5 Questions Before Building (A Sustainability Check)

The first question isn't "can you build it?" — it's "can you stick with it for a year?"

The trap of the "decent idea"

GitHub is full of "dead projects." Repos that started with a fancy README, last commit six months ago. Korean dev communities like Disquiet or Brunch blogs are full of retrospectives like these: "Finished an MVP in 6 weeks but couldn't maintain it." "Got good feedback but gave up 3 months later."

Everyone starts with a "decent idea." The problem is that nobody asks from the start whether it's a "decent" idea or an "I can stick with this for a year" idea. This post isn't 5 questions to judge the quality of your idea — it's 5 questions to confirm whether you're entering a relationship with this idea that you can't easily break.

Q1. Can you stick with it for a year?

"A good idea is good. But the question is whether you can love it." That's the core of Paul Graham's YC philosophy. He has seen 1,000+ founding teams, and the common pattern in failures wasn't bad ideas — it was that the team couldn't sustain that idea for over a year.

Let's sketch a persona — Seoyeon Park (35, 6 years in marketing). Over the past 4 years she has written down 6 ideas of the form "I wish a tool like this existed." She wants to pick one. But the question she really needs to ask is not "is this idea good?" — it's "can I think and talk about this for 10 hours a week for a year?"

Because MVP build (3 months) + first user acquisition (2 months) + feedback and iteration (6 months) = at least 11 months.

Toss CEO Sunggun Lee is famous for "8 failures, 9 tries." In his retrospectives, most of those 8 attempts failed because of "market not ready" or "founder not focused." The 2011 attempt UlaBla (an SNS) was a decent idea, but Korea was being absorbed into Facebook at the time, and his energy was scattered. Even if market timing is right, an idea dies without founder focus.

What's the signal that you "can stick with it for a year"? Are you thinking about this idea on the bus to work on Monday morning? Are you analyzing competitors during lunch? Are you reading user reviews at 11 PM? Then you have sustainability. If not, this is "something I have to do," not "something I want to do."

Q2. Can you name someone who would pay?

"Good" and "sellable" are different. Paul Graham repeats this advice:

"Is anyone in enough pain about the problem you're solving to actually pay?"

Most failed ideas: the founder genuinely believed they had spotted a real problem. But it was just their problem. Say persona Junho Lee (28, backend dev for 3 years) says, "I'm going to build an AI prompt management tool for developers." Good idea. But did he ask 2,000 developers up front whether 100 of them would pay 9,900 KRW a month?

Sahil Lavingia, in The Minimalist Entrepreneur, advises: "Start with community. Find people who would pay first, then listen to their specific pain." And ask how much they'd actually pay for that pain. If the answer is "I'd use it for free, but paying I'm not sure," that's a problem not worth solving.

A line that comes up often in Korean startup post-mortems: "The product wasn't bad. There just weren't any customers."

"People are showing interest" and "people will pay" are not the same. Instagram likes, survey responses, "what a great idea" — these are barely signals. The only signal is pre-payment, signup-then-purchase, or beta usage fees actually being collected.

Q3 & Q4. Is AI making it possible? + What's your unfair advantage?

The new condition for evaluating ideas in 2026 isn't "can you do it?" — it's "can AI help you do it?" But this is a double-edged sword. If AI can do it, your competitors can build the same thing with the same tools.

If your idea is "you can build it by stitching together 5 AI tools," then practically your competitive edge is near zero. Another developer can build the same thing in 3 months.

So the question becomes: "What's the unfair advantage that only I have in solving this problem?"

The unfair advantage should be one of these:

1. Deep understanding of a specific industry: "6 years in marketing" → building marketing tools
2. Existing customer pool: "100 people in my startup community"
3. Speed of building: "I'm faster than others at Bolt"
4. Understanding of the Korean market: localizing overseas tools for Korea

If you don't have any of these, an MVP will get you to market, but competitors will show up within 3 months.

Q5. Can you decide not to build it?

This is the most important and most often skipped question. "Can you decide not to build?"

Paul Graham said: "The best ideas are the ones you can't easily give up. Conversely, ideas you can give up are ones you shouldn't start in the first place."

If you can't decide "I won't build this," that's not the idea being strong — it might be your ego being strong. Giving up a project you've already started is psychologically stressful. So before you start, define clear kill criteria.

For example: "If I don't have 10 first users a month in, I quit." Or: "If signups don't grow in 3 months past the last commit, I'm done." Objective criteria like this protect you from the sunk-cost trap later — "I've already invested too much, I have to keep going."

Deciding not to build isn't weakness. The problem with many failed Korean founders was actually that "they couldn't give up giving up." The resource for your next MVP comes from the decision to kill this one.

The 5-question decision tree

Evaluate your idea with these 5 questions.

Q1: Can you stick with it for a year? → No → next idea
Q2: Can you name someone who would pay? → No → re-do market research
Q3 & Q4: Possible with AI, and what's your unfair advantage? → No advantage → intense competition
Q5: Can you decide not to build it? → No → check your ego
All pass → GO. Start the MVP.

The key is that you should have a clear answer to every question. If even one is hazy, don't start.

Real-world workbook, finish in 30 minutes

For the idea you're currently considering, fill in the following. If 4+ of the 5 questions return a "strong signal," you can start. 3 or fewer — think again.

• Q1: Can you think about this idea for a year? (Which day last week did you think about it?)
• Q2: Specifically, who would pay? (Name, company, income range)
• Q3: Is it impossible without AI? Sufficient with AI?
• Q4: What's your edge? (Industry experience / customer pool / speed / local understanding)
• Q5: What's the kill criterion? ("By ___ date, if ___ doesn't happen, stop")

Wrapping up

An idea that passes the 5 questions is now ready to validate. But validation has a "method" too. The next post covers using AI as a research partner to survey the market in 30 minutes and validate your hypothesis. We'll start answering "will these people actually pay?" with data.

References

1. Paul Graham, "How to Get Startup Ideas," paulgraham.com, 2012. paulgraham.com/startupideas.html
2. Sahil Lavingia, The Minimalist Entrepreneur: How Great Founders Do More with Less, Penguin Random House, 2021.
3. "7 Startup CEOs on Why I Failed," Korean Economic Daily, 2022.08.24.
4. "Failure Cases: Common Mistakes in Solo Founding," ki-magine blog.
5. "Startup Idea Validation Framework: 5 Powerful Tests," Startup Validation Framework, 2026.

Previous: 90% of MVPs Are Fake: The Real Definition of MVP
Next: Using AI as a Mirror, Not an Answer Machine (Hallucination-Proof Research)

About the characters (Seoyeon Park, Junho Lee)
Characters in this series are fictional personas created by Fail School — composite characters compressing patterns common to people in similar situations. However, company cases (Toss, Kmong, etc.) and citations (Paul Graham, Sahil Lavingia, etc.) are all real.

Minchul Kim, CEO of Freeive, Fail School