The Modern Craft of AI-Assisted iOS Development: Metrics, Methods, and Reality

When our team set out to build a production iOS application from scratch, we made an unconventional choice: embrace AI coding assistants as core development tools rather than occasional helpers. The result? A shipping application where 90% of the code was AI-generated, with measurable improvements in development velocity and surprising insights about what "writing code" means in 2026. This talk shares our real-world experience incorporating terminal-based AI agents and Model Context Protocol (MCP) integrations into daily iOS development workflows. You'll see concrete metrics from our Git history showing commit frequency, PR throughput, and feature delivery timelines compared to traditional development. More importantly, you'll learn what worked, what spectacularly failed, and how we maintained code quality and architectural integrity while letting AI generate the majority of our implementation. We'll explore practical applications including: automated code style enforcement, intelligent task decomposition with external system integration, structural research for complex feature requirements, automated PR templates and review comments, and high-level architectural decision support. The focus is on staying in control - using AI to amplify your craft, not replace your judgment. Whether you're skeptical about AI coding tools or already experimenting, you'll leave with actionable strategies for incorporating these technologies into Swift development while maintaining the professional standards that define quality software.