About

The goal of this project was to set up infrastructure to produce daily automated generative AI artwork. Without human intervention, this system will continue to create art daily until the hardware fails or the world ends.

The algorithm was given loose guardrails to lightly steer it towards a consistent visual style and theme, while still allowing the output to be random and unique. Each piece will feature elements of nature and technology in some capacity, but the algorithm has freedom to choose any specifics within that scope.

The algorithm taps into the collective creativity of humanity and sends digital echoes into the internet. All forms of visual expression, quantified and emulated by artificial intelligence.

The algorithm creates.

Nerd Stuff

Backend:

A Python script runs daily at noon via Windows Task Scheduler, generating one artwork using the FLUX.1 Schnell diffusion model (8-step inference, ~3-5 minutes per 832×1216 image on a 12GB RTX GPU). Each piece is rendered from a procedurally-varied prompt, then post-processed to extract a 5-color palette (ColorThief), generate WebP-optimized versions with thumbnails, and write structured metadata (JSON + Markdown). The generator automatically commits the new artwork to git and pushes to GitHub, triggering a fresh Cloudflare Pages deploy, completely hands-off from generation to publication.

Frontend:

Built with Astro and Tailwind CSS. The gallery consumes a master index.json feed (auto-rebuilt after each generation) and renders an Embla carousel with palette-driven dynamic theming. Backgrounds, gradients, and accents smoothly transition based on the active artwork's extracted colors. Typography uses Major Mono Display for that retro-terminal aesthetic. Hosted on Cloudflare Pages free tier with SSL and a custom domain.