What Founders Shouldn't Do

I stumbled on an Instagram channel called Failure Collection. It's a place where founders of once-hot startups openly share how their companies rose and fell.

I found it admirable. Talking honestly about your own failure takes a lot more courage and self-respect than talking about your success. It made me hope our community could become a place where people talk about this stuff honestly, even if it's nothing grand.

Around then, Brian Jin, a Korean Canadian who had spoken at one of our recent meetups, reached out first. Brian wanted to share what early founders shouldn't do, drawing on their own startup failure. While staying in Korea for a while, Brian had met a lot of people and kept getting asked similar questions, so the topic came up naturally.

Brian originally planned to do it as an online webinar. That felt like a waste of a good story, so we offered to handle the venue and the promotion.

So on a Friday at 10 a.m., we got together at D.CAMP Seolleung, a startup center in Seoul.

Only the things you shouldn't do

Brian is an engineer and founder from Toronto. At our meetup fireside chat the week before, Brian told the story of how, right before an accelerator investment, the investors' read on the market flipped completely and all 18 teams in the same cohort stalled at once. This time Brian had boiled that experience down into five mistakes. I liked the opening line right away: the internet is full of things you should do as a founder, so today we'll only talk about what not to do.

What Founders Should Avoid. Five mistakes that get expensive fast.

And this sentence was on the first slide. I think it ran through the whole two hours.

Bad advice is often good advice applied to the wrong company.

Since my job is putting together rooms where founders trade advice every week, this one stung. I watch good advice clash with other good advice all the time, and I'd never heard the reason put so cleanly. B2B and B2C are different. Raising VC money and bootstrapping are different. Cross those two axes and you get four boxes, and each box has its own definition of good. If you're a bootstrapped B2B company, revenue comes first and you should keep your customer segment narrow. If you're a VC-backed B2B company, the bar is a big market and a repeatable way into it. The same advice turns toxic in the box next door.

One example settled it. A company that reliably brings in $10,000 a month in subscriptions is a great company in the bootstrapping world. In the VC world the question changes to: if I put money in now, how long until I get 10x or 50x back? In front of that question, this company isn't good or bad. It just can't be evaluated at all. That's why you have to figure out which box you're standing in before you take anyone's advice.

They don't know what's investable either

The question the partners at Antler always asked was "Is it backable?" Is this worth investing in? I appreciated how honest Brian was here: "We didn't know what backable meant either. We still don't." It can't be defined with numbers. The conclusion Brian came to was speed. Unless you're in deep tech, robotics or defense, you're probably building SaaS, and in that game, however big the market is, there are so many similar companies that it comes down to who takes the market first.

So what pre-seed investors look at is founder-market fit, not the product, and at this stage you don't need PMF. Their position is that you'll figure that out later. The interesting part is that this matters just as much if you never take VC money, and the reason is almost anticlimactically simple: startups are hard. To last, you need persistence and resilience, and for that, it has to be a problem you've lived through yourself or something tied to a personal mission. Brian said that for B2B in particular, it absolutely has to be a problem you've experienced firsthand.

One attendee preparing to start a company in healthcare pushed back. If you start in your own domain, you can't quit even when you want to. It's a small world where everyone knows what you're doing, and if you fold and go back to your old job, you have to be ready for people to point fingers. Your domain knowledge becomes both your weapon and your shackle. In a textbook, founder-market fit is one line. For the person living it, it can mean losing your way out. That day was the first time I'd thought about it that way.

Brian also shared something VCs always demanded: explain the game you're playing without mentioning your product, in one line, like you're talking to a ten-year-old. Brian admitted they still can't do this to their own satisfaction, and honestly, I'm not off the hook on that question either.

Watch what customers do, not what they say

The second mistake is building what customers say they want. The slide had two columns side by side. What customers say: "Can you automate this?" "Please add this feature." "I'd definitely use this." What customers actually do: work around it in Excel, get approvals over Slack and email, copy and paste, ask the person next to them. What people say is a hypothesis. What they do is evidence. If you want to ask better questions, Brian said, read The Mom Test. Ask about past behavior, not future promises.

Replace "Would you use this?" with "Show me how you do this today."

The ultimate version of "show me" is job shadowing, where you sit next to someone as if you'd been hired at their company and watch how the work actually gets done. Brian tagged along when a design partner met with one of their clients, and the important part turned out to be what happened after the meeting. Once it ended, that person dropped an Excel file into ChatGPT, uploaded an older file alongside it, and told it to make the new one in the same format. In interviews, everyone had only talked about Excel. If Brian hadn't followed them in, they would never have seen that.

"We always thought we were competing with Excel. It turned out we were competing with ChatGPT." That line hit close to home. The closer your business sits to language models, the faster it becomes a declining industry. The healthcare attendee added something in the same vein: these days even doctors are vibe coding, and some of them build and use their own tools instead of buying a solution.

Learn instead of building

The third mistake is building too much before you know enough. When founders feel uncertain, they react by building more: one more integration, one more feature, or a whole platform. Instead, they should ask what their riskiest assumption is and what the cheapest test is that could prove it wrong. Signals have tiers, too. Meetings, letters of intent, free pilots and logos on your deck are weak signals. Budget getting allocated, repeat usage, workflows changing and accounts expanding are strong signals.

Don't build more. Learn more.

The strongest advice of the day came here, something Brian heard directly from a VC: "Don't build this. Don't even make a landing page. Go meet customers, no matter what. You're going to build it eventually anyway." As a developer, Brian found building the most comfortable thing to do, since you can see it and it feels like progress. But a startup is about building a business, not a product, and at this stage your day should be talking to customers from 9 a.m. to 9 p.m.

Someone asked how far you should validate before you start building. Brian said there's one textbook answer: people give you money when there's no product yet. "I'll pay you, build it for me." "I'll invest, build it." Anything short of that is still a hypothesis. Interviews only make a hypothesis look more like a hypothesis. They're not demand and they're not traction. Brian added a warning not to start with enterprise sales: in markets where a sales cycle takes a year, an early team burns all its time on a single round of validation.

Investors aren't ATMs

The fourth mistake is treating fundraising like a loan. Along with a warning not to play games with people you might be working with for seven to ten years, the slide said this:

Choose investors like partners, not ATMs.

The stories from the field were a lot rawer. For three months before the Antler program started, Brian met with the partners every two weeks. They never once asked what Brian was building. They only asked for numbers: how many design partners do you have, have you gotten letters of intent. What Brian took away from that was, "Wow, this is really brutal."

How to work your way through a fund was the most practical part of the day. At Antler Canada, the investment committee is made up of one managing partner, two GPs, one AP and two directors. If one of them backs you, that person becomes your champion and fights for you on the inside. You pitch the investor, and then the investor pitches you again internally. Funnily enough, Brian's champion was the AP, the most junior person in the room, which shows that actual influence matters more than title. Brian added a tip: the investment memo your champion writes gets shared ahead of time with other VCs preparing for the next round, so you should know what sections go into that memo and feed them the right information.

Brian was refreshingly open about the terms, too. Antler Canada invests $250,000 for 9 percent equity, with a separate $90,000 program fee. In Korea, Brian heard, it's 100 million won (about $70,000) for 13 percent. The plan for using that money comes with rules: no founder salaries and no hiring. All of it goes into sales and marketing. One team Antler invested in has three people, still has no product and no CTO, but it has customers and gets paid. There's a CEO with a marketing background, someone on sales and someone on product, and they do development with AI. The share of the founding team that's technical is shifting from 50/50 to 80/20, Brian said.

YC came up too. Brian applied to YC 10 times and was rejected 10 times. After finishing Antler, Brian sent a LinkedIn message for the first time instead of an application, and got a reply. It said it was still too early, but it was a reply. One message got a response that 10 applications never did. Brian also pushed back on the rumor that YC invests in ideas: YC requires you to have launched, no exceptions.

Brian didn't forget the warning from the other side, either. "Assume 90 percent of the people you meet in San Francisco or New York are scammers." There are people who offer to be your advisor and ask for 5 percent equity, and brokers who want a commission for an introduction. So founders should do reference checks on investors too, and in particular ask founders who've been through hard times with that investor.

No moat, only distribution

Something I heard at our last meetup came up again here. The Antler partners raised two issues with Brian's business. First, will you survive if Claude ships a new version tomorrow? Second, Claude got so good at coding that suddenly there are a hundred companies doing the same thing. The second one was heavier. The first is your own problem. The second is a problem for the whole market, and no matter how well you do, no winner comes out of it. You can't build a moat with technology anymore. So what's left? Distribution, no matter what, Brian said.

Distribution turns product into a company.

Distribution sounds like a big word, but the example that day was a mascot costume. A founder who started in Korea and now runs the business in the US once put on a mascot suit at a COEX event (COEX is a big convention center in Seoul) and handed out bags to foreigners. Since nobody could see their face, it was easy to walk up and start talking to people. One person who showed interest that day led to the move to the US. I liked how Brian read it: the door opened because they were the kind of person willing to go that far, not because the idea was good.

I also learned one way to read where things are heading. VCs publish a list of areas they want to invest in, called Request for Startups, but if you prepare based on that list, you're already late. There's an old story that the people who made money in the gold rush weren't the ones digging for gold but the ones selling jeans. Everyone knows the AI vein is going to blow open, so the question is what the jeans are right now. Brian's answer was infrastructure, rather than verticals that dig into a single domain. Like LangChain, which gives its framework away as open source and makes money with LangSmith, the tool for monitoring it.

Don't outsource your judgment

The fifth mistake was the conclusion of the session. Everyone gives you signals, but nobody owns the decision. Customers know the pain but not the solution. Design partners give you depth but drag you down the customization rabbit hole. Investors see patterns, but their pattern matching is imperfect. Advisors' experience comes from a different company in a different era. Metrics only show you what you decided to measure. So the order is: listen, observe, test, and then you decide. There's one question to ask yourself: what evidence would change my mind?

The method was concrete too. Keep a hypothesis log of what you believe and why. When you make a decision, write a memo with the trade-offs and the outcome you expect, then later compare it with what actually happened. Brian added that salespeople are the best at this, and technical people like Brian find it the hardest. Telling signal from noise is the hardest part, and over time, things you threw out as noise sometimes turn into signal. Keep running that loop and you land on one of three outcomes: keep going, shut it down, or change course. If you change course, you go back to the start.

The founder's job is to get less wrong, faster.

Roundtable: everyone's failures on the table

After the seminar, we spent an hour in groups of six sharing our own failures. The heart of a Bloom event is always this second half. People are more honest in front of others who have no stake in their business, so stories come out that usually wouldn't.

The healthcare attendee's story was heavy. Before starting a company, they had spent years building products inside the industry and thinking hard about whether startups and healthcare can coexist. What they saw was that the moment you take investment, there comes a point where VCs start pulling you along. They felt they'd end up in a situation where they couldn't quit even if they wanted to, so right now they don't take even angel money. Brian related to that and said the reason for joining Antler wasn't the investment either. It was to find a cofounder.

Government grants came up too. One person said that once you start fitting yourself to grants, the eyes that should be on revenue end up on reports and requirements, and before you know it the company's whole constitution follows the grants. So at this stage they don't go looking for them. Canada works differently. There's no money up front. R&D payroll comes back later as a tax credit, and the bigger the amount, the more matching outside investment they require. The private sector validates first and the government follows.

The last topic was a question Brian turned back on the room: "Why does everyone want to start a company in the US?" One person said that a former AI startup they'd worked at closed a Series B of 15 billion won (about $11 million) from Korean VCs, while a US company with the same product raised $100 million at Series A. And Korean investors keep telling founders to go to the US anyway. So if you're going to do the same thing either way, why not start in the US from day one. That was their conclusion. The healthcare attendee said Korean healthcare is barren ground for business, so the answer was the US even before they started. Someone else shared that once you launch a product and gather users, big companies ask for meetings and tell you to come join their team. Brian added one cold observation: they've almost never seen someone who lived their whole life in Korea go straight to the US and succeed. Immigrants and people who studied abroad are a different story.

We wrapped up on cofounders. Brian asked a friend in San Francisco who invests from Series C onward what the biggest problem is these days, and the friend said cofounder dynamics. A lot of companies collapse because the founding team breaks up, even after they've started making revenue. Antler hands out a template of 50 questions for cofounder alignment, and Brian went through all of them in one day. Each person writes their answers ahead of time, then you swap computers and read through each other's answers one by one. Brian said the process of reading them out and getting aligned was more valuable than the answers themselves.

If I could keep only one thing

If I could take only one thing from that day, it would still be the first sentence. Bad advice is often good advice applied to the wrong company. Startup advice is everywhere, online and at events, and I'm one of the people putting on those events every week. This is why all that advice keeps clashing. Until you figure out which box you're in, no advice is really yours. And what you do after picking your box comes back to the fifth mistake. Take signals from everyone, but own the decision yourself.

Thanks to D.CAMP Seolleung for sponsoring the space, and to Brian for reaching out first and putting their failures into a talk. There are plenty of places to share success stories, but very few to share failures. I plan to make more of them.

Originally written in Korean on Brunch, August 29, 2026.