I read this piece: https://blog.naver.com/quantdaddy/224448090689. It’s titled “Building a Web App as a Non-Developer, in an Age When Implementation Is Free.” I agree with the broad strokes, but there are two places where I think it’s worth adjusting the lesson to fit this particular candidate’s situation.
What the piece argues
- The cost of building is heading toward zero. So you should try things — small things, constantly.
- But doing a lot of whatever is just noise. What survives in the end is “core product thinking.”
- An eye for choosing problems: the ability to pick problems worth solving that you’re well suited to solve
- Domain judgment: the ability to tell signal from noise
- The imagination to picture the user’s experience: being able to see who opens this, when, and in what mood
- When you work with AI, the human part is the two ends. The planning at the very front and the verification at the very back. And the author says the real problem after building something was “who do I get this in front of, how, and how does it turn into money?”
Where I agree
The picture of “what’s left for humans is the planning up front and the verification at the back” is the most accurate part of it. And it maps directly onto this candidate’s situation.
- Career: the destination they’ve settled on — “measuring effects tied to money” — is that back end. Judging whether the thing you built actually worked is not something AI can take responsibility for on your behalf. It’s the same reason I told them to deliberately collect measurement experience, whether at Enliple or CJ.
- Résumé: the case-card format they’ve already settled on (model · input · output · improvement · measurement) is exactly this lens. “Why I chose that problem and how I verified it” is worth more than “here’s what I built.”
Two things worth adjusting
1. “Implementation is free” only holds up to the toy stage. The cost of building a demo has fallen close to zero, but operating is not free. The things I’ve lived through are the evidence.
- PHP 5.3 and utf8mb4 constraints on a live e-commerce server
- A database outage caused by crawler bots
- Swapping out a recommendation table with zero downtime
- On the job board, too: 211 postings the collector had quietly dropped
“A first version that runs” is free. “Something that runs every day without getting it wrong” is still expensive — and that’s where the gap in skill still lives.
2. “Try a lot of things, constantly” is a strategy for people with time to spare. If family comes first and you’re the sole earner starting next May, the time available to pile up attempts is limited. So I think the advice should be rewritten like this:
- Rather than increasing the number of attempts, tighten the criteria for choosing problems.
- Before you start, decide in numbers when you’ll walk away.
Honestly, the part that stings most for this candidate
The question the piece leaves at the end — “who do I get this in front of, and how?” — is the biggest blank right now.
- The speed of building is already there. There are more than forty projects, and several of the Web Doctor tools are actually running.
- And yet Web Doctor has 13 published pieces, roughly 9,500 views, and one inquiry. The bottleneck isn’t implementation. It’s problem selection and distribution.
- To borrow the piece’s own framing: what’s needed now isn’t buying more lottery tickets. It’s reading which ticket has a tail.
So, if I’m turning this into practice
- Write three lines before building anything new: Who wants this? What are they getting by on right now? How do I reach that person? If I can’t write the third line, I don’t build it.
- After building, judge success only by the number of inquiries and DMs. That’s a standard already set, and it’s the same conclusion this piece reaches.
- The same principle at the next company: become the person who “chooses which problems to solve and proves the effect,” not the person who builds more models. That’s the road to AI team lead.
To sum up, the piece’s conclusion — “planning and verification are what remain” — is right. But for this candidate, the weakest part of the planning side is “choosing the problem and getting it in front of people,” and that’s also the part whose value will rise the most.
댓글