Noizz

The pitch for local AI is easy: free, private, runs on your own machine, no subscription. The thing that actually stops people is the first hour: which open-source model, which quantization, how much RAM, and why there are nine files that are all somehow called Q4.

noizz is where I publish the guides I wished existed when I started: which model realistically fits an 8, 16 or 32GB machine, Ollama and LM Studio setup walkthroughs, and a quantization explainer written for humans. The guides get stress-tested in public, and that includes the embarrassing part: we shipped a stale model recommendation once, got called out on Reddit, and the page now opens with that correction and gets re-checked against current releases before anything is recommended.

Business-wise it is a brand discovery and comparison site with a paid membership, but the local-AI guides are free with no signup, on the theory that the guide that saves your weekend is the only ad that works.

For anyone here running models locally: what nearly made you quit in the first hour? That list is my roadmap.