An international primary school teacher with 11 years of experience has developed a custom, AI-powered reading recommendation app for their classroom using "vibe coding"—the practice of generating software by describing requirements in plain language to artificial intelligence. Initiated in May 2025 and launched in November 2025, the application matches students with books from their school’s existing 10,000-book library catalog. The teacher utilized tools like Microsoft Copilot and Anthropic's Claude on a 12-year-old Mac to construct the system, which ingests library catalogs via CSV files and analyzes them against student profiles containing reading ages, interests, and curriculum topics.

The resulting application generates a personalized list of approximately 50 ranked book recommendations for each student. To protect privacy, student profiles and progress data’including books read, reviews, points earned, and comprehension quiz scores’are restricted to teachers and school librarians. Building the system required overcoming technical challenges, such as managing API calls for book cover images and recovering from lost files, with the developer noting that switching from Copilot to Claude significantly improved development reliability.

This project highlights a shifting paradigm in educational technology, where educators without formal software engineering backgrounds can bypass expensive, rigid proprietary software by building bespoke tools. The rapid evolution of LLMs between late 2025 and mid-2026 underscores how natural-language programming is lowering the technical barriers to software creation. For schools, this suggests a future where software can be highly localized and tailored to specific classroom curricula without requiring massive IT budgets.

For South African schools facing tight budgets and diverse reading levels, this approach demonstrates how teachers might eventually bypass expensive software licenses to build localized learning aids aligned with the national curriculum.