"Write some code" isn't a strategy. Most developers experimenting with agentic AI have no consistent framework for deciding what to automate, what to review closely, and what to keep entirely hands-on. This leads to either over-reliance on AI-generated code or underuse of a genuinely useful tool. This toolkit gives you that framework!
What You'll Need
- Mac running macOS 26.3 or later
- Xcode 26.3 or later (agentic coding features)
- Apple's sample project: Landmarks: Building an app with Liquid Glass (SwiftUI)
What's Inside the Toolkit
- Pre-Built Lesson Plan: A ready-to-run, hour-long session plan with an introduction, hands-on activity, and reflection, built around Apple's Landmarks sample app.
- 3 Structured Learning Phases: A step-by-step facilitation framework, from a 15-minute framing discussion to a 30-minute hands-on build in Xcode, designed for classroom teaching or team upskilling.
- Human-in-the-Loop Design Framework: Guided prompts and benchmarks to evaluate what AI should automate (boilerplate, repetitive logic) versus what must remain developer-driven (architecture decisions, final review, ethical judgment).
How It Works
- Set Up the Activity: Download Apple's Landmarks sample app, build it, and choose a feature to implement, such as adding imperial units, a Favorites list, or missing action buttons.
- Automate the First Draft: Let Xcode's agentic coding tools generate the code from a natural-language prompt.
- Review & Personalize: Apply the human touch: critique the agent's changes, iterate on the prompt, add documentation and tests, and decide what still needs a developer's judgment.
Designed for Real Classrooms (and Dev Teams)
- Judgment-First: Built around a repeatable framework for deciding what AI should and shouldn't touch, not just a coding tutorial.
- Completely Customizable: Swap in any feature idea or sample app, and scale the activity to fit your session length.
- Low Barrier to Entry: Uses standard Xcode agentic coding features: Xcode familiarity is a plus, but not required.
Discussion
- Do AI agents solve for self-efficacy?
- What security measure should an engineer consider before automating a task?
- How do you assess understanding when an agent wrote the first draft?
Credits
- Teaching kit developed by Azhar Amir Kimanje, AVELA
- Post by Talile Geloto, AVELA
Attach up to 5 files which will be available for other members to download.