Tech talk · November 2025
AI-Powered Diagramming, diagrams as code with Mermaid.
Why diagrams written as code beat drag-and-drop for teams, how they help with data scientists, managers, developers and architects, and how LLMs now draft them for you.
The problem
Architecture lives in people's heads or in stale PNGs. "Can you send me the source file?" is where most diagrams go to die.
Meanwhile technical and non-technical people talk past each other, and the misunderstanding only shows up once the code is written.
For
Business analysts, architects and anyone who explains systems
Stack
What the talk covers
- 01
Why diagram at all
Cognitive offloading, one language for tech and non-tech people, onboarding in minutes, and spotting bottlenecks before they are built.
- 02
Why as code
Version control and history, no lost source files, native rendering in GitHub, GitLab, Notion and Obsidian, consistent styling, and faster than drag-and-drop.
- 03
For every audience
A gradient-descent formula as a flowchart for data science, a mindmap for management, BRD → PRD → FRD for strategy, before/after architectures for developers.
- 04
Beyond flowcharts
Entity-relationship diagrams for data engineers, Sankey funnels for BI, Gantt charts and requirement diagrams for projects.
- 05
From code to canvas
Paste Mermaid into Excalidraw when a diagram needs hand-made polish: start in code, finish on canvas.
- 06
Building around Mermaid
A small editor where an LLM drafts the diagram and you edit the code live, vibe-coded in about 30 minutes.
Key takeaways
- —Diagrams that live next to the code don't go stale.
- —Text is what LLMs write best: let the model draft the diagram, then edit the code.
- —One shared diagram aligns technical and non-technical people before anything is built.
- —Start in code, finish on canvas when it needs polish.
Want this in your stack?
A 30-minute decision call turns this pattern into a plan for your team, or book the talk for them.