I blogged recently about experimenting with local LLMs inside the Micro.blog for Mac app. This is useful for post categorization and other things. It works, but it’s not as good as I’d like without requiring everyone to have an M4 with 48 GB of RAM. Might put it on the shelf for a bit.