Freshmen Build Four AI Apps in Two Hours with Huawei AI Shell and openJiuwen

Makerizon organized an AI hackathon for freshman students with almost no project experience, challenging them to build working applications around real needs within two hours using AI development tools. All four projects completed successfully and are now open-sourced.

Projects Built

PostureGuard – Intelligent posture monitoring system using visual recognition; detects posture, alerts on forward lean, and generates AI adjustment suggestions.

Drowsy Detection – Fatigue recognition via facial feature analysis; monitors eye keypoints and closed-eye duration to trigger fatigue warnings.

Four-Quadrant Task Kanban – Task management tool based on the Eisenhower matrix; supports task categorization, priority management, and drag-and-drop sorting.

Water & Stretch Reminder – Health reminder app fused with LLM capabilities; provides timed reminders for drinking water, standing up, and AI-generated copywriting.

Toolchain and Division of Labor

The development relied on three Huawei tools working in concert:

AI Shell – Pre-installed runtime environments, dependencies, and mainstream large models. Students described requirements in natural language, executed commands, and started services without configuring local environments from scratch.

CodeArts – Provided AI-assisted coding capabilities to implement application logic.

openJiuwen – Huawei's open-source agent development framework (SDK and related capabilities). It handled prompt orchestration, output parsing, and integration of LLM abilities into concrete business scenarios (e.g., generating posture advice after forward lean, triggering reminders after prolonged eye closure, producing health reminder copy). Developers configured agents around specific tasks instead of building an entire agent system from zero.

The article emphasizes that the first hurdle for beginners is often environment setup, dependency installation, and version conflicts—not writing code. AI Shell removes that barrier, CodeArts accelerates coding, and openJiuwen bridges LLM capabilities into product features.

Outcome and Resources

All four prototypes ran successfully within the two-hour window. The projects serve as reference implementations for students and makers learning AI application development. Source code is available at:

https://gitcode.com/MakerStudio/Maker-jiuwen

The event demonstrates that with mature AI development tools, beginners can complete the first step from idea to runnable application much faster, though the prototypes are not yet production-grade and the exercise does not equate to mastering software engineering in two hours.