Matisyahu's Blog

This morning I woke up at 4:45am NZ time so I was all logged in to watch the live stream at 5:00am NZ time. If you’ve been following the news regarding WWDC then you’ve got. good understanding that the WWDC this year isn’t about new features but rather it being a ‘Snow Leopard’ style release where the focus is around finally getting Apple Intelligence working along with refining the rough edges of Apple’s platforms along with bug fixes, optimisations focused around making the system work better by making ‘under the hood’ changes. This blog post is what I noticed after having watched both the keynote and the platform state of the union (the blog isn’t structured in any particular order but rather just grouping common topics together).

When it comes to Apple Intelligence I think it is important to note that there were many rumours regarding the next generation of Apple’s models would be based on Gemini (I speculated that myself but then again it was more guess work rather than statement of fact since I was only dealing with unverified rumours) but in an after keynote interview (link) with Craig Federighi, Amar Subramanya and Mike Rockwell, and Sebastien Marineau-Mes there was greater insight in the role that Google played in terms of Apple Intelligence.

What was interesting was the focus by various speakers regarding Apple Intelligence where there appears to be a concert effort to have as much of the processing being done on device and the private cloud computing being used as a ‘last resort’ rather than first port of call. I think this is very much driven the reality that running an AI datacenter is incredibly costly and given how AI provides are already looking at moving customers to ‘pay as you go’ models or subscriptions with severe limitations, I think Apple want to set realistic expectations. With the focus on local models I also wonder whether the use of cloud based AI is going to be something of a crutch until hardware eventually gets to the point that having to use cloud based AI will become either a thing of the past or only used for niche issues that users are willing to pay additional for.

What I found interesting is the emphasis being made by Craig Federighi regarding the importance of AI being tastefully integrated where it based sense – a swipe at Microsoft’s and Google’s attempt to ram it into everything regardless of whether the user wants it nor not? the demos that were given were actually things that I could see myself doing – practical use of AI rather than it being a tick box project of “I don’t know whether people will use it but we’ve got to ram it in there to keep the shareholders happy”. Apple appears to be very much aware of the backlash that Microsoft is receiving with commentators referring Microsoft as Microslop by many online – make sure if you’re going to integrate AI that it is actually useful and it is unobtrusive for who don’t have interest in using it.

There is talk about usage limits but no details have been given how however it appears, at least as it is being reported on Macrumors, that it maybe linked to the tier plan you’re on when it comes to iCloud+. With that being said, it’ll be interesting to see whether in conjunction with the rate limits whether there will be the ability to pay for extra usage and if they go down that route then what meterage will be employed – a set number of requests per month for an allotted amount? If they are going in that direction then it’ll have to be some sort of way in which the average person an understand it. with that being said, given that we don’t know the specifics I guess it’ll come down to how things turn out during the public beta once launched so then Apple can get some sort of vibe check regarding what sort of load they’re getting from early adopters and how that’ll be extrapolated out to the general population once the stable version is released.

The demos of the on device models are a big improvement – there was a video a while ago which compared a Samsung Galaxy S26 Ultra to an Apple iPhone 17 Pro Max regarding editing a photo. The editing of the photo were basic things like removing an element out of a photo and the examples that were given the results of the iPhone 17 Pro Max were pretty horrible. I’ve seen even more recent examples and the models haven’t gotten much improvement over time. The presentations were impressive however it’ll be interesting to see what the results are in the real world when the public beta is released and people start sharing photos they’ve edited.

At the moment the process of automating is done via Shortcuts however that requires the end user to have a reasonable level of technical skills however there will be the ability to create multistep automation simply by asking your phone to do just that. It’ll be interesting to see how reliably it does it but for me I would sooner manually make any sort of short cuts and read the documentation if I have trouble understanding how to do it.

One feature I didn’t expect to hear a lot about but was actually talked a lot about is the device management features that parents can use for their children’s devices. It is interesting the various speakers and how they talked about responsible use of technology, a heathy use of technology at a young age so there is a good mixture of using technology in conjunction to playing outside with friends, exercising and other non-tech based activity. The impression I got was that the announcement was as much an announcement of features as it were a signal to regulators (who are currently being pressured by voters to keep ‘big tech’ in check) that they’re taking their concerns seriously and are proactively doing something to empower parents rather than waiting for regulatory agencies to step in and start passing mandates on what the tech industry needs to do.

When it came to the optimisation side of the equation they have examples of how they’re optimising how text is rendered, replacing old C based technology such as the font rendering engine with a modern Swift based font rendering engine.AppKit controls, SwiftUI controls and UIKit Controls with a unified backend, allowing users to tweak the tinting/glass effect when it it comes to the Liquid Glass design language. I think for the under the hood changes it’ll be a situation of waiting until the release of version 27 of their platforms to ‘feel’ the change. It’ll be interesting to see whether we’ll see a noticeable improvement in the number of security bugs being found in areas where it is common for those bugs to be found once they’ve been replaced with one rewritten in Swift.