Local AI Weekly

Local AI Weekly is an upcoming newsletter for Linux users and open-source enthusiasts who want to use AI without handing their data to Big Tech.

What's inside each issue?

Here's what I plan to initially. I'll make changes later on.

Hands-on experiments
Real tests on real hardware. I have access to ZimaCube Pro 2 with Nvidia Ada GPU, Raspberry Pi with TPU and AI HAT, SBCs with NPUs and Asus Zenbook that has integrated NPU. I am not a fan of benchmark charts, I would prefer to show actual setups you should be able replicate at home, with my experience on what works and what doesn't.

Open model news
New model releases, local AI tools worth your attention, and self-hosted setups that have gotten useful. This will be curated so you don't have to follow a dozen sources yourself.

AI news and updates in general
It would be wise to stay updated with what's happening in the (commercial) AI. So quick commentray on such news of significance.

Memes

I love memes, both creating as well as sharing them. A newsletter without some fun element would be boring. You'll get trending memes on AI in each edition.

Why a separate newsletter?

If you are already subscribed to FOSS Weekly, you already know about the "AI, Homelab and Hardware" section.

We have been covering local AI topics for nearly two years now but we will be putting more focus and effort on it as the interest in local AI has increased considerably and I do not want you to left behind.

But there is also an apprehension against AI, local or not, among many readers. And I do not want to push AI down everyone's throat.

And hence this separate newsletter only for the interested ones.

What is local AI?

If you are still wondering about this "local AI" term, think of it as a new kind of open source software. Usually, when you use chatGPT or Claude Code, the model is on their server, the processing takes on their server and naturally, your data lives on their server.

The concep of local AI is that you download the LLMs on your system. You interact with the local AI on your system, the processing is done on your system and the data lives on your syste. It doesn't go to a remote server managed by an enterprise.

Depending on the model, it could be 5 GB or 50 GB or 500 GB in size. Also depending on model, your hardware may not always be able to run all the models. And there are different models, suitable for different use cases.

This is what we will test on It's FOSS. Like you discovered new distros, applications and Linux tricks, you will also discover new models, ways of using AI in your workflow and more.

FAQ