Spent the last couple of weeks learning to fine-tune a model from scratch; tried it on my RTX 3060, hoping I don't make it worse. The end results were pretty surprising. I'm thinking about improving/bumping the score even further. I strongly believe that asking my agent to run the vacuum shouldn't need a round trip to a frontier model in the cloud. Would love your thoughts on this.