Good experiments need constraints. Without them, a lab becomes a collection of half-finished demos: technically interesting, perhaps, but disconnected from anything people actually need.
KozakMedia Lab starts from a different position. It is a place to test visual ideas, publishing systems and applied automation against real outcomes. Can information be made clearer? Can a repetitive process become calmer? Can an image do more than decorate a page?
The point is not to put more technology on display. It is to make useful work easier to see, understand and act on.
A deliberately quiet foundation
The site is built with Astro and TypeScript, then compiled into static HTML, CSS and a small amount of JavaScript. That choice is intentionally unglamorous. There is no public database to maintain, no plugin catalogue to secure and very little that can fail at runtime.
Static does not mean fixed. Individual experiments can still use interactive components when interaction earns its place. The default, however, remains a fast document that works without waiting for an application to start.
This gives the Lab a useful rhythm: write and design simply, add complexity selectively, and keep every release easy to reverse.
Photography with a clear boundary
The opening visual language comes from a deliberately selected Immich album. Only images placed in that album are considered approved source material. The website receives exported copies—not originals—and each public derivative has private metadata removed before it is resized and compressed.
Every published image is recorded in a small manifest with its source asset ID, public filename, alt text, usage and export date. It is a modest detail, but it makes future updates repeatable and keeps the boundary between a private library and a public site explicit.
Build beside the existing system
The first version was built in its own Git repository and Docker container. The previous WordPress site remained available throughout, with a fresh database and file backup taken before traffic moved to the new container.
That rollback-first approach matters beyond websites. Safe experimentation is rarely about eliminating uncertainty. It is about isolating the change, making the result observable and knowing exactly how to step back.
What comes next
The next stage is to use the site, not simply decorate it. We will publish practical notes about visual documentation, self-hosted media workflows and small AI-assisted tools. The Lab will also become home to focused interactive experiments—each one explaining the problem it addresses, the limits it accepts and what we learned by building it.
This first release is a foundation. The useful part begins now.