I created a whole macOS app with Flutter just using AI, here is my report

Building My First MacOS App: 100% AI-Created Render Manager

For the first time ever, I ventured into creating an app for MacOS, and the craziest part? It was entirely built by AI. I didn't write a single line of code myself. Of course, my background in software development was invaluable in guiding the process, but it’s absolutely insane what’s possible with AI today.

Why a Render Manager?

Rendering complex 3D scenes in Blender is a demanding process, and existing render managers often lack the specific features or flexibility needed for personal workflows. As a Blender user on MacOS, I struggled to find a solution that integrated seamlessly with my tools, offered customizability, and streamlined rendering tasks. This app was born out of necessity, and its development highlights both the capabilities and limitations of AI-assisted programming.

About the App

The app is a render manager specifically made for MacOS. Why? Because I couldn’t find an existing solution that met my needs. Here’s what it does so far:

• 📦 Drag and Drop: Easily add Blender files to the render queue.
• 🔐 Blender Addon: Integrates directly with Blender’s render menu for seamless file addition.
• 🔅 Configurable Settings: Full control over render configurations.
• ⏹ Mid-Process Stop: Ability to halt rendering at any point.
• 🔧 Custom Output Paths: Define output paths and naming patterns.
• 🔄 Auto Retry: Retries failed renders automatically with customizable attempts.
• 🌐 Network Rendering: Distribute workloads across multiple machines.
• 🚀 TurboRender Support: Compatibility with advanced features like Temporal Denoising.
• 📱 Telegram Integration: Remote monitoring and control via Telegram.
• ⚙️ Command-Line Parameters: Support for custom Blender CLI commands.
• 🏠 Factory Settings: A troubleshooting mode for resetting configurations.
• 🎯 Scene Selection: Select scenes and cameras per project.
• 📁 Organized Outputs: Flexible file organization using pattern variables.
• 🔍 Real-Time Monitoring: Track render progress live.
• 📊 Detailed Logs: Comprehensive render logs and error reporting.
• 🎮 Frame Control: Define start, end, and step values for frame ranges.
• 🎨 Multi-Scene Support: Manage individual settings for multiple scenes.
• 🔄 Automatic Data Fetching: Seamlessly retrieves scene data.
• 📸 Camera Selection: Choose cameras for specific scenes.

Building with AI: Insights and Tips

Patience Is Key

Working with AI requires a lot of patience. It’s also not cheap—each request incurs a cost, and the AI often makes errors. Without version control like Git, the process would be chaotic. Sometimes, the AI replaces sections of code with placeholder comments, so reverting changes is critical. Using Git or similar tools is non-negotiable when relying heavily on AI-generated code.

Splitting Tasks

When working with AI, break down your tasks into small, manageable chunks. This minimizes misunderstandings and helps the AI focus on specific objectives. Restarting chats frequently to maintain instruction clarity also helps.

Debug Logs Are Crucial

It’s essential to let AI create as many debug logs as possible. You’ll often find yourself copying logs into the AI to provide context for what went wrong. This iterative debugging process is vital for resolving issues efficiently.

Where AI Excels

• Refactoring: AI handles refactoring impressively well, turning messy code into clean, maintainable structures.
• File Splitting: Breaking down large files into smaller ones is essential for managing AI costs and reducing errors.
• Prototyping: Quickly generating boilerplate code or testing ideas is where AI shines.

Challenges

• Instruction Fatigue: The longer a chat continues, the more the AI seems to ignore custom instructions. Restarting the chat frequently and focusing on small, manageable tasks helps.
• Test-Driven Development (TDD): TDD didn’t work as smoothly as expected. However, when this improves, coding with AI will likely reach a whole new level, enabling faster iteration with automated unit tests and quick fixes.

Cost Management

AI tools incur costs for every request. To keep expenses under control:

1. Plan Ahead: Draft clear, concise instructions before interacting with the AI.
2. Optimize Code: Use modular design to reduce the complexity of individual AI tasks.
3. Test Locally: Minimize back-and-forth iterations by testing generated code locally.

Moving Forward

If you’re a Blender user on MacOS, this app might be interesting for you. Since it’s built with Flutter, I plan to port it to Windows and Linux soon. Expanding the app’s reach will involve addressing cross-platform nuances and optimizing performance.

Tools and Setup

I utilized VS Code alongside Cline and Claude for this project. These tools provided a collaborative environment for AI-assisted coding, although their performance varied. Integrating a robust IDE setup and automation tools can significantly streamline AI development workflows.

Final Thoughts

Despite its challenges, this experience has been incredibly rewarding. AI-driven development has its hiccups, but the speed and efficiency of creating an entire app are unparalleled. If you’re considering using AI for your projects, I’d say give it a shot—just remember to bring patience and a solid backup system!

Let me know if you’re interested in trying the app or if you have any feedback. The future of coding is here, and it’s looking pretty exciting.