AI Coding Experience: What Can Non-Engineers Do with AI Coding?

My Programming Level? 🤔

It's probably not zero, because I've self-hosted websites before, so I can understand basic HTML and CSS, and can make some modifications to UI code. I also took a basic Python course a while ago and learned a few concepts.

That's my background, but honestly, throughout the AI coding process, I mostly relied on AI to complete things. When I encountered problems, I'd throw them to AI first.

The biggest gain from this process was learning how to ask questions, follow up, and solve problems. Figma Plugins are what I've written the most. Although AI does the writing, because I've done it many times, I've gotten faster. I now know how to handle problems when they arise. While my programming ability might not have changed, my "AI coding" ability has improved significantly - something I didn't expect at the beginning.

Below I'll share several tools I've written, explaining the reasons and some insights.

▍PDF File Splitting Tool 📄

I wanted to upload a PDF file to NotebookLM, but the file exceeded 200MB and couldn't be uploaded. I first Googled PDF splitting tools and found a few, only to discover that these tools require payment for files over 100MB or 200MB.

My old solution would be to pay, give up, or slowly search for free alternatives. But now my solution was to open Cursor and spend about half an hour making one myself. The result was a personal PDF splitting tool that perfectly solved my problem!

Insights from this:

1. AI is very convenient for writing single-function tools like this, with low failure rates. In the future, these paid tools for small needs might be directly or indirectly replaced by AI coding services. I also discovered these tools aren't as complex as imagined. In the past, due to technical barriers, I could only pay for solutions. Now there are different approaches.
2. The problem I wanted to solve already had existing tools, so I knew I could try having AI write a tool to help. If you want to solve a problem and don't see related tools, it's either too niche or too difficult.

▍Face Cropping Tool + File Naming Tool 🖼️

I wanted to use an AI generation tool to train my personal model, but the model needed my photos as training data. The requirement was "face" photos - other parts needed to be cropped out for better results. Cropping photos is simple for me, but when you need to crop 20 photos, it becomes annoying.

I also chose Cursor to help me here. I wrote a tool that reads photos from a specified folder, crops out faces, and exports them to another folder. The images needed a specific "filename format," so I also wrote a tool to help me rename them. None of this took much time, but now I can use these tools whenever I have similar needs.

Insights from this:

1. I discovered that the benefit of programming is being able to solve these problems in the shortest time in the future. One-time creation produces long-term benefits. Programming solves the "repetitive labor" problem.
2. The AI applications mentioned here aren't just coding but also AI image generation. We can see that the help and impact of these AI tools reinforcing each other is really significant.

▍Countdown Timer Tool ⏳

My programming inspiration usually comes from wanting to solve a problem. This tool was actually because I saw countdown timers on crowdfunding pages that looked ugly. I searched online and found most free versions have watermarks. I thought this tool didn't seem complex, so I decided to make one.

But I got stuck for a long time because this tool needed to be embeddable in Notion and website pages. I didn't actually understand embedding and spent a lot of time on it. Later I used a tool called "bolt.new" to write it because it can publish online, which finally let me complete it successfully.

Insights from this:

1. I still don't really understand embedding, only knowing it might be because this tool can't work "offline." But I wasn't sure, so I tried switching to a tool that could help publish online, and it finally worked.
2. From this I can tell that if an engineer did this, it would be super fast. In the past, they could write it, but since it still takes some time, the benefit wasn't great for them. Now with AI, non-engineers can also make such tools and improve their efficiency.

▍Google Apps Script 📊

This is probably the most recommended "AI coding" application method for beginners because it's easiest to get started with and can solve many problems. I tried using it to write some auto-execution features for Sheets.

For example, when I needed to create a post-course survey recently, I first gave the course content to AI and asked it to help me think of questions. After the questions were ready, I gave them to AI again and asked it to generate the survey (thanks to the Generative AI Annual Conference gift pack). This let me complete the survey creation super fast, saving tons of time.

Insights from this:

1. The benefit of writing Google Apps Script is that Google already has many useful tools and features. The programs we write can automate these tools, producing huge benefits. It's better to directly leverage existing platform resources rather than trying to write Google Docs, Sheets, Gmail yourself - that's too difficult and unrealistic.
2. Any small task can be automated. Even if it originally only took you one minute, if it can become one-click execution, the accumulated time savings can be considerable. Most importantly, the "mental energy" saved is even more valuable.

▍Figma Plugin 🧩

This is what I've written the most. I now have 3 tools officially published on Figma that users worldwide can use. This is something I never dared to imagine before, and honestly, I've gained a great sense of achievement.

Although I encountered many frustrations along the way, compared to the gains, these failures are actually fine. Before AI, I couldn't write a single one, so every success feels like a bonus. I can only say this feeling is amazing.

Insights from this:

1. Maybe because I'm not an engineer, the "every success is a bonus" mindset actually removes pressure. I dare to try anything and am not afraid of failure.
2. I've written the most Figma Plugins because Figma is currently my main design tool. I was testing whether I could empower my design work through programming, solve my problems, and make workflows smoother. It proved possible. A designer being able to write programs for themselves is really awesome. I see more possibilities, and because of this, I especially like Figma - I see more flexibility and potential.

My published Figma tools:

• LAYER RENAMING: Streamline Your Workflow With Effortless Layer Management.
• IG MULTI IMAGE SPLITTER: Seamlessly Create And Split Instagram Carousel Posts.
• MULTIDESIGN ASSISTANT: Swap Colors, Adjust Sizes, Scale Proportionally, Add Circular Text Backgrounds.

You can find these tools by searching for the English titles in the image. Feel free to try them out. I'm continuously improving and hope to release more powerful and practical tools in the future to help myself and more people.

Finally, I want to say that Figma is very useful for AI coding because it lets me create plugins with more beautiful interfaces. AI makes me feel that knowing design is awesome - it helps me in so many ways.

This is my AI coding experience. Just a few months, but it showed me so many possibilities. Because of this skill, I now see many things and problems from a very different angle.

I think about whether this work or problem can be solved with programming. Because of AI, I also have the opportunity to solve problems through programming. I feel this is just the beginning. The future will definitely be more convenient, and we'll all be able to easily solve our own problems by creating various tools.

Thinking about such a future arriving one day makes me excited and expectant.