Something I am not understanding about the Foundation Models framework: How does the LLM know what output to expect from a Tool, given that ToolOutput is not generic, and doesn't take a Generable type? You would think knowing what is returned before you call the tool would be essential. Or are we just supposed to describe the output in the description?

Ugh. When your embedded LLM has an existential crisis.

What a difference a day makes: Was optimistic about Foundation Models yesterday, and today I think I know why they didn't ship the improved Siri. The local model really is pretty thick. I thought it would be capable of stringing together tool calls in a logical way, and sometimes it is, but other times it fails to understand. Exactly the same prompt will work one time, and fail the next. Sounds like what Apple was saying about the new Siri.

Having played with the new Foundation Models framework and thought about ways we can use it in our apps, I think it could be Apple’s Trojan horse for AI. It barely gets a mention in mainstream media, understandably, but it leverages Apple’s developer base. I think we are going to see very creative uses in apps, and Apple just have to iterate year on year (eg add private cloud compute next year).

I have a legacy macOS app. I want to use some new stuff, and it seems the macros only work well with Swift 6. So, is it possible to update to Swift 6 and simply disable all the concurrency checking? My attempts so far have been a bit forlorn. Seems to always check some stuff (eg awakeFromNib WTF?!)

Just noticed this. Bit of AI fun I assume.

Oh dear. This could be a real problem. Installed iOS 26 beta, and the whole screen looks dull to me, even on maximum brightness. I have multifocal contacts, which exacerbates it. Hope this is just an early beta glitch.

Wow, this Musk-Trump falling out is making the White House Zelensky ambush look like a friendly get together.