Started playing around with running ollama on my Macbook Pro. gemma3 is more powerful than I expected for the size and load it puts on my system. It can even generate a "hello, world" example in Go. I tried qwen3-coder and my system began visibly stuttering when I asked it to implement a PR. Not sure I've got the right setup for a local agentic setup, but having gemma3 running to ask it questions seems like it has some viability.