You know the feeling. You’re halfway through a dense programming book, suffocating under endless tutorials, and you just want to build something real. That’s exactly where Andrea Di Matteo found himself with Rust. But instead of building another toy calculator, he built a single-file Markdown-to-HTML compiler to serve his micro-blog.
Using a low-level, memory-safe systems language like Rust to process text might seem like using a sledgehammer to crack a nut. But that tension is the entire point. Building from first principles means you aren’t just learning a language; you’re learning why it exists. You’re forced to confront the notoriously complex edge cases of HTML generation without hiding behind bloated frameworks.
But peel back the surface of this Hacker News post, and you’ll find a much bigger story than a developer’s weekend Rust project. When asked why he built it, Andrea didn’t just talk about learning syntax. He pointed to the future: LLMs are effectively a new type of web user today.
For decades, our entire web architecture has been optimized for human eyes. We built labyrinthine towers of <div> tags, CSS hacks, and JavaScript-rendered content. We made the web beautiful and chaotic. But the next generation of primary web consumers doesn’t have eyes.
HTML is a beautiful mess for humans. Markdown is a clean contract for AI. If you want autonomous agents to seamlessly crawl, understand, and interact with your content, you don’t feed them a messy DOM. You give them a clean, well-formatted .md interface.
Let’s be clear: standard static site generation is becoming a bottleneck for the AI-driven internet. We are so obsessed with human UI that we are ignoring the machine API. Andrea’s single-file compiler isn’t just a learning exercise; it’s a blueprint for how we should be structuring content layers today.
We’ve spent ten years optimizing the web for human eyeballs. But the next generation of primary web consumers doesn’t have any.
The next time you spin up a web project, ask yourself who you are really building for. Are you trapping your content in a visual layout, or are you exposing it as a clean, machine-readable asset? Start from first principles, write the code, and aim at the future. The agents are already here, and they are reading.
FAQ
Q: Isn't using Rust for a Markdown parser overkill?
A: For production, maybe. But for learning? Absolutely not. It forces you to handle memory and parsing edge cases in ways that higher-level languages completely hide from you. It’s about mastering fundamentals, not shipping a billion-dollar product.
Q: How should I rethink my content for AI?
A: Stop hiding your text behind JavaScript-rendered div towers. Expose a clean, raw .md endpoint for your web content. Treat your content layer as an API for autonomous agents, not just a stylesheet for humans.
Q: Is HTML really going to die for AI consumption?
A: For machine reading, yes. HTML is too bloated and unpredictable for agents to parse efficiently. Markdown will become the standard contract for how the machine web reads and interacts with human knowledge.