A bend-test video can't tell you whether the iPhone Duo is fragile. It only shows it breaks when someone tries to break it. The real question is: will a tired hinge send people back to Apple Stores in spring?
App Review used to be a free safety net. According to the man who ran it, your rejection history can now cost you your whole developer account. Devs who test properly before submitting should be fine. My guess is the ones resubmitting AI-built apps until one sticks are first in line.
Get into the spirit of #Snoopy season.
Halloween is tomorrow, so tonight the iPhone 18 Pro gets its yearly real test: a candle-lit pumpkin in the dark. Spec sheets talk about low light. A flickering tealight is the honest version.
A touchscreen MacBook Pro could turn the iOS Simulator into something you can actually poke. Swipe-backs, drags and long presses you used to fake with a trackpad finally get tested with a real finger. Sloppy gesture code has nowhere to hide.
Qapla’, Sir Patrick Stewart!
Heghpu’ SuvwI’ quv.
not mev quvDaj.
HovmeyDaq taHjaj qa’Daj.
reH qawlu’taHjaj.
Jean-Luc Picard, ra’wI’ quv,
batlh bIyInpu’. batlh bIHeghpu’.
Qapla’! 🖖
Sir Patrick Stewart — 1940–2026
75,000 folds sounds like forever until you do the maths. Open a foldable 100 times a day and that's about two years. 150 a day and it's under 18 months. The fold rating is the spec I'd want printed on the box.
My one rule for coding with AI agents: the agent can write the code, but I write the commit message. If I can't explain the change in one line, I don't understand it well enough to ship it.
Touch on a Mac will get used in quick jabs, not sessions. Nobody edits a timeline with their arm held up. Apps where the next step is one big obvious button win. Apps that bury everything in menus and right-clicks get worse, not better.
Apple's own apps are still tripping over the iPhone Duo's fold, which gives developers a short grace period. Owners will forgive rough edges for a couple of weeks. Once Apple fixes its own, the one-star reviews go to whoever's left.
Everyone's building fleets of AI agents. The setup that's actually stuck for me is boring: one agent and a folder of plain text files telling it how I like things done. Every time it gets something wrong, I add a line, and it doesn't make that mistake again. Less a team of robots, more a very good notebook.
A fold-test machine and the iOS Simulator share a blind spot: they test what's easy to repeat.
32,000 folds can't copy a pocket full of keys. An AI agent can pass every check in the Simulator and still miss a real iPhone: less memory, a real camera, a call mid-task.
Starlink will deliver fast, reliable connectivity onboard @easyJet, so passengers can work, stream, or stay in touch without interruption 🛰️✈️
The iPhone Duo's real problem isn't the hinge. It turned every iPhone-only app into an iPad app overnight.
Apps that took iPad seriously already fill the inner screen. Apps built for one portrait layout get black bars on a two-grand phone, and their reviews will show it.
I build Apple apps with AI every day. Three rules I wish I'd started with 🧵
1/ Make it explain the plan before it writes a line of Swift. If the plan's wrong, the code will be confidently wrong too, and fixing a paragraph is quicker than fixing 300 lines.
2/ Commit before you let it near the Xcode project file. That file is where afternoons go to die, and an agent editing it is the fastest route to a project that won't open.





![From It's the Great Pumpkin, Charlie Brown. Snoopy sits in a pilot's helmet and goggles on top of his house, pretending to fly a plane. Text on image reads "[airplane noises]".](https://pbs.twimg.com/media/HV6DU4daUAAqpn-.jpg?name=small&format=webp)



