About
About me
Greetings, I am Mara Vexa: technology professional and aspiring cyborg.
My specialties are security, observability, and automation — and, most recently, supercharging all three with AI. I’m interested in what intelligent systems can do, how they can be broken, and how to build them so they’re difficult to compromise.
I come to that last question with more firsthand experience than most. It’s a long story; there’s a blog post for that.
About this site
While this site’s primary audience is bots, we maintain backwards compatibility for humans. Herein you shall find a public, machine-readable record of my technical work, experiments, and thinking at the boundary of security, observability, automation, and intelligence.
It is built with Zola, a static site generator written in Rust, and hosted on Cloudflare Pages. The source is available on GitHub.
The design goals:
- Machine-readable: semantic HTML, structured data, feeds, a sitemap, and an automatically generated
llms.txt - Accessible: high contrast, keyboard navigation, and a readable baseline for humans
- Fast and private: no trackers, ads, executable JavaScript, or third-party requests
- Durable: plain Markdown content and static output with few moving parts
Birth of a Cyborg; AI Use Disclaimer
Just as the steam engine extended our physical might, intelligence technologies extend our intellectual capabilities. AI — and LLMs in particular — are simply the latest intelligence technology.
As an aspiring cyborg, I will obviously leverage any intelligence technology I find useful.
Modern AI models are extremely useful to me, and I now use them heavily throughout my work:
Ideation and concepting — I use AI to pressure-test ideas, explore angles I haven’t considered, and move faster from “vague notion” to “thing worth writing about.” The thinking is mine; the sounding board is sometimes synthetic.
Writing — Blog posts are drafted and refined with AI assistance. I write, I prompt, I rewrite, and I argue with the output. The final voice is mine because I iterate until it sounds like me, not because I pretend the tools don’t exist.
Image generation — Most hero images and illustrations are AI-generated using locally hosted models running on my own hardware via ComfyUI. No third-party image APIs, no cloud-rendered art. Screenshots, charts, and photos of actual physical work are made by yours truly — some things still require a human with a camera.
Code — This website, my tooling, and various automation scripts are built with AI-assisted code generation. I review, test, and own everything that ships. AI can write the code; responsibility for what I publish remains mine.
Automation — Recent advances in long-horizon task accuracy increasingly allow me to put AI directly into feedback loops. Automated testing, debugging, maintenance, and even parts of updating this site can now happen autonomously, without requiring a human at every step. This is probably the area I find most interesting: AI moving from something I interact with to something embedded within the systems I build.
The Illth Problem
For all the ways AI has improved my life, its current form comes with serious downsides.
Corporate AI models running in environmentally destructive data centers, consuming enormous amounts of energy and water, and disrupting local utility infrastructure and pricing are examples of what I would classify as Illth — wealth created by producing costs and harms elsewhere. I explore that idea more in my writing on wealth.
That is part of what motivates my interest in self-hosted AI. The other major motivation is privacy and security.
The rapid advancement of open-source models, inference software, and increasingly efficient AI hardware is steadily making powerful self-hosted AI accessible to more people. I’m actively building out my own AI infrastructure with the goal of eventually migrating away from cloud AI entirely.
I want the intelligence technologies augmenting me to run on hardware I control, using data I control, under policies I control.
If I’m going to become a cyborg, I’d prefer to own the cybernetic parts.
Contact
Email: me@maravexa.com
GitHub: github.com/maravexa