Today we spent some time working on upgrading out local AI model. I had been looking at a model, Qwen 3.8 Flash Next, but I didn’t think I would be able to fit it onto my GPU setup. Then I figured what the Hell let’s try it out. And…. It worked.

We got a 15-20% bump in intelligence, with faster decode! This is great news. The local model intelligence is absolutely at a point where it can be useful for real work.

Stay tuned. I do want to publish a full write up on our path to running our own local AI. Eventually I’ll get there.