Weekly

A focused newsletter on open models you can run on your own machine, for people with a GPU and a weekend

What an issue holds

Once a week: a new open model worth downloading with its size and licence, releases of the tools that run them (llama.cpp, Ollama and the rest), a technique that measurably helps, and what a GPU or memory upgrade actually buys you in tokens per second. For someone who already runs models at home, not for someone choosing a cloud service.

What it leaves out

Not cloud-only model announcements you cannot download and run, not AI safety and policy debates, not leaderboard drama with no numbers, not corporate funding news.

Where the news comes from

Hugging Face blog, Ollama blog, Simon Willison's weblog, Interconnects, Together AI blog.

Nobody gets this one yet. It starts when you do.

Start this one