No Coding Skills? How to Earn 30K RMB a Month Taking AI Freelance Projects

A creator who can't code used AI tools to complete the entire process—from client communication to project delivery—in one month, earning 30,000 RMB. The core of this methodology isn't technology; it's about using AI to turn ideas into deliverable products.

Why do people who can't code often make more money with AI?

In the video, the Bilibili creator puts forward a counterintuitive idea: many of the people who are best at making money with AI can't code. Some have built dozens of products single-handedly with AI; some, with no technical background, made a small tool, put it up for sale, and orders haven't stopped since; others built things like resume optimizers and viral copywriting assistants and sold them directly for money.

What these cases share is: one person using AI to turn an idea into something sellable. Last month, the creator took on a project using the AI tool Trae (referred to as "Tree" in the video) and delivered a complete project directly. He emphasizes that many people have ideas but can't produce anything deliverable—and that's exactly the problem AI is meant to solve.

Step one: decide in three minutes whether to take the job

The creator received an inquiry from a nail salon client: customers were all booking times over WeChat, and double-bookings kept happening—could something be built so customers could schedule and book on their own? He didn't rush to start working; instead, he first asked about the budget.

Budget directly determines whether the job is worth taking. Beyond that, he runs through three checks: can the features be narrowed down, is the tech stack clear, and can the deliverable be verified. Only after passing all three does he expand the client's one sentence into a requirements description and hand it to Trae.

He typed into Trae: "I've taken on a project—judge whether this direction is feasible." Then he attached the client's basic information and requirements. Trae quickly produced a detailed analysis, including a feature flow diagram, a recommended overall architecture (what tech to use, how to implement it), and finally reminded him of three things that needed to be confirmed with the client before moving to the next step.

Using AI for requirements communication multiplies efficiency

Following the H5 solution Trae recommended, the creator had AI put together a three-question survey to send to the client for confirmation.

The survey was generated quickly, clearly spelling out which questions needed to be discussed with the client and what kind of wording to use. The creator could just copy and paste—"I don't need to think through any of these details myself." Whenever he hit a question he didn't understand during the conversation, he could ask AI directly, greatly improving communication efficiency.

After the client's feedback came back, he organized it and sent it to Trae, which combined it with the earlier information to write an actionable technical plan. Then he had AI break the plan into a task list and build the entire project step by step.

The H5 solution refers to a mobile web application built on HTML5 technology, which users can access directly through built-in browsers like WeChat without downloading or installing a native app. For small offline merchants like nail salons, the advantage of the H5 approach lies in low development cost and low promotion barriers—customers can complete a booking by scanning a QR code or opening a link, and the shop owner doesn't need to maintain app store listings or version updates. Compared with mini-programs, H5 is slightly weaker in interaction experience and WeChat ecosystem integration, but it wins on being lightweight and fast to deliver, making it very well suited to small booking systems with clear requirements and a narrow feature set.

A working system in 7-10 minutes—but the human makes the decisions

Once AI started executing, a working system emerged in 7-10 minutes: customer side, shop-owner side, admin backend, and scheduling—all there.

The creator especially emphasizes one key point: I'm in charge at every step—I can pause, change, or overturn anything AI does at any step. "That's where the confidence to take on jobs comes from—not AI making decisions for me, but me checking things from behind."

If he's unsatisfied with certain pages or details, he can have AI revise them directly. This "human-AI collaboration" model lets people who don't understand code control the project's pace.

Delivery: AI helps you list items, package everything, and write the talking points

Once the project is done, delivery isn't just about sending over a file. The creator told Trae directly: "The project is done and I need to deliver it to the client—what exactly should I deliver?"

AI lays out the delivery checklist in detail, helps you package the project, and finally lists the points to note and the talking points for communicating with the client, making sure nothing goes wrong. All you need to do is follow the instructions.

The same process works for anything

According to the creator, this process isn't just for big jobs—it works just as well for small ones. For example, when an online store's order export comes out a mess, with mismatched column names and duplicates, you can use Trae to make an Excel cleanup tool: drag the table in and it automatically removes empty rows, removes duplicates, and sorts by date. A small tool like this can also be sold online.

Media professionals can build a small tool for breaking down and rewriting viral content; for online store clients, you can make marketing-oriented QR code mini-games for traffic generation. "Demand is infinite—as long as you've mastered this method and use the right tools, the same process can do anything."

AI programming tools (like Trae) are a class of integrated development environments that support natural language interaction: users describe requirements in Chinese, and the tool generates code, explains logic, and even scaffolds a complete project. Trae's typical working mode is "conversational development": first analyze the feasibility of requirements, then output a technical plan and task breakdown, and finally generate runnable code step by step. Such tools lower the barrier to programming, but they don't mean full automation—the quality of the output depends heavily on how precisely the user describes requirements and how well they control the process, which is exactly why the original article stresses "I'm in charge at every step."

Can this method really be replicated?

Judging from the process shown in the video, the creator's core skills aren't actually coding—they're requirements breakdown, client communication, and project control. AI handles the technical implementation and document organization, but judging "whether to take this job," deciding "which direction the plan should go," and checking "whether each output is correct" still depend on human experience.

For beginners who want to try AI freelance work, this process offers a reference framework: first use AI for feasibility analysis and solution design, then confirm requirements with a survey, then have AI break down tasks and generate code, and finally use AI to organize deliverables. The key is to maintain control over every stage rather than tossing the whole project to AI and walking away.

Trae, mentioned in the video, is an AI programming tool that supports generating complete projects from natural language. The creator's claim of "generating a working system in 7-10 minutes" reflects performance under a specific level of requirements complexity; real projects still require iterative revisions based on feedback.

Background notes

Requirements convergence is a key concept in software project evaluation, referring to progressively clarifying vague, sprawling client demands into a set of features with clear boundaries and controllable scope. Judging whether features can converge essentially means checking whether the client's requirements contain a lot of uncertain "we'll figure it out later" items, and whether those uncertainties will cause the project scope to keep expanding. For freelancers, features that can't converge mean uncontrollable delivery timelines and costs—a risk signal that must be ruled out before taking a job. The original article lists "can the features converge" as one of the three checks precisely to prevent projects from spiraling out of control during execution.

One-line Summary

A creator who can't code used AI tools to take on and deliver projects in one month for 30K RMB, with the core being requirements breakdown and project control.

Paragraph Summary

The article describes how a Bilibili creator who can't code used the AI programming tool Trae to complete the entire process—from client communication to project delivery—in one month, earning 30,000 RMB. The core of the method isn't technology but using AI to help judge project feasibility, generate client surveys, break down tasks, generate a working system, and organize delivery checklists. The author stresses that the human always holds decision-making power and can pause or revise anything at any step. The same process works for big jobs like a nail salon scheduling system and for small tools like an Excel cleanup tool that can be sold—the key lies in requirements breakdown, client communication, and project control skills.