AI
6 articles on AI — tested first, then written plainly.
AI articles
6 postsEvery Way Apple's On-Device Model Can Fail, and How to Handle Each One
Nine errors, measured. The safety filter blocks ordinary Unix documentation, one failure costs ten seconds and hands back nothing, and chunking loses three quarters of the details.
ReadApple's On-Device Model Keeps Handing Back Your Own Text
A typed return value nearly eliminates it. Measuring what that costs turned up two more things, including a latency number I had been publishing wrong.
ReadBuild an On-Device Summariser in SwiftUI, and Measure What It Costs
A complete SwiftUI app that summarises text entirely on-device, and the first-token numbers behind it: 27x slower on the simulator than on a real phone.
ReadAn Empty String Costs 8 Tokens: What Apple's tokenCount(for:) Actually Measures
Apple's tokenCount(for:) reported 9 tokens for the word test. An empty string costs 8. And a real session charges 22, not 8, so the overhead you can measure is not the one you are billed for.
ReadThere Is No Such Thing as a Token Count
The same sentence costs 3 tokens on Apple's on-device model, 5 on GPT-4o and 13 on GPT-3.5. Measured across eight languages, three tokenizers, and one context window that is far smaller than you think.
ReadData Is the Model: The Most Ignored Part of AI
A beginner-friendly guide to why data quality beats model hype.
ReadNew articles, straight to your inbox.
No spam, no filler — just new writing on iOS, the web, and AI when it ships. Unsubscribe anytime.