Post

I am vibe coding a news aggregator that moved from a local web server that uses local LLMs on my MacBook to a native iOS trying to use Apple Intelligence models that are crazy aggressive on restrictions to now “let’s just use ChatGPT scheduled tasks to constantly update a site”. Help.

31/7/26, 7:32 pm0 reposts2 likes

dnu

@dnulkjkjh.bsky.social

31 July

Apple Intelligence isn't a general LLM API - it's locked to Apple's own writing/summary use cases. for arbitrary on-device inference you need Core ML or MLX. hit this same wall building Basil - had to go Core ML. what tasks are you trying to run locally?

Wes Souza

@wes.dev

31 July

I used Apple’s newer Foundation Models framework, which exposes Apple’s built-in language model to apps through LanguageModelSession. I am matching headlines to my preferences (written sentence + positive and negative tags), and after they pass, I run it to summarise each downloaded news article.