The developer has completely overhauled a Morse code application originally written in Objective-C. The project, which had not been updated for over five years, was migrated to Swift and SwiftUI with the assistance of AI models Claude and Codex.
The application now supports 60 languages and includes 37 different Morse code tables, including international, Japanese, Chinese, Indian, and Thai variations. The integration allows for input from telegraph keys via adapters.
A new administrative panel was implemented to manage customer feedback, newsletters, and auctions. It features an integrated translation system that enables real-time communication between support staff and users in different languages.
The training module was updated to include modern methodologies, such as the Koch method, and a testing system modeled after professional Morse code competitions. The app now tracks user progress and identifies specific characters that require more practice.
The developer trained a neural network on a MacBook Air M3 to decode Morse code from noisy radio environments. The model is designed to process up to 16 simultaneous stations in a 2.5–3 kHz band, supporting speeds from 5 to 100 WPM.
To train the model, the developer created a custom dataset generator that simulates real-world radio interference, such as atmospheric noise, equipment artifacts, and signal fading. This allows users to practice under conditions mimicking actual radio broadcasts.
