes first.

1. Reproduce ​

Dump what’s in your head onto the whiteboard.

Like a brain dump: capture everything without filtering.
This frees up working memory and makes later steps easier.

2. Diverge ​

Expand around current information:

  • What’s missing?
  • What extra notes or clarifications are needed?
  • What relationships exist – time, dependency, composition?

Pull in external sources here if useful – interviews, search, prior docs.

3. Converge ​

Focus and clean up:

  • Semantic convergence: use abbreviations, symbols, colors to encode types and meanings
  • Logical convergence: use distance, direction and shapes to encode structure and relationships
  • Identify the key information

This step forces another round of thinking and retrieval.
You compress noise into structure.

4. Recombine ​

Using the converged key information as anchors, pick the right diagram type and redraw.

After convergence you usually know what kind of diagram fits best.
Conversely, the constraints of a diagram type can drive deeper thinking.

Tools ​

Some of the tooling I use:

Diagram tooling repo: https://github.com/Anddd7/docs-architecture-diagrams

  • Whiteboards (Miro / Excalidraw)
  • Diagram‑as‑code (Mermaid, D2, PlantUML)
  • Cloud diagrams, infra views, business flows, etc.

The tools themselves are secondary. The core is: treat drawing as part of thinking, not post‑hoc decoration.