I started with LLM-tinkering on my PC, then decided to push its limits with image generation. So most of the outputs displayed here were produced locally, without the cloud. They were produced in CPU-only mode (brute-force rocks lmao), with some clever RAM management.