
Something really messed up happened yesterday.
I had a chance to talk with a young man in his mid-twenties. His mother runs a daycare center, and she reached out asking whether her son should attend an AI bootcamp or some AI-related program for his job search. She offered about 80,000 won for travel expenses, so I went partly for a change of scenery and partly just to get out, and ended up doing a career consultation with the guy.
To be honest, as everyone knows, I'm not really in a position to be giving career advice, and I know my own situation well. The young man was attending a very good university in Seoul, but he had been sending out so many applications without landing anything that his self-esteem had taken a real hit.
So we talked about whether he should stay in Seoul for an AI bootcamp or come back down to Changwon in South Gyeongsang Province.
But as we talked, something felt off. The terms were too strange. On top of the AI bootcamp, there was an additional fee for what the instructor called "private lessons," which supposedly let you do real hands-on work. It didn't make any sense, so I pressed him on it, and he told me the instructor takes on freelance projects and then farms the work out to students. On top of that, the students were the ones paying for the privilege.
I already can't wrap my head around why you'd need a bootcamp to learn AI when everything about how to use it is right there in the Claude docs and Codex docs, but beyond that, what practical skills are you actually supposed to build by using AI? Are they teaching you how to use Python's openpyxl?
Even if I give every possible benefit of the doubt and assume the instructor is teaching genuine foundational programming principles rather than just surface-level usage, it's still nonsense.
For starters, I'm skeptical of what schools market as "foundational CS for programming." Real-world programming has always been closer to plumbing: grabbing pipes someone else made and connecting them together. Genuine academic depth only matters when you're designing a completely new algorithm from scratch, and the average programmer never does that in their entire career; even exceptional programmers might do it a handful of times at most.
The vast majority of real work is picking up tools that the world's greatest minds have already built and putting them to use. Realistically speaking, when Korean developers try to design new algorithms, almost all of them perform far worse than existing open-source implementations. There's an old saying in this field: "don't use anything with a K in front of it," and it exists for good reason. Even in pure CS and coding, the core skill is already about connecting things others have built well, so what deep secret knowledge could possibly justify 1,000,000 won a month in private tutoring for "AI utilization methods" that amount to making API calls to models the big tech companies have already built?
Some might push back and say it's not just simple API calls but advanced techniques like the trendy agent "harness design." That argument makes even less sense.
Look at where AI harnesses are headed today. The phase of stringing together a few prompt lines like in the early days is long gone. The trend now is toward adopting Blackboard architectures where agents collaborate over shared memory, and bundling workflows into structures like DAG (Directed Acyclic Graph) state machines. There are even approaches combining formal verification at the compiler or type-checker level to ensure that an AI's outputs don't violate the system's invariants.
The methodologies vary widely, and when you're actually building an AI harness, you need to handle state management as a baseline while designing the entire execution system from the ground up. How are students supposed to grasp any of that?
This is essentially a convergence of distributed systems and compiler theory, so what on earth is this instructor doing messing around with students like this?
In the end, what that instructor is selling as "real AI work" is a perfect contradiction. If it's simple API plumbing that anyone can do after reading the official docs, charging 1,000,000 won a month is a scam. On the other hand, if they're genuinely teaching harness architecture with DAGs and formal verification, then they're selling a fantasy that no entry-level developer can possibly absorb, so that's also a scam. No matter how you look at it, the conclusion is the same: it's nothing but predatory exploitation, dressing up buzzwords like "harness" and "agent" to look cutting-edge while emptying the pockets of desperate young people and conscripting them as cheap subcontractors for the instructor's own freelance work.
The premise that an individual could be made to surpass software maintained by the world's best programmers never made sense to begin with, but setting that aside, is this really the right way to treat desperate young people, charging them 1,000,000 won a month like this?
The mother had asked me to advise her son because I had been brought in as an AI instructor and she assumed I would know. What I had basically done for her was build an MCP in Python that auto-generates newsletters for the daycare. I charged very little because I needed to build a reputation and wanted a long-term client.
The instructor takes on freelance contracts and subcontracts the work to students under the banner of "real-world experience," yet instead of paying those students, he charges them 1,000,000 won a month for so-called "private tutoring fees"? This isn't education. It's holding young people's desperation and anxiety hostage while running what amounts to a new-age digital pimp operation and a vicious pyramid scheme. At least the old "passion pay" culture didn't reach directly into your pocket. Now we've arrived at an era where young people desperate for a line on their resume pay out of their own pockets to do a fraudster's subcontracted work for him.
1,000,000 won a month is not pocket change. I never understood the lecture-hawking crowd, and I still don't. This guy claims he came from a major programming company, but it's not like he designed the entire system at that company himself. What I can't grasp is this: if he wants small change, why not just offer something affordably priced on the market instead of squeezing it out of people?
This bootcamp is a filter that makes students "buy the signal of real-world experience with money." Neither side is creating value; both are just controlling scarcity to collect a toll. But the student told me he had to sign up for something like this because companies demand experience.
I was at a loss for words. The job market is far removed from my world, so I couldn't say anything with confidence, but I started thinking: what AI utilization, really? There are so many flawed AI use cases even at large corporations... I tried to continue the thought but it just circled around in my mouth and never came out. I can't even figure out my own path forward.
I just showed the student how to look up Git usage, how to sign up for GitHub, the docs on using Git, and how to use Claude Code and Codex. Then I told him to build things through conversation using those as a foundation and get a feel for vibe coding.
I didn't explain things like how AI-generated code tends to be overly defensive, or how the code can feel off depending on the person. I just told him to try building something. On the bootcamp question, I told him to find one that actually pays him. Government-subsidized programs at least give you a stipend, I said, and in my view the whole "real-world freelance projects" pitch doesn't add up. In practice, getting freelance work channeled through a single instructor personally is almost never a legitimate arrangement.
This is especially true when you consider that typical programming contracts, right from the initial estimate, require listing the people needed, headcount, and feature requirements. Real programming contracts start by specifying conditions such as whether it's a progress-based payment contract and what counts as completion, then list the project participants, and listing an intern there makes no sense. And if it's purely the instructor's personal freelance work, the actual amounts are small and the scope is predictably limited, so whatever "AI practical skills" you pick up in that setup amounts to nothing more than the instructor using the student's expensive AI subscriptions on the cheap. It's not hard to generate AI code once you've drafted a UML diagram. But if a student doesn't even know what object-oriented programming means, how are they supposed to code anything? Setting aside FP, domain-driven design, and data-oriented design, shouldn't you at least have a basic understanding of what code looks like and how it's structured? Without even that foundation, I honestly don't know what they think they're learning.
I've already worked in this field, but these days I worry about where my next project will come from. That young man wants to get his start, but he says even getting the first opportunity to begin is out of reach.
Maybe we're both facing the same problem from different positions.
That's probably why I couldn't tell him with any confidence what he should do. I don't know how I'm going to survive going forward either.
But one thing keeps nagging at me.
A young person who wants to work and earn money has to pay money just to get a chance to try working.
Is that really what we mean by education, real-world experience, and job preparation?
I honestly don't know. It just felt really messed up.