I've been running Linux and Ubuntu for nearly 20 years. I can date it almost exactly, because I still have the email from my Dad back in early 2007 telling me he'd just installed Ubuntu. That's where this started.
The big trend the last couple of years is obviously AI and the large language models, and this balancing act between what you can run yourself on your own machine versus what you get from a state-of-the-art model that you have to pay quite a bit for. Everyone has seen the challenges with the cost structure. Copilot had to change their license. Cursor had to change their license. Anthropic is changing theirs. It's hard to find the right mix where you give a lot of value to customers and you're also able to charge them an appropriate amount.
Open source and AI is another way to solve this. So I've been investing in building out my own local stack of AI tools I can run on my machine.
I'm on an M-series Mac, the Apple Silicon. You may have heard of tools like Ollama that let you run open source models. There's another one called oMLX that's designed specifically for Apple Silicon, and in my testing it ran better. You download the models straight from Hugging Face. The ones I run are the mlx-community builds, which are already quantized down to 4-bit, so a 31B model that would be around 60GB at full precision comes down to about 17GB on disk. I didn't have to convert anything or do any prep. I just pulled them down and pointed oMLX at the folder. Right now I'm running Gemma 4 and Nemotron, plus a few others.
