Student Question:
I’m a mid-level UX designer at a tech firm in Shenzhen, earning around 220,000 yuan a year. Not bad, but not where I want to be. A friend told me I can earn solid extra income doing AI training tasks — labeling datasets, writing and rating prompts, reviewing and ranking AI-generated responses. Platforms like Outlier and others are paying $15 to $50 USD per hour for this kind of work, and some people claim even more. I have the skills these platforms are looking for. Is this a smart way to build side income?
Master Chi’s Response:
Before I answer your question, let me tell you what your friend did not tell you — and what these platforms will certainly never tell you: you are not describing a side gig. You are describing a confession.
When you sit down to rate AI outputs, to write prompt-response pairs, to rank which machine answer sounds more like a real professional — you are not performing labor. You are performing a transfer. Your years of taste, instinct, and professional judgment are being compressed and fed into a system whose entire purpose is to render people with your exact profile unnecessary. The platforms pay you in dollars. You pay them back in the one currency they cannot yet manufacture on their own: genuine expertise. And once enough of you have made this trade, the expertise itself becomes a line item with a depreciation schedule.
This is what I call permission collapse. Not “being replaced by AI” — that framing is passive, it implies the machine is advancing regardless of your choices. Permission collapse is something more precise. It means you reached out, created an account, verified your email, and handed over the instruments of your own displacement. Voluntarily. With a tax form.
Have you watched what happened to the professional translators who spent two years submitting their finest work to machine translation training sets for a licensing fee? Have you watched what became of the illustrators who sold ten thousand pieces into AI art datasets because the one-time payment felt like found money? Do you imagine they look back on that as a wise allocation of their expertise?
Let me share the story of a community member I will call Wei. She was a content strategist at a consumer goods company in Hangzhou — salary in a similar range to yours, bilingual, the kind of professional her team quietly depended on to make the work actually land. Sharp instincts, genuine taste.
Two years ago, Wei started taking AI annotation contracts on evenings and weekends. Writing prompts, rating outputs, flagging errors in AI-generated marketing copy. The pay was real. She cleared an extra 6,000 to 8,000 yuan in good months and felt, as she described it to me over dinner in Hangzhou last autumn, like she was “staying ahead of the curve.” She was proud of herself. She thought she was playing it smart.
Eighteen months in, her company began piloting AI tools for precisely the kind of work she performed. The tools were not spectacular — but they were adequate. Adequate was enough for her manager to ask why the team needed four content strategists. Wei was not laid off. She was “transitioned” into an AI oversight role. Reviewing outputs. Flagging errors. At the same salary, with a lower title, no promotion track, and a ceiling she could see from her desk. She had spent eighteen months training her own successor and called it a hustle.
Now let me tell you about the other person. A UX designer, not unlike yourself. He was shown the same annotation platforms around the same time, considered them for about a week, and decided they were not a transaction he wanted to make. He took those same evening hours and began building something different: a small consultancy helping mid-sized e-commerce brands understand which AI tools were worth deploying in their product workflow, and how. He charged 3,000 yuan for an initial half-day audit. Within eight months he had raised the rate to 12,000 yuan. By the time Wei was explaining her “transition” to her parents, he had more inbound inquiries than he could manage. His destiny framework had shifted entirely — not because his skills were so different from hers, but because he understood which direction to point them.
Same professional background. Same free hours in the evening. The difference was one word.
He who teaches his shadow to walk will find, one morning, that the shadow has learned to cast him.
There are two positions in the AI economy. The low-tier position is feeding the machine — annotation, rating, prompt writing, data cleaning. It pays by the hour, it scales with volume, and it slowly dissolves the market value of the expertise you are selling. The high-tier position is directing the machine — telling it what to do, in whose service, toward what specific outcome, inside what set of constraints the machine cannot perceive on its own. Directing builds an expertise layer the machine cannot replicate because the machine does not know your client’s industry, your client’s competitors, your client’s internal politics, or your client’s actual goals. It can produce faster. It cannot know better.
Every year, that gap narrows slightly. Which is exactly why this particular door — the door of building a directing practice — closes a little more each month you spend feeding instead.
Here is how you move through that door.
First: stop selling your title and start selling your specific read. “UX designer” is a commodity label. What you actually possess is a particular, hard-won understanding of how a certain class of users behaves under a certain set of product conditions in a specific market context. That is not something any platform can extract from you through a rating form. Get precise about what you genuinely know that a generalist does not. The more precisely you can name it, the more specifically you can price it — and the further you move from the territory where AI is already adequate.
Second: produce one public judgment artifact per month. Not a portfolio piece — those prove execution, not thinking. A teardown. A public analysis of why a major app’s recent redesign will not achieve what its team believes it will. A dissection of where a competitor’s product fails its actual user at a specific friction point. Publish it where your clients read. One community member maintained this practice for nine months consistently. By month eleven, inbound inquiries from startups who had found her work were arriving without solicitation. By month fifteen, she was billing 38,000 yuan a month on project work — from a starting point of roughly your salary.
The act of producing public judgment also trains you, each month, to operate at a level above execution. This is the compounding effect. You are not just marketing — you are genuinely getting better at the only thing AI cannot yet commoditize, which is the courage to say “this is wrong and here is exactly why.”
Third: sell implementation certainty, not design deliverables. Your clients today are drowning in AI outputs they cannot evaluate and do not know how to deploy. They need someone who can take what the machine produces and translate it into something that works inside their actual business — with its actual constraints, stakeholders, and users. This is fundamentally a consulting problem, and it prices accordingly. You already work inside a tech firm. You already see where AI tools are creating confusion and chaos for product teams. That internal knowledge is your opening. Begin there, and expand outward.
Master Chi should confess something here. In my mid-thirties, building the early practice, I made a structurally identical mistake — not with AI, but with a business arrangement where I provided access to my methods and frameworks in exchange for a fee that felt substantial at the time. The fee was real. The cost was also real, and I did not see it clearly until half a year of positioning had already leaked away. There is a class of transaction that pays you in one currency while taxing you in something harder to measure. Recognizing that class — learning to feel its shape before you sign — is a discipline that takes years to develop. I am handing it to you now, at the cost of the lesson I paid.
Your major life cycle (大运) in the years immediately ahead will reward the practitioner who built the directing capacity. It will offer the feeders a lower rate each renewal cycle until the rate reaches zero. This is not speculation. The pattern is already visible in the translator cohort, the illustrator cohort, the data entry cohort before them. The shape of this collapse follows a logic that the BaZi (Four Pillars of Destiny) practitioner in me recognizes immediately: you cannot accumulate Chi fortune by giving away the source of it. The energy flows one direction in these arrangements, and it does not flow toward the person filling out the rating form.
You have a real skill. The 220,000-yuan version of you is not the ceiling — not remotely. But the road up is not through platforms that pay you by the hour to accelerate your own replacement.
Stop feeding the machine. Start directing it. There are clients already looking for someone with exactly your background who knows which side of this to stand on. Go find them before someone else does.
