Microblog – page 7
Please make MapKit work with #VisionPro! I’d love an interactive 3D globe, or a slice of 3D terrain to which you can add annotations, lines and overlays. #WWDC
Something like this example from the State of the Union, but powered by #MapKit. Hope this will be one of the sample apps.
Was the Maps app included in the list of apps supported by #VisionPro? That would be absolutely stunning. #WWDC
Did I just hear Tim Sweeney scream?
Fascinating. How does that Vision Pro FaceTime experience look like for the other people though? You are voice only?
Alright. That’s a wrap. #WWDC
Did not expect Electron to get Sherlocked. #WWDC
Woohoo, @qzervaas !
#Mastodon question: If I mark a toot as unlisted and use a hashtag, does it appear in the home feed for people who follow that hashtag?
Judging by this comment from Gargron, I think unlisted toots will NOT show up for people who follow the hashtag: https://github.com/mastodon/mastodon/issues/5463#issuecomment-350572440 (though the comment predates that feature)
Bit of a shame as that means Apple developers around the globe will have no choice but flood their local timelines with #WWDC messages.
While trying to track down an annoying very rare test failure, I just noticed—and immediately embraced—the “Run Test Repeatedly…” option when right-clicking a test on #Xcode. Chef’s kiss!
0.5 sleeps till #WWDC.
As much as I like living in Sydney, the timezone can be a pain.
With #WWDC around the corner, I’ve cleared my deck and wrote up my thoughts on Microsoft’s Build conference. Interesting guidance for adding #LLMs to apps and the upcoming tools in Azure do look like promising building blocks.
https://adrian.schoenig.me/blog/2023/06/04/microsoft-build-for-app-developers/
Another lovely morning
I think my favourite feature about Ivory are its little sounds. Such a calming and unique experience.
Good morning
I’ve been using Markdown pretty much daily for over a decade and I still regularly mess up the link annotation, swapping where the URL and where the title goes. I have no sense why one would be in regular brackets and the other in square brackets, or why one would be before the other.
Add some stage I’d like to migrate Longplay from a storyboard to SwiftUI, but, boy, I also do not want to.
As great as it is to have different ways of interacting with ChatGPT, it makes it hard to find previous discussions as each of them have their own histories. Where did I have that chat? On ChatGPT directly, Raycast, or WitKit? A shared history would be *great*.
♫ Happy birthday to me ♬, sings this blog. 11 years!
I wonder if future #LLM-driven IDEs will look more like GitHub than a code editor: you file an “issue” to tell the bot what you need, it creates a PR for you to review, you test it, review it, make some changes, ask it to change other things and refine your specs in the issue until you’re happy and merge.
That’s the workflow I find myself using with ChatGPT, but all the copying and pasting between browser and IDE, and repetitions in the chat make it clumsy. GitHub solved most of the UX for that already.
https://mlc.ai/mlc-llm/ has a TestFlight beta for running an LLM locally on a modern iPhone. Remarkable that it works.
While GPT4 is leaps and bounds ahead in terms of quality, those locally run model could catch up. LLMs likely follow an S-curve in terms of their quality. Makes me wonder if GPT4 is near the top or still near the start or middle.
Bored with the long assertion failure messages that XCTest produces? I used to copy them into Kaleidoscope to diff them, but I’m now using Raycast AI for it and I love it:
- I’ve added an ^K hot key for its “Search AI command” action
- I can then select a failure in Xcode, hit that hot key
- Type a prompt to have it explain the failure (or pick from my recent prompts)
- Tada!
Marvin is a Python library that has “AI functions”: functions were you write the *body* as a #ChatGPT prompt. It does not use ChatGPT to turn that prompt into code, but rather it passes function input and body to ChatGPT, and then passes the response back to your code as a string, JSON, or data strutures.
Fascinating and feels like embedding the equivalent of a magical “I’m feeling lucky” button in your code. https://www.askmarvin.ai/guide/concepts/ai_functions/
Was just greeted by this on a website. I guess it was inevitable that soon entire newsletters are written by #ChatGPT. Isn’t the key selling point of newsletters the human touch and insightful commentary?
Very exciting what’s ahead for ChatGPT with plugins to feed in dynamic information and actual facts, as with the Wolfram plugin. Fascinating way of having APIs interact with each other.
https://writings.stephenwolfram.com/2023/03/chatgpt-gets-its-wolfram-superpowers/
Really cool how widgets demonstrate the flexibility of @maparoni in such a pretty way:
- You can point it at arbitrary APIs by describing how to extract location information from JSON/XML APIs.
- Use formulas to filter/sort/style the locations.
- The widget then draws the map using an appropriate projection.
(Years of working on this in my spare time… 🤪)
Many, many thanks to @nicklockwood for his brilliant and super fast Expression library: https://github.com/nicklockwood/Expression
Many, many thanks also to @simonbs for his excellent Runestone library: https://github.com/simonbs/Runestone
If anyone’s curious about the map projections, I’ve open sourced the component for drawing them in #Swift: https://github.com/maparoni/GeoProjector
MacWhisper is a fantastic app. I absolutely made the day of a colleague when I introduced it to them, as it’s such a productivity boost for someone preparing official meeting notes. Well done, @jordibruin!
According to ChatGPT, a group of e-bikers is called a “Spark” whereas a group of drivers stuck in traffic is called a “Honk”. Seems about right.
Looking forward to the inevitable revival of the iPhone Mini form factor. Who cares if they call it the Apple Watch Pro Max Plus.
Boo, the iPad Calendar app crashes in the iPad-sized views nowadays when you tap a date around 1582 and October is not showing its proper three-week duration anymore. All this swiping for nothing. (Still fine on iPhone though.)
Oh boy, #iOSDev can be so frustrating, in particular when trying to debug widgets. My widget displays a pretty version of a document stored in iCloud or the app’s sandbox. Works beautifully… until you restart your phone. I was using the built-in URL option in the intent config and naively assumed this would work with file URLs. But I need to recreate fresh new URL “instances” in my widget, even if printing them looks the same as what comes from the intent config. Took me a few hours…
Finally I can go back to fixing those bugs where I can blame myself.
”[…] that’s not the only place the App Stores fall short. Search has never been good […]”
Couldn’t agree more with @macsparky. The crazy thing about this is that the App Store will turn 15(!) in a few months. Apple does not seem institutionally able to tackle this. The App Store needs competition on iOS (and hopefully the EU will force Apple).
https://www.macsparky.com/blog/2023/02/more-app-store-troubles/
I’m apparently 20 years late to this, but Masters of Doom is an excellent book about how id software essentially created the FPS genre. Oh the nostalgia! (Wil Wheaton always makes for a great audiobook, too.)
I wonder if anyone else has a single repo with a mix of these languages.
The main web app is written in #Elm, with an #Elixir / Phoenix wrapper for additional pages, and the main logic is provided by an AWS Lambda written in #Swift (sharing the same code with the iOS/macOS app). Was a fun and rewarding challenge building this. Elixir will likely grow and take a bite out of Elm.
The things that testers of Bing Chat encounter are utterly fascinating. In particular when its claims are challenged and it is being fed proof by pointing it to articles that show it’s wrong. Who would have thought that training a language model on the world’s archive of decades of heated discussions would lead to such a “personality”?
A worthy read, including the links: https://arstechnica.com/information-technology/2023/02/ai-powered-bing-chat-loses-its-mind-when-fed-ars-technica-article/
My goodness, the new performance monitoring in @getsentry looks sweet. Let’s see if this helps me track down some performance issues reported by beta testers of Longplay.
These prompt injection attacks on Bing Chat and Chat GPT are absolutely fascinating. Shows that there’s more to the large language models than just “the most likely sequence of paragraphs following this prompt”. Reminds me of the concept of “emergence” where what you get is more than the sum of the individual parts.
What I find most fascinating about this isn’t so much the prompt injection attacks themselves, but rather that these systems are configured (programmed?) using a bullet list of sentences.
Oh my, hit a nasty #UIKit crash when using #SwiftUI in a collection view using the UIHostingConfiguration.
Crashes whenever I change interface orientation on my phone. Can be reproduced with this trivial snippet:
cell.contentConfiguration = UIHostingConfiguration {
Text(“Noooo!”)
.frame(width: 150)
}
The frame bit triggers a UICollectionViewRecursion warning and then crashes.
Added details on my blog and filed a radar. Hopefully it’ll be fixed, as UIHostingConfiguration is otherwise a good way to get the most out of both SwiftUI and UIKit.
https://adrian.schoenig.me/blog/2023/02/10/uicollectionviewrecursion/
ChatGPT
For some things #ChatGPT is just stunning. (Google was no help with this.)
His website is a treasure trove: http://worrydream.com/
Can’t wait for the kids to grow up, so I can have my midday naps on them. Only seems fair.
Swan Song on Apple TV+ is a compelling SciFi thought-experiment with heart. It also might be the closest thing to a vision statement from Apple since the Knowledge Navigator: The AR user interfaces providing an enhanced desktop, AR multiplayer games over FaceTime, self-driving cars, and even a circular watch.
5yo discovered a book in the library about building computer games for kids using Scratch and is now super keen to do a bit of coding with me and his big sister every day. It’s lots of fun and full of giggles when we try the games. I enjoy those moments of sharing my programming passion with the kids as it’s been to abstract for them previously. Scratch is really well done, too.
This is the book, definitely recommended: https://www.dk.com/uk/book/9780241317747-computer-coding-games-for-kids/
https://www.perplexity.ai is neat. It’s a bit like ChatGPT but with less bullshitting thanks to a list of sources for the stated facts (and without constantly expiring sessions). It feels like a glimpse into the near future of search. This could bring an axe to Google’s main source of revenue.
Warm welcome to the Fediverse, @cabel and @pschiller!
I’d love it if an eager Objective-C veteran would fork #Adium and distort it into a #Mastodon client for macOS. I miss that icon. Get started here: https://github.com/adium/adium
Meanwhile, @terhechte shows that this kind of UI could work quite well for Mastodon: https://mastodon.social/@terhechte/109762188190735435
Venting about the state of 1Password on Mac
Disappointed by 1Password 8 on my Mac. The app isn’t too bad, but the Safari extension is just incredibly unreliable. Barely ever does what it should and I end up going to the app most of the time. I do rely on it for the teams and family sharing features, and I am not aware of a good alternative.
Sad to see an app go from “ooh, this is nice” to “please, Safari, remember my login so I don’t have to use this”.
Settled into a nice routine for Mastodon where I can follow a good number of people, be a completionist, without being overwhelmed by it:
I have my @ivory timeline set to filter boosts, so it’s quite calm when I check it during the day. In the morning, I also check @quintsns’s https://quintsns.pianeta.uno which provides an overview of top links & posts of the last 24h to see what I’d have otherwise missed. Great combo!
subscribe via JSON









![We're launching our exclusive newsletter, written by ChatGPT. Sign up below and we'll add you to the waitlist.
Prompt: [Your email address]
Button: Join the waitlist](/images/micro/110105676358562660/1.jpeg)








![Adrian: How long would it take by car to drive to the moon
Perplexity: It would take around 153 days to drive to the moon at 65mph without stopping[1]. At 60mph, it would take around 6 months[2][3]. If you took the M6, it would take 12 months[4]. With no stops for bathroom breaks or sight-seeing, it would take about 100 days[5].
(Footnotes refer to source links that perplexity.ai injects into its replies.)](/images/micro/109809253619992388/1.jpeg)