How I Built an E-commerce Toolkit with Zero Coding Experience - Dev Log #001

🚀 Project Introduction

Let me formally introduce my new project for the first time: www.selofy.com

The #1 community-driven toolkit for e-commerce SMBs & solopreneurs.

Dedicated to providing comprehensive services for global DTC independent website small and medium e-commerce sellers and related practitioners:

• 🏷️ Brand References
• 🔌 Plugin/Theme Recommendations
• 🤝 Service Partner Recommendations
• 📚 Tutorial Sharing

Simply put: Help users find competitors, find plugins, find services, find tutorials. These are four pain points that independent website players encounter throughout their entire lifecycle, especially newcomers in the pioneering stage.

💡 Project Background

Around the end of 2024, I began to have the idea of independently developing some application tool-type "projects" - web applications, Coze workflows, Shopify apps, etc. I tried each one superficially, and without exception, each attempt ended in failure.

Personally, I actually have many paths to choose from, such as continuing to delve into knowledge monetization and community building, self-media, service providers, cross-border e-commerce, etc. But I still chose to light up a new skill tree among various wandering attempts.

🎯 Three Reasons for Choosing This Path

1. Personal Character Factor

I'm that typical three-minute enthusiasm player. Once I figure something out, my interest inevitably drops to freezing point within a few months. What others call "the best startup method is continuous replication" - I can't do it at all. So while maintaining some basic income, I absolutely love challenging new things.

2. External Opportunity Assessment

The best way to catch the AI wave is to practice AI deeply. On the other hand, without AI, developing a project of this scale alone would be almost impossible. There's so much about AI I want to share with everyone - more on that later.

3. Business Model Prospects

If everything goes smoothly, it fits my ideal project that can create passive income and long-tail value for me. With continuous comprehensive operation and iteration, this project can generate revenue through affiliate programs, advertisements, membership, etc. Of course, the premise is that everything goes smoothly and I can withstand setbacks and invest patiently.

🎯 Project Positioning

Overall, this project is the milestone starting point for lighting up multiple new skill trees:

• "Half-baked Independent Developer"
• "Full English Self-Media Operations"

But I will still focus on the independent website cross-border e-commerce industry. Future self-media tutorial content, community operations, etc., will all have a more sustainably growing base.

🛠️ Tech Stack Details

Core Philosophy

99% almost no-code development throughout. Since there are still many features that haven't been iterated, currently I mainly used the following:

Technical Architecture

ComponentChoiceNotes

‍CMS Frontend SaaSWebflowNo-code website building platform

‍ServerHostinger KVM4 cores/16GB RAM/200GB storage

‍UI/UX TemplatesFlowbasePaid template subscription

‍CMS DatabaseNocoDB (open-source self-hosted)Data management

‍AutomationN8N (open-source self-hosted)Workflow automation

‍Main AIClaude/Claude CodeDevelopment assistant

‍LLM API CallsZhipu GLM, Silicon Flow (Deepseek)Why these two? Because there are discounts!（for Chinese users)

‍Data SourcesNative crawling + AI processingApify calling SimilarWeb and other trusted third-party data

‍Other ToolsDokploy, Crawl4AI, AnyCrawl, Browserless, etc. (open-source self-hosted)-

💰 Investment Cost

Time Cost: Time is the biggest cost. As a newbie with no no-code experience, almost every small step got stuck for a long time. Therefore, the entire learning and practice process took at least 4 months.

Money Cost: Mainly SaaS, server, Flowbase, and LLM API token fees, which were relatively modest.

🏗️ Development Process Details

Overall, I was learning while practicing, without any systematic 0-1 learning process.

Phase 1: Frontend Development (First 2 months)

Core Work: Webflow learning and practice period

Although Webflow is already an excellent no-code website building platform, it was particularly challenging for me with no frontend knowledge.

Main Process:

1. Searched the entire web for Webflow template frameworks that met my needs and related JS (features like filter, chart, etc. required finding specialized solutions)
2. Used Flowbase paid template subscription
3. Claude helped me generate demos
4. In my completely chaotic initial state, I paid freelancers to help me design in Figma, then copied and pasted to Webflow
5. As I self-learned and time progressed, I eventually almost completely rebuilt everything myself

Phase 2: Backend Data Content

After the frontend was almost done, I confidently thought that content-wise, with AI assistance, it should be easy to handle. Wrong.

Challenges:

• Data fields were extremely large - each brand alone has over 80 fields
• Involved many dimensional capabilities: server knowledge, basic coding languages, etc.
• Continuously tried various solutions, from Feishu multi-dimensional tables, Supabase, etc., before finally settling on the current system

Core Work: Researching crawlers, N8N, NocoDB tables, etc.

‍

Phase 3: Operations and Traffic Generation

Just set up overseas social media, will mainly focus on deep operational traffic generation through SEO, self-media, and other forms later.

Phase 4: Future Plans

Still have a large number of website features and content undone:

• CRM membership system
• More content fields valuable to users (but might be troublesome to process)

💭 Deep Sharing

😰 Challenges

The core challenge is still psychological. For anyone, psychological devastation is always the most deadly:

• Giving up due to suffering
• Giving up due to questioning value
• Giving up due to laziness

However, frankly speaking, it went much smoother than I expected. I thought I was a person who fishes for three days and dries nets for two days, someone who retreats when facing difficulties, but this time I was unusually determined.

Sources of Determination:

1. Previous multiple attempts at new directions "working out" made me feel I can do anything
2. I have very firm values: I don't care about short-term benefits at all. Multiple practice results tell me that as long as you can truly create value, someone will pay for your value someday

Solving Technical Challenges:

Indeed, sometimes getting stuck for a day or two can be devastating, but as long as you have specific problems, there are always concrete solutions:

• Pay to solve it
• Use AI to solve it
• Learn more deeply to solve it

There's no problem you can't solve.

🎉 Joy Points

Actually, throughout the entire process, I was in a state of extreme excitement. I felt everything was so fresh and fun.

Addiction Level:

• For about two months, I worked until 2-3 AM every day, sitting motionless
• Even more "addictive" than playing games as a child
• When traveling, I would bring my laptop
• While watching Fuji Rock in Japan, I was thinking about data solutions in my head
• Even at hotels, I would quickly find places to start researching
• Now in Sanya accompanying my parents, I still find a café every day to research

🤖 For AI Coders

💰 More Opportunities

As I shared when doing independent websites, whenever you deeply practice an industry and self-loop a project from 0-1, you can discover countless opportunities to monetize during the process:

• Developing Apify data interfaces
• Designing N8N workflows (selling C-end templates or providing B-end customization)
• Any small problem you solve in the process can be packaged into an application

Small Example: The day before yesterday, to clean up my X account, I wanted to delete all previously posted tweets. But almost all solutions online cost around $10 per month. But if you know a bit of tech, you can create a script yourself. I knew nothing, had Claude make it for me in 10 seconds, and it works amazingly well for personal use.

💪 Effort Behind Opportunities

Of course, I also know many coders will pessimistically mock: "Development isn't hard. But how to get traffic and make people pay is hard."

Nonsense - if this were easy, millions of programmers wouldn't need to go to work, wouldn't fear unemployment, and could all just do independent development.

But don't rush:

• After developing, did you really do SEO, write content, share on self-media?
• Did you really persist for more than 3 months?
• You probably don't even dare to public build
• Or you spend money on ads? You definitely don't want to either
• You don't do anything and don't want to spend money - even gods can't help you

🤖 Deep Thoughts on AI

🌪️ How to Catch the Wave

How to catch the wave? Jump into the eye of the storm first. Don't worry about whether you can catch it, just jump!

The external opportunity that keeps me determined is that I firmly believe AI is like mobile internet from ten years ago. It's just that AI is more suitable for international soil, so I resolutely chose overseas as my market for development and operations.

Another point that makes this Taurus feel very good - currently, AI's early stage is like mobile internet's early stage. To compete for users, there are tons of AI productivity resources you can freeload. Freeloading is earning. The later it gets, the more expensive it becomes, so freeload like crazy. Gemini API has free quotas, occasionally new LLMs come out with various promotional activities, tokens are like free money, giving me a feeling that the more I spend, the more I earn.

🔥 Claude Usage Experience

Claude is the AI you must subscribe to at your current stage - not an option, not "maybe each one suits different people." Stop talking, subscribe to Claude and go hard.

• Previously, when Claude Code only supported Max, I spent nearly $1000 a month using it intensively. Felt amazing
• Coders ask why use Claude Code when this project doesn't involve code development?
  • Use it for batch work!
  • It can make HTTP requests!
  • It can call MCP!
  • As long as your production environment has API or MCP, you can have it execute directly!
• Currently using Pro version can also use Claude Code, incredibly cost-effective
• By the way, to smoothly experience international AI in the future, I specially got a Nigerian virtual credit card

Other AI Tool Comparisons

• Gemini CLI: Free but mediocre to use, often acts like an idiot and argues with you, but free with large quotas, not unusable
• Domestic Qwen Code: I haven't used it because I fear they don't even have basic search and fetch external network functions

🎯 Industry Choice in AI Era

If AI is mobile internet, you can think about which industry you'd choose under mobile internet. Remember those various waves from those years? O2O, mobile internet finance, sharing economy, etc.

Currently, for me, undoubtedly, I can only choose AI + independent website cross-border e-commerce.

I often secretly study and observe many coders. They're exactly like newbies selecting products in cross-border e-commerce, always choosing the same few things - jewelry, pets, blah blah blah.

But, family, listen to me:

You need to use code's basic capabilities + AI empowerment to quickly dimensionally reduce and attack a traditional industry. Don't always develop "AI image generation," AI this, AI that. I've probably seen 1 million AI tool navigation sites. Why insist on competing in that space?

Can't you make an AI-driven navigation site for other industries?

This is like how I often advise everyone on product selection. Which industry or lifestyle do you understand better? For example, if you like skiing, can't you make an AI-powered global ski resort navigation site? There are so many possibilities.

AI is a tool. You don't have to produce an AI tool. Instead, use AI to produce better, more efficient solutions for traditional industries.

🌏 AI's Practice Environment

The only practice environment for AI is overseas. Listen to advice.

What are you doing domestically?

• No web ecosystem
• Mini-program ecosystem is too closed-loop, you can't get traffic
• App development is the same

Finally, AI individual developers domestically can only become AI bloggers selling courses. Not that selling courses is bad...

🎯 Final Words

This is my first independent development sharing. Many people might mock:

"Haha, this garbage project, you talk about it like you've already achieved good results. What's there to brag about? Come back and talk when you make money."

This is the biggest challenge entrepreneurs face - another pessimistic self in your head always jumps out for self-denial:

"Sigh, I definitely can't do it. I'm not doing this well. Others will laugh at me. Forget it, I won't do it, won't talk about it. Embarrassing."

🚀 But who starts at the peak from entry?

Who is perfect from the beginning? Even Elon Musk was globally mocked and criticized like a sieve when he started new energy and SpaceX.

When you can treat entrepreneurial projects as a dynamic, sustainable, long-term game, you'll feel relieved and then peacefully optimize and operate step by step.

💪 Keep going. Mutual encouragement.

‍