I see the party of small government has effectively shut down Fable and Mythos. Yet another reason to consider local models. Although I wouldn't be shocked to see a ban on distributing Chinese models at some point too.

A little DIY instead of sitting at the computer today. There are some things AI can’t help me with.

So after actually looking at the contents of the PR that adds MLX FoundationModels support it's pretty obviously from Apple and I'm dumb. If you're thinking about implementing your own custom LanguageModel it's a ton to wade through but worth a look - https://github.com/ml-explore/mlx-swift-lm/pull/334 #mlx

Confused about where MLX fits in Apple's ML ecosystem / plans. The MLX frameworks live under ml-explore, not Apple's repos which, okay. But then WWDC session 339 shows an MLXFoundationModels framework. Found a post suggesting it's part of mlx-swift-lm, but the only mention there is an open PR that doesn't appear to come from Apple. Maybe Apple has a yet to be released framework in the works? Anyone know the deal?

I gave the new Foundation model my standard test asking it to create a simple unit converter view. Yes I know it wasn't designed for this. Anyway it actually came close to making something that compiles. It came less close to making something functional. I added it to my collection at: https://github.com/somegeekintn/LMTestCompare

In days gone by I'd wait for the first .1 release before upgrading macOS but here I am now with the beta installed on day one. Have to decide if and in what way my little local LLM project moves forward.

My phone should be able to send AI requests to my Mac when it’s available instead of private cloud compute aka a data center.

Language model protocol you say? 🤔

Since recently becoming jobless I've been looking forward to WWDC to see how much it might motivate me to try and er... re-job…ify? Anyway not much so far it turns out.

Why are all these lunatics keeping their phone in their back pocket? 🤔