Rashad Shirizada·Oct 7Why Your Core AI Model Is Slow on First Launch — and How to Precompile ItUse Xcode 27’s coreai-build workflow to precompile models, then measure cold-start behavior with Core AI Instruments.
Rashad Shirizada·Oct 6Stop an Old SwiftUI Search Task from Replacing a New ResultUse task cancellation and a query identity check so a slow response cannot overwrite a newer search.
Rashad Shirizada·Oct 5Ask for Consent Before Sending a Failed On-Device AI Request to the CloudA SwiftUI fallback flow that makes the data boundary explicit instead of silently changing where a prompt is processed.
Rashad Shirizada·Oct 2How to Recover from a Context-Window Error in a SwiftUI ChatA concrete retry flow for LanguageModelError.contextSizeExceeded that keeps the conversation useful instead of silently failing.
Rashad Shirizada·Oct 1Stop Reusing LanguageModelSession for One-Shot SwiftUI TasksUse a fresh session per task — or your “simple” AI feature quietly carries old context forward.
Rashad Shirizada·Sep 30How I Test a Foundation Models Meeting-Summary Feature in SwiftA 20-case suite that catches missing owners, dates, and invented action items.
Rashad Shirizada·Sep 29Don’t Hard-Code One AI Provider in Your iOS App: Use a Backend PolicyA small Swift boundary that lets a feature start on-device, fall back deliberately, and stay testable.
Rashad Shirizada·Sep 28Your Xcode Project File Is Finally Agent-Readable: A Safe Migration to .xcprojMove from conflict-prone project.pbxproj edits to reviewable JSON without breaking CI.
Rashad Shirizada·Sep 11Your On-Device AI Feature Is Probably Untested: A Practical Evaluation Strategy for SwiftIf your AI feature “felt good” in a demo, you have not tested it yet.
InLevel Up CodingbyRashad Shirizada·Sep 9How to Test AI Features in Swift When the Output Is Never the Same TwiceA practical evaluation strategy for reliable Foundation Models features in Swift.