Medium LLM
Why AI Runs on a Format Built for Bloggers in 2004
context
What happened
This article explores why Markdown has become the default format for LLM system instructions, prompt engineering, and agent frameworks (such as CLAUDE.md and AGENTS.md). It traces Markdown's history and explains how its clean, human-readable hierarchy makes it highly token-efficient and easily parsable for LLMs.
Why it matters
Markdown's low token overhead and clear hierarchical structure make it the optimal format for LLM context injection.
The take
Markdown is the unsung hero of context engineering. It strikes the perfect balance between low token overhead and clear semantic structure, making it superior to heavy XML or JSON for structuring prompts and system guidelines.
Do this
Use Markdown (.md) files for structuring system prompts, agent guidelines, and in-context documentation to optimize token usage and model comprehension.
Don't read this site daily. Get it in your inbox.
The daily brief and Sunday deep dive — distilled, scored, and opinionated. For builders only.