Those Who Swift - Issue 254
Weekly note ✏️The hiring process has become complicated. Even the biggest tech companies struggle with it, and candidates often need to take unusual steps to stand out. Sometimes that means building something memorable (like a OpenClaw). Peter Steinberger has joined the OpenAI team to push the limits of AI agents, and according to his blog post, OpenClaw will remain open source. His next goal is to make the agent as accessible to users as possible. At the same time, interviews are far from easy. You tailor your CV with AI to match a role, then an ATS and another AI filter it on the other side. During technical interviews you are either banned from using AI tools or heavily restricted. And then you read news like Spotify stating that much of its code is now written by LLMs. It’s an interesting decision, though hopefully more marketing than reality: without careful review, such tools would produce unpredictable and poorly managed code. In practice, developers still need to review, refine, and document everything until it becomes a maintainable codebase. Maybe the lesson is to look at job searching from a different perspective. As Stan Shelipov suggests, instead of chasing individual vacancies, try finding the ecosystem where you want to belong. So he made a list of iOS Companies to track. Connect with the "Those Who Swift" team - Justas Markus & Anton Gubarenko 👋 Swift Around the Web 🌐Tracking Token Usage in Foundation ModelsArtem Novichkov shows how to monitor token usage when working with Apple’s Foundation Models, helping you understand prompt and response costs, debug unexpected behavior, and optimize model interactions for performance and efficiency. Copy-on-Write in Swift ExplainedSagar Unagar breaks down Swift’s copy-on-write behavior, showing how value types like Array and String avoid unnecessary copies by sharing storage until a mutation occurs. The article clarifies performance implications and why structs can remain both safe and efficient in real apps. Adopting AI as an Indie DeveloperIn this issue, Antoine van der Lee shares how indie developers can embrace agentic coding and AI tools to dramatically boost productivity. He discusses reducing context switching, giving agents more context, automating repetitive tasks, and building personal AI workflows that act like extra teammates Coding 👨💻Little SwiftUI Tip: Interact With the App StoreThis quick tip by Itsuki shows how to open and interact with App Store pages directly from SwiftUI: using system APIs to present app listings or reviews in-app, improving user flow without forcing a full context switch. Building a Toast Component in SwiftUIThis nice component by Artem Mirzabekian shows how to create a reusable toast notification in SwiftUI, handling presentation, auto-dismiss timing, and animations. He demonstrates a clean overlay approach so you can display temporary feedback messages without disrupting your app’s layout. Apple 🍏Get Ready with the Latest Beta ReleasesApple has released new developer betas for iOS 26.4, iPadOS 26.4, macOS 26.4, tvOS 26.4, visionOS 26.4, and watchOS 26.4. Developers should test their apps and build with Xcode 26.4 beta to ensure compatibility with the updated SDKs. Other cool stuff 🧰If You’re Not Versioning Your SwiftData Schema, You’re GamblingMohammad Azam explains why SwiftData schema versioning is essential the moment your app stores real user data. Using VersionedSchema, SchemaMigrationPlan, and custom migration stages lets you safely evolve models, reshape existing records, and add constraints without crashes: turning data migrations into a deliberate part of your app’s architecture. SwiftUI Foundations: Build Great Apps Q&AAnton Gubarenko walks through key SwiftUI fundamentals Q&A with best practices, focusing on architecture, state management, and performance to help developers create scalable, maintainable apps from the start. Isolate SwiftUI Animations to Specific AttributesNatalia Panferova explains how to limit SwiftUI animations to only the properties you intend, preventing unrelated layout changes from animating. By scoping animations to specific values, you keep transitions smooth and predictable while avoiding visual glitches. AI 🤖Anthropic Learning PortalAnthropic Skilljar is the official training platform offering structured courses and tutorials on using Claude and building with Anthropic’s AI tools: covering prompting, safety concepts, and practical developer workflows for getting started quickly. OpenClaw, OpenAI and the futurePeter Steinberger joins OpenAI as his tool hitting the likes. OpenClaw, an open-source tool designed to improve development workflows by automating repetitive project tasks and integrating with modern AI-assisted tooling. Git Worktrees + Agentic AI — Stop Single-Threading Your BrainWesley Matlock explains how combining Git worktrees with AI coding agents enables parallel development workflows instead of one-task-at-a-time iteration. By splitting work into clearly bounded lanes and orchestrating multiple agents, developers shift from typing code to supervising execution: making architecture, decomposition, and scope control the new core skills. Sentry Acquires XcodeBuildMCPSentry has acquired XcodeBuildMCP, a tool that helps AI coding agents interact with Xcode builds and diagnostics. The move aims to improve debugging workflows by letting AI tools understand compiler errors and project state more directly, bringing smarter automation to iOS development. Tutorials 📒How to Migrate to @Observable Without Breaking Your AppThis guide walks through safely moving from Taking First Steps Into Metal ShadersThis beginner-friendly guide introduces Metal shaders in Swift, explaining how vertex and fragment shaders work and how to connect them to SwiftUI rendering. It walks through setting up a Metal pipeline and rendering custom graphics so you can start adding GPU-powered effects to your app. Video 🎥Sharing SwiftData Content Between UsersPart two of this series shows how to export and import SwiftData data between app users. It covers presenting a UIKit share sheet from SwiftUI and handling incoming files with Yet, another thing…🤖🧃💵Opus 4.6 Vending BenchAndon Labs benchmarks Claude Opus 4.6 on real coding tasks using their “vending machine” test suite, comparing reliability, autonomy, and error recovery against other models. The results highlight improvements in long-running agent workflows and show how modern LLMs are moving from code assistants toward autonomous engineering collaborators. Thanks for reading Those Who Swift! Subscribe for free to receive new posts.
|
