Apple Technical Partner · Since 2017
AI already writes code. Deciding which code to accept, what to fix and why is still your job. That’s what we teach: Swift, SwiftUI and AI agents, with judgment.
import SwiftUI@MainActor @Observablefinal class Classroom { var lessons: [Lesson] = [] func load() async throws { lessons = try await Lesson.all() }} struct ClassroomView: View { @State private var classroom = Classroom() var body: some View { List(classroom.lessons) { lesson in Text(lesson.title) } .onAppear { Task { try? await classroom.load() } } }}Julio commented on lines 16–18
Use .task instead: it cancels itself when the view disappears. And you get rid of the stray Task.
Suggested change
Removed: .onAppear { Task { try? await classroom.load() }}Added: .task { try? await classroom.load() }import SwiftUI final class CourseProgress: ObservableObject { @Published var completed = 0 let total = 124} struct CourseProgressView: View { @StateObject private var progress = CourseProgress() var body: some View { Gauge(value: Double(progress.completed), in: 0...Double(progress.total)) { Text("Class hours") } }}Julio commented on lines 3–4 and 9
With @Observable the view only redraws for what it reads, and @Published goes away. While you’re at it, @StateObject becomes @State.
Suggested change
Removed: final class CourseProgress: ObservableObject { @Published var completed = 0Added: @MainActor @Observablefinal class CourseProgress { var completed = 0Removed: @StateObject private var progress = CourseProgress()Added: @State private var progress = CourseProgress()import Testing struct CourseworkTests { @Test("Remaining hours are computed correctly") func remainingHours() throws { let coursework = Coursework(hours: 124, completed: 100) let remaining = coursework.remaining #expect(remaining != nil) #expect(remaining! == 24) }}Julio commented on lines 7–9
#require stops the test if it fails and hands you the unwrapped value. Goodbye, !.
Suggested change
Removed: let remaining = coursework.remaining#expect(remaining != nil)#expect(remaining! == 24)Added: let remaining = try #require(coursework.remaining)#expect(remaining == 24)import Foundation struct Program: Codable, Sendable { let name: String let seats: Int} func loadCatalog(from url: URL, completion: @escaping @Sendable ([Program]) -> Void) { URLSession.shared.dataTask(with: url) { data, _, _ in let programs = data.flatMap { try? JSONDecoder().decode([Program].self, from: $0) } completion(programs ?? []) }.resume()}Julio commented on lines 8–16
With async throws, the caller finds out when something fails. Right now a network error turns into “there are no programs” and nobody knows.
Suggested change
Removed: func loadCatalog(from url: URL, completion: @escaping @Sendable ([Program]) -> Void) { URLSession.shared.dataTask(with: url) { data, _, _ in let programs = data.flatMap { try? JSONDecoder().decode([Program].self, from: $0) } completion(programs ?? []) }.resume()}Added: func loadCatalog(from url: URL) async throws -> [Program] { let (data, _) = try await URLSession.shared.data(from: url) return try JSONDecoder().decode([Program].self, from: data)}import SwiftUI struct BookingButton: View { let book: () -> Void var body: some View { Button(action: book) { Image(systemName: "calendar.badge.plus") } .buttonStyle(.borderedProminent) }}Julio commented on lines 6–8
A button with no text makes VoiceOver guess. Give it a title and show only the icon: it looks the same and says exactly what it does.
Suggested change
Removed: Button(action: book) { Image(systemName: "calendar.badge.plus")}Added: Button("Book a seat", systemImage: "calendar.badge.plus", action: book).labelStyle(.iconOnly)import Foundationimport FoundationModels struct Summary: Decodable { let ideas: [String]} func summarize(_ lesson: String) async throws -> Summary { let session = LanguageModelSession() let text = try await session.respond( to: "Return a JSON with three ideas: \(lesson)") return try JSONDecoder().decode( Summary.self, from: Data(text.content.utf8))}Julio commented on lines 4–6 and 10–13
With @Generable the on-device model hands you a Summary directly. No JSON to parse, and nothing to break on the first odd answer.
Suggested change
Removed: struct Summary: Decodable { let ideas: [String]}Added: @Generablestruct Summary { @Guide(description: "Key ideas", .count(3)) let ideas: [String]}Removed: let text = try await session.respond( to: "Return a JSON with three ideas: \(lesson)")return try JSONDecoder().decode( Summary.self, from: Data(text.content.utf8))Added: return try await session.respond( to: "The three key ideas of this class: \(lesson)", generating: Summary.self).contentimport SwiftUI struct LiveBadge: View { var body: some View { Label("Live", systemImage: "video.fill") .padding(.horizontal, 16) .padding(.vertical, 10) .background(.ultraThinMaterial, in: .capsule) .shadow(radius: 8) }}Julio commented on lines 8–9
You can’t fake Liquid Glass with a material and a shadow. .glassEffect() reacts to what’s behind it and respects Reduce Transparency.
Suggested change
Removed: .background(.ultraThinMaterial, in: .capsule).shadow(radius: 8)Added: .glassEffect(in: .capsule)import SwiftUIimport SwiftData struct LessonsView: View { @Environment(\.modelContext) private var context @State private var lessons: [Lesson] = [] var body: some View { List(lessons) { Text($0.title) } .task { let byOrder = FetchDescriptor<Lesson>( sortBy: [SortDescriptor(\.order)]) lessons = (try? context.fetch(byOrder)) ?? [] } }}Julio commented on lines 5–6 and 9–13
@Query runs the fetch and keeps it up to date: when a new lesson arrives, the list updates on its own.
Suggested change
Removed: @Environment(\.modelContext) private var context@State private var lessons: [Lesson] = []Added: @Query(sort: \Lesson.order) private var lessons: [Lesson]Removed: .task { let byOrder = FetchDescriptor<Lesson>( sortBy: [SortDescriptor(\.order)]) lessons = (try? context.fetch(byOrder)) ?? []}import SwiftUI struct StartView: View { let start: Date var body: some View { VStack(alignment: .leading) { Text("Next cohort") Text(date(start)) .font(.largeTitle.bold()) } } private func date(_ day: Date) -> String { let formatter = DateFormatter() formatter.dateFormat = "dd/MM/yyyy" return formatter.string(from: day) }}Julio commented on lines 8 and 12–17
A fixed-format DateFormatter ignores the reader’s language and is rebuilt on every render. With format: .dateTime everyone reads “October 5” or “5 de octubre”, in their own format.
Suggested change
Removed: Text(date(start))Added: Text(start, format: .dateTime.day().month(.wide))Removed: private func date(_ day: Date) -> String { let formatter = DateFormatter() formatter.dateFormat = "dd/MM/yyyy" return formatter.string(from: day)}final class Seats { private(set) var available = 40 func book() -> Bool { guard available > 0 else { return false } available -= 1 return true }} func openEnrollment(_ seats: Seats) async { await withTaskGroup(of: Bool.self) { group in for _ in 0..<50 { group.addTask { seats.book() } } }}Julio commented on lines 1 and 13
Swift 6 won’t compile this: 50 tasks touching available at once is a data race. With an actor, each booking waits its turn.
Suggested change
Removed: final class Seats {Added: actor Seats {Removed: group.addTask { seats.book() }Added: group.addTask { await seats.book() }import Foundation @MainActorfinal class Transcript { var words = 0 func analyze(_ text: String) async { words = countWords(text) } private func countWords(_ text: String) -> Int { text.split(whereSeparator: \.isWhitespace).count }}Julio commented on lines 8 and 11
Counting the words of a 4-hour class on the main actor freezes the UI. With @concurrent the work leaves the main thread and the result comes back on its own.
Suggested change
Removed: words = countWords(text)Added: words = await countWords(text)Removed: private func countWords(_ text: String) -> Int {Added: @concurrentprivate func countWords(_ text: String) async -> Int {They’re the ones we’ve seen most since 2017, in individuals and in teams. None of them is a build error: they all have a fix, and we know which one.
Pick the line that sounds most like you. Wherever you’re starting from, you don’t have to figure it out alone.
Everyone starts with a first line. Tell us what you’d like to build and we’ll help you map the way, step by step and at your pace.
What you already know counts, a lot. In one conversation we’ll see what carries over, what changes on Apple platforms and where you should start.
It happens to almost everyone after a few years. We’ll look at where you are together and what you need for the next step.
That’s normal: everything moves fast. The good news is that judgment doesn’t expire. Let’s talk about working with agents without losing control of your code.
Take a look at the team catalog and tell us what your team does. We design the training from there, without pausing your projects.
It’s the most common starting point. Browse the catalog and, in one call, we’ll help you decide what to tackle first.
Waiting for your answer…
None of these sound like you? That’s fine: hardly anyone knows where to start, and that’s what we’re here for. Our team is by your side to help you see it clearly. Talk to us.
We’ve trained teams at Accenture, Deloitte, Telefónica, ING, Carrefour, Indra, NTT Data, Globant, Sngular, Alten, Red Eléctrica and Mercadona.
Enrollment open
For experienced iOS developers: specify, direct and verify the work of AI agents, and bring intelligence inside your app.
View program Swift Agentic ProgramJulio commented on “Swift Agentic Program”
Agentic coding for experienced iOS developers: specify, direct and verify AI agents without losing control of your code.
Cohort in progress · next cohort to be announced
For starting from scratch or rebuilding your foundations, and ending up directing AI agents with judgment.
App for iPhone, iPad and Mac
The academy’s app: courses, mini-courses, tips and snippets across five levels, on iPhone, iPad and Mac.
Four rules that don’t change with every WWDC. Everything else does.
From junior to senior, on your platform or ours. We start from what your team already does.
Academic leadership
Academic director and founder
More than 30 years programming and more than 10 as an expert Swift consultant. Author of Aprendiendo Swift, a visionOS pioneer and a public voice for the Apple ecosystem.
“Programming is a form of communication that teaches us to think.”
I’ve been programming for more than 30 years and I still learn something at every WWDC. Here you won’t copy code: you’ll understand it. And when AI comes in, you’ll know what to ask of it and what not to accept.
— Julio
Julio commented on “2026”
We review the programs after every WWDC. Today: Swift 6.4 and Xcode 27.
“From not knowing how to build an app to working with REST APIs and concurrency. I started from zero.”
“It teaches you to think like a developer, always starting from judgment.”
“I learned things I didn’t know I didn’t know, and filled in the gaps I had.”
From FORTRAN to Xcode 27: every abstraction took a task away from programmers, none took the craft. Why this time is different and what we did with the SEP.
Liquid Glass and the Foundation Models framework: the two WWDC 2025 announcements that redefine Apple development. Analysis and must-watch sessions.
Two-thirds of the apps rejected from the App Store fail on performance. What quality means on iOS and how to build it in from day one.
The next step
One hour with us to put things in order: where you’re coming from, where you want to go and the shortest path between them. You take it away in writing.
A written route recommended for your case. Even if you follow it on your own.
Jon Ander Orcera · Project Manager at the academy
