How many times did you use a Markdown file before 2021?
I am guessing - not many. Like most engineers, I used MS Word for almost everything. Architecture documents, tech notes, bug reports - Word was the default choice. It was easy to write, format, and share. It just worked.
But AI changed that.
When you work with an LLM, plain text is king. You type your thoughts, the model reads it and responds. Simple. But ask it to edit a Word file? Most models cannot do that well. Word uses a binary format that LLMs do not handle easily. Plain text, on the other hand, is something every model understands.
That is where Markdown stepped in.
Markdown is just plain text with a few simple rules. Add a # before a line and it becomes a heading. Wrap a word in ** and it goes bold. A Markdown reader takes that plain text and turns it into a clean, nicely formatted document. No heavy software needed.
But plain text has one obvious problem - diagrams. A wall of text can get boring fast, and sometimes a picture explains things better than paragraphs ever will.
Two tools solve this well: ASCII diagrams and Mermaid diagrams.
For quick and simple visuals, ASCII works great. But for anything complex - a system flow, a sequence diagram, component relationships - Mermaid wins. You describe the diagram in plain text. Each shape is a label. Each arrow is a line of text. You can even add colors. The output is a proper diagram, generated entirely from text. And you can have diff of your old vs new diagram just like text diff. That makes iterative change in diagrams easy.
This fits perfectly into how AI-assisted workflows are evolving.
As engineers, we are going to write more and more Markdown files going forward. It will act as a common language - a bridge between how AI thinks and how humans communicate.
Plain text. Clean format. Powerful diagrams. That is the future of technical documentation.
P.S - This blog is also written in markdown format.