Weeknotes 2024/25
Last week the WWDC and Apple Intelligence were of some focus; Microsoft recalled Recall, since security >> AI; everybody is looking for an AI icon; Stable Diffusion Horror; AI in Software Engineering; Mental Models around Complexity; oh and some site developments.
If you’re reading this on the site (as opposed to your newsreader), you may have noticed a few visual and structural changes. I hope you like it! It’s not the Astro rewrite though. One more configuration headache and less than straight-forward solutions forced me to cut my losses and move the changes as best as possible back to the Eleventy-based code base. Still it was not for nothing. I learned a lot and also learned what I don’t want to bother with anymore. It also gave me enough of a different angle on the same issue, which translated into a much quicker progress in Eleventy — porting back the changes took me about two days.
Now they will be some regression for the next few days and not all of the static pages are ready, but it felt good enough to push it live alongside this weeknotes.
The link selection was especially tough this week. Apple surely added to this problem, by the sheer volume of announcements and inevitable reactions to it. In some cases, I felt that the topic of a link would warrant more than a few sentences of context.
WWDC: Apple Intelligence, Swift goes places
Before we get to the links, just one observation: in the iPad event, Apple was pretty adamant about their “we shipped NPUs for years now”, claiming an advantage. As WWDC has shown, this advantage is a bit more nuanced, at least for consumers of their products. To take advantage of Apple Intelligence, a device with an M-Series Apple Silicon is required, the only non-M exception is the latest A17 Pro used in the latest iPhone Pro models. The rest is out of luck.
🧑💻 WWDC24 Highlights - Apple
The highlights straight from the source.
📚 WWDC Notes | Documentation
Not just 2024. All the way back to ’16. Bookmark it!
🤖 Introducing Apple’s On-Device and Server Foundation Models - Apple Machine Learning Research
In which Apple explains how their models interact.
🔐 Private Cloud Compute: A new frontier for AI privacy in the cloud - Apple Security Research
This is probably the biggest thing coming out of WWDC. It feels sound and even the most cynical observer has to admit that a lot of thought and resources went into this, especially compared to the competition.
🤫 Here’s how Apple’s AI model tries to keep your data private - The Verge
A more condensed version of the above.
🤓 AI for the rest of us - by Nathan Lambert - Interconnects
One of the many reactions to Apple Intelligence.
🤔 Thoughts on the WWDC 2024 keynote on Apple Intelligence
Simon Willison on the announcements. As usual, he’s one of my most trusted sources when I form my opinions.
🔬 Observations on Siri, Apple Intelligence, and hiding in plain sight (Interconnected)
Matt Webb with his take, emphasising Apple’s firm border between “Personal Knowledge” and “World Knowledge”.
🐣 What’s new in Swift 6.0? – Hacking with Swift
As usual Paul Hudson delivers the best overview of what is new and why it warrants a major version. Also interesting his observation that Swift spent half of its now 10 years in versions 5.0 to 5.10.
🏠 Swift.org - New GitHub Organization for the Swift Project
Swift is now old enough to get its own house, eh, organisation on GitHub. Finally no more sifting through all of Apple’s open source projects to find the Swift nuggets.
🚀 A Swift Tour: Explore Swift’s features and design - WWDC24 - Videos - Apple Developer
A nice overview video that is not app-centric: the examples build a library, an HTTP server, and a command line interface.
🗜️ Go small with Embedded Swift - WWDC24 - Videos - Apple Developer
Baby steps to move WWDC beyond its regular app ecosystem. While it has applications inside Apple, I guess it was mostly curiosity for most of the regular WWDC crowd.
🧑💻 Explore the Swift on Server ecosystem - WWDC24 - Videos - Apple Developer
Yay! I want more of this in the future. Apple should push it more, since it is also more immediate applicable for app developers at large (at least compared to embedded).
🐧 Swift.org - Getting Started with the Static Linux SDK
Related to the Swift on Server topic, this will make deploys to Linux servers a little easier to handle. My hope that this will also help to finally bring Swift to FreeBSD.
♟️ macOS Sequoia brings an update to Apple’s Chess game
This is a code base from the 1990s, before the technology was acquired by Apple. Astounding.
AI of the non-Apple kind
🔐 Microsoft to prioritise security over AI
Brad Smith’s testimony had no surprises. Satya Nadella is now personally accountable for the security practices of the company. That’s probably the only way they can ensure that some department isn’t going rogue, because it’s “not their job”.
⏳ Microsoft Recall delayed indefinitely
That announcement came shortly after the above testimony and the article outlines how it’s related. I wager a guess anyway that Apple’s announcements did contribute: compared to their pretty detailed treatment of common concerns and plausible countermeasures, Microsoft’s Recall, similar in its intention, felt like a hastily stitched together brute-force implementation.
🤖 Apple joins the race to find an AI icon that makes sense | TechCrunch
Maybe someone should ask the AI. The emojis choice is also hard, you are more or less forced to use the robot 🤖.
😱 New Stable Diffusion 3 release excels at AI-generated body horror | Ars Technica
🚨 Trigger warning! The title is not an exaggeration. The results are truely horrifying.
▲ Getting started with AI: Advice from the experts at Vercel Ship – Vercel
A few nice tidbits. If you take only one thing, take Sunny Madra’s quote: “We always tell people, start with your data. This is the chance for you to fix your data models or even your data management. Do that work first because then it’ll really pay off.” I mean, only if you want your AI initiative to fare better than your data-drive business intelligence initiatives.
AI in Software Engineering
🧑💻 Software Engineers Remain Indispensable in the Age of AI - The New Stack
The AI solutions in this space stepped in the same trap as no-code or any other “citizen programming”: writing code may be tedious at times, but it’s the easiest part of the whole game.
🧑💻 Generative AI Is Not Going To Build Your Engineering Team For You - Stack Overflow
Charity Majors not just debunks the (current) impact of Generative AI on Software Engineering, but adds some solid advice on building solid teams, depending on your company context. Mentoring juniors is probably the biggest ROI a company can have and I love every minute of it.
Last but not least
😞 The Stanford Internet Observatory is being dismantled
For political reasons 😔 Another win for disinformation.
🤷♂️ Whatever Elon wants, Tesla gets - The Verge
The article mentions that Elon Musk is just the tip of the corporate iceberg. Corporate conduct lost its moral compass, but it’s only a lens on society. The characteristics have spread far and wide and I bet you have seen this behaviour on a smaller scale in your company or the general public. It’s time to say no, unfortunately Tesla’s shareholders didn’t have the backbone for it.
🤷♂️🤷♂️ Tesla investors sue Elon Musk for diverting carmaker’s resources to xAI | Ars Technica
Given the previous article, how much will this bother him?
🚫 Why Not Open Source?
I agree with all of the points made. The same reasoning is behind the “open source, closed contribution” model, where the project is proper open source (not just source-available licensing BS), but the maintainers decided to not accept contributions and encourage forks.
🏝️ Sidebar is taking a break
I feel it. The curation of links can be pretty time consuming.
🔗 htmx ~ htmx 2.0.0 has been released!
While I’m not using it at the moment, this makes me happy. I like the whole hypermedia systems concept. I must carve out a little time in the future to explore in more detail where it can be beneficial to other approaches.
🤔 A Note on Essential Complexity | olano.dev
Related to the above, the distinctions between different kinds of complexities is not new: essential complexity vs. accidental complexity, i.e. the inherent complexity of the problem vs. complexity arising from the specific implementation.
It maps to my own mental model of internal complexity vs. external complexity, although I have added a third class: contextual complexity. Internal is the inherent complexity of the problem. Problems in products (and processes) arise when you try to reduce this class of complexity. External complexity arises from the specific implementation and might be necessary or unnecessary. To distinguish between necessary vs. unnecessary external complexity, I sometimes use the term of contextual complexity for the necessary kind.
Happy Tuesday!
📊 Link Shortlist: 141