Boltpage

Working with AI full-time means producing and consuming documentation at a pace most tools cannot keep up with. Specifications, project notes, conversation logs, technical references: Markdown became my default format because it is fast to write, structurally clean, and covers everything daily documentation requires.

The problem was not the format. The problem was that nothing rendered it well.

Every Markdown viewer available either ran in a browser or depended on a browser engine under the hood. On a machine already running Docker containers, Node servers, and one or more AI workloads, summoning a browser that consumes 2 GB of RAM to display a text file is not a practical inconvenience. It is absurd. Native alternatives were no better: slow rendering, poor formatting on long documents, or bloated editors pretending to be IDEs when all I needed was to open a file, read it, and move on.

So I built one.

The application

BoltPage is a desktop Markdown viewer and editor built on Rust and Tauri. It supports GitHub-flavored Markdown with syntax highlighting, multiple themes, multi-window operation with independent preferences per window, live preview, and automatic detection of external file changes. It also handles JSON, YAML, TXT, and PDF in view-only mode.

The frontend is vanilla JavaScript. No framework dependencies. Markdown parsing runs through pulldown-cmark, syntax highlighting through syntect. It ships native on macOS, Windows, and Linux.

It is fast, it is light, and it does exactly what it is supposed to do.

How it was built

I built BoltPage in Rust. I had never written a line of Rust before this project.

The entire application was developed with AI coding agents under my direction. Not blind-coded, not one-shotted, not generated in an afternoon. Every piece of code was reviewed, tested, and manually approved. Tauri was the framework choice because I was not going to supervise the construction of an entire native renderer from scratch as a side project. Tauri gave the scaffolding. AI wrote the implementation. I made the decisions.

This is what AI-assisted development looks like when done with intent. You direct, you review, you decide. The agent writes. The result is production software that ships cross-platform with CI/CD, code signing, and proper release management.

The real product

The repository includes the actual system prompts I use to direct AI coding agents: ai_truthfulness.md and ai_software.md. They are not the sort of prompt engineering you see circulated on social media to impress people. They were written to work, and they are battle-tested across several projects. BoltPage is simply the one I chose to open-source.

The prompts are the real point of the distribution. The application is useful. The methodology is transferable.

What this taught me

• AI does not replace the need to understand what you are building. It replaces the years it would take to learn a language well enough to build it. The decisions, the architecture, the judgment calls: those are still yours, and they still matter more than the code.
• Rust is as demanding as its reputation suggests. But AI turns it from a months-long investment into a viable choice for a focused project. The compiler is unforgiving, and that is precisely what makes it work: if it compiles, it runs.
• The industry defaults to platforms when most people need tools. BoltPage does one thing. It does not sync to the cloud, it does not manage notes, it does not have a plugin system. It opens files and renders them. That is enough.

BoltPage is open source under MIT.

Sometimes, AI rocks.