The use of artificial intelligence in software development is demonstrating an increased capacity to accelerate application creation and porting, with developers leveraging both commercial and open-source AI tools.
While these tools offer significant productivity gains, they require active management, iterative refinement, and developers' foundational understanding of coding principles to address challenges such as user interface design, performance optimization, and AI behavior.
AI in Commercial and Proprietary Development Environments
Claude Code Assists Mac Application Porting
One developer utilized Claude Code, an AI coding tool, to port an existing iPhone application to a Mac. The original iPhone app, built with Swift and SwiftUI in Xcode, manages 3D printing filament spools, tracking 120 spools via NFC tags and camera images. The objective for the Mac version was to monitor material availability from the desktop, integrating with existing print job initiation workflows.
The porting process involved branching the Git repository and instructing Claude Code to develop the Mac application, omitting NFC and camera functionalities but including iCloud synchronization. The initial AI-generated Mac app was functional within ten minutes. However, it exhibited poor user interface design and performance issues, notably slow loading of the spool list.
Development proceeded through iterative cycles of building, identifying Xcode errors, and providing feedback to Claude Code. Key challenges and their resolutions included:
- UI Scaling: Ensuring UI elements scaled correctly with resizable Mac windows, addressed by establishing a knowledge base within Claude Code for responsive behavior.
- Interface Adaptation: The Mac application adopted a three-pane interface, distinct from the iPhone's screen-to-screen transitions, requiring extensive prompting.
- Cross-Platform Conflict: An issue where the Mac's three-pane implementation inadvertently broke the iPhone app was resolved by instructing the AI to create separate interface screens for each platform.
- Performance Optimization: Slow loading of spool lists due to full-resolution photos was mitigated by implementing a thumbnail caching system, generating and storing 128-pixel versions in the background, which also necessitated features for list refreshing and cache purging.
The developer characterized the experience as primarily "managing" the AI, akin to overseeing a remote human programmer, involving assignments, questions, error identification, and extensive testing.
Despite occasional "obtuse and stubborn" behavior from the AI, a functional Mac application was developed in approximately 8 hours of active work over one month, described as significantly faster than manual coding.
Xcode 26.3 AI Enhances iOS App Development
Another developer reported enhanced capabilities in Apple's Xcode 26.3 AI for creating a new iOS application. This contrasts with previous experiences using Xcode 26.1 AI, which was deemed largely unusable for complex tasks. The new project involved a sewing pattern manager application designed for iOS, incorporating features such as photography, NFC tag tracking, and machine learning for image processing. The AI was utilized for tasks like identifying pattern envelopes, straightening images, cropping, extracting vendor names and pattern numbers via OCR, and distinguishing pattern numbers from barcodes.
The development process, which involved migrating an existing codebase and adding new features, resulted in the insertion of 52,947 new lines of code and the deletion of 10,626 lines across 689 files in less than two days of partial work. This work included integrating Apple's latest AI/ML libraries for image analysis.
It was estimated that this level of work would have taken a lone developer four to six months without AI assistance.
Challenges encountered with Xcode 26.3's integrated AI assistant included system hangs, lack of visibility into changes, and excessive token consumption due to its use of multiple background agents.
A resolution was achieved by instructing the AI: "Do NOT use background agents or background tasks. Do NOT split into multiple agents. Update me regularly on each step. Do NOT run steps that take more than a minute or two without having an update heartbeat." This rule reportedly improved productivity and stability.
Usability features of Xcode's AI assistant include image pasting, automated code builds and error fixing, and integration with voice dictation tools for hands-free prompting. Apple is anticipated to release the full version of Xcode 26.3 on the Mac App Store in the near future.
Local and Open-Source AI Coding Solutions
Goose and Qwen3-coder for Local AI Development
The establishment of free, local AI coding agents using Goose, an agent framework, and Qwen3-coder, a coding-centric large language model, has been explored as a potential alternative to paid AI coding plans. This setup requires a powerful local machine. Jack Dorsey has been noted for endorsing this combination.
The setup process involves:
- Ollama: Installing the Ollama application, downloading the
Qwen3-coder:30bmodel (approximately 17GB) within its interface, enabling "Expose Ollama to the network," and configuring context length. Ollama must remain active during use. - Goose: Installing the Goose application, navigating to "Other Providers," selecting "Go to Provider Settings," and configuring Ollama as a provider by selecting
qwen3-coder:30bas the model.
Initial testing, which involved prompting Goose/Qwen3 to develop a simple WordPress plugin, required five attempts for successful completion, with the first three attempts failing to produce a working plugin or follow directions entirely. Agentic coding tools like Goose, however, allow for iterative corrections.
Performance varied by hardware; an M1 Mac with 16GB of RAM experienced significant performance issues, while an M4 Max Mac Studio with 128GB of RAM reported good overall performance, with prompt turnaround times comparable to cloud-based or hybrid AI coding solutions for simple tasks. Further analysis with larger projects is suggested to assess their full potential as replacements for paid alternatives.
General Observations and Recommendations
Across various AI coding tools, a common theme is the shift from direct coding to a management role, where developers guide, correct, and refine AI-generated code.
This iterative process, involving continuous refinement and rigorous testing, is deemed crucial due to the potential for AI to omit requested features or introduce issues. While these tools offer substantial productivity benefits, allowing for significantly faster development cycles, they require developers to possess proficiency in coding and an understanding of underlying technology for effective guidance on architectural and design choices.
New programmers are advised to manage expectations regarding immediate market-ready application development.