I, a human, write every word of this blog in my own voice. Nonetheless, I built an AI Slop Machine.
(Boring! Skip to the pretty chart) American and Chinese AI labs continue to make increasingly strong open-weights large language models (Qwen, Gemma, gpt-oss, and others). You can self-host these LLMs on consumer-attainable computers and GPUs. They’re pretty good at agentic knowledge work tasks, and when you use them, several of the things that people say are bad about AI don’t apply.
source & further reading
cmart.blog — original article
LLM: Learningful Links and Musings
Inference Cards