How a Builder Spending $1,800 a Month on AI Survives

AI has made building easy, but landing contracts is as hard as ever. On the Bloom podcast, I sat down with Jeongmin Lee, a builder who has been working on his own since May this year, with no investment and no government grants.

He studied biomedical engineering, started a company as a student, interned at a VC firm, served as CSO at a startup, and is now a sole proprietor doing B2B work. For 50 minutes I asked him where the turning points of the third AI wave were, what he means when he says the fourth boom is when companies use people as APIs, and how one person can earn trust in the B2B market.

His introduction came with a caveat. His original goal was to see whether one person could run a B2B business. After trying it, he found the limits of being one person are pretty severe, so these days he's careful about being called a solo builder. He's even been thinking he should just call himself a builder and leave open the possibility that someone is behind him.

I didn't really like programming

He majored in biomedical engineering and did some programming for things like medical image analysis, but back then he didn't really like it. He laughed that somehow he ended up living like this anyway. In the old days, he said, he couldn't have claimed he could develop software. He could do modeling and write training code, and that was about it. Now he builds web apps himself and delivers them to clients.

The standard track in his major is grad school. Most people go on to more research or become researchers at pharma, chemical or electronics companies. He skipped that track and started a company as a student.

I thought a startup was the ultimate résumé item

Looking back, he admitted he sometimes wonders why he did it. For people who entered college around 2016 or 2017, engineering students barely did competitions. Grades and senior thesis projects were considered far more important. But he suddenly got hooked on competitions and threw himself into everything from advertising contests to volunteer groups. He described himself as aiming wide and shallow rather than deep.

Bouncing around like that, he started to think that founding a startup might be the ultimate résumé item. In 2018, almost by accident, he was selected in Naver D2SF's student startup competition (Naver is Korea's biggest internet company, and D2SF is its startup program) and got office space near Gangnam Station plus a stipend. It was private support with no equity, so it was closer to basic seed support than an accelerator. The idea was digital healthcare: detecting schizophrenia with mobile devices.

I realized I had no expertise, so I went to a VC

The lesson from his first startup was that he had far less expertise than he thought. While serving as a social service agent (an alternative form of Korea's mandatory military service), he thought about what expertise you need to start a company. If he was going to start one, he'd be raising money, so he figured he should understand the investment ecosystem. He applied for an internship at Bluepoint Partners (a Korean early-stage VC) and got in.

He supported the investment managers and stayed about half a year. The firm gave interns real opportunities, and he was one of the few interns to bring a portfolio company to the investment committee. That deal got dropped, but he connected the company with another firm, and it ended up raising money there. He wasn't trying to build a career as an investor. He chose it because he thought it would help him understand the funding ecosystem when he eventually jumped back into starting a company.

We pivoted the day after ChatGPT came out

Next came Smoretalk, where he spent four years including the pre-founding period. The mission was to solve mental health problems with AI. It started in 2022, before ChatGPT.

When ChatGPT came out, the team saw a crisis, not an opportunity. They got scared that it would do everything they did, and they pivoted that same day toward images. He remembers the date exactly because they pivoted on December 1, Korea time. Smoretalk now does media AI for advertising, and he left around that point.

ChatGPT, Manus, and Opus 4.5

I asked whether today's AI capabilities arrived early or late. He said they came at the expected pace, but that expected pace already reflected enormous expectations, so in the end it really did come fast. It's fast, and the quality is almost scary, and he thinks the gap in how well people understand that is huge right now.

When I asked him to point out the moments that built momentum, he walked back through the turning points of the third AI wave. First the ChatGPT launch, then Manus. The ad showing an agent running Instagram on its own on screen was a real shock, and he wondered whether it was already here. When Claude Code came out, only certain people were using it. For him personally, the big one was the release of Opus 4.5.

Up to that point, people said we were at the stage where you can delegate, not the stage where you get replaced. From 4.5 on, the people ahead of the curve started thinking, this is really serious, an era just changed.

Non-developers started using it in earnest this January, and even heavy users were shocked from then on. News about Claude Code started showing up in stock prices, and FOMO hit people all over again.

Next is when companies use people as APIs

When I asked how the fourth boom will come, he said it seems to have already started. It could take the form of personal assistant services, but the real marker is when companies handing work to AI becomes normal. He means an era when handing work to AI end to end is taken for granted at ordinary companies, beyond the early adopters.

From the point when a person becomes an API that uses a Claude or a GPT. Companies will use humans as APIs.

His evidence was how companies do the math. If companies thought AI wasn't useful, they wouldn't bother optimizing costs. They'd just give up. Instead, with Chinese open-source models now performing well and getting cheap, they've started optimizing for performance and price on top of them. Companies never do something they don't need, so if they're doing it, it's because they need it. That's why he sees AI infrastructure as having become a core company asset. On the social side, he added, the bigger impact will be companies hiring fewer people.

I thought B2B was the hardest place for an individual to go

The hypothesis he set when he went solo was how far one person and AI could grow together. He wanted to see how much one person could really earn alone and how much influence he could have on society.

His read was that one person might actually earn more with a consumer app than with B2B, but that kind of story doesn't travel well in the B2B world. It becomes a hot topic at the B2C level, but executives at companies don't feel it. So he wanted to make his name properly known among senior people at Korean companies too.

The reason was proof. If he met senior people and actually got hired to run a project, that would count as a proven task. And he believed the people who move society are those at the top of those organizations. His calculation was that if he rose to being someone with business influence, not just an entertainer, he could try a lot more later.

The fact that this is the hardest place for an individual to break into was itself the reason to choose it. It's very hard for young people like him to get to that level. If he could do it with AI, it would be a case of AI making possible something that otherwise couldn't have happened. He was putting himself up as the test bed and checking his hypothesis under the toughest conditions.

When I asked for a four-month check-in, he said his expected revenue for this month and next had gone up a lot, but he held back on giving specific numbers. Instead he summed it up like this:

I thought I'd starve, but I didn't. I made the right call. And honestly, I'm actually making pretty good money.

What companies hand to one person: education, events and ads

Still, he thinks the areas where one person is allowed in are fixed. Officially, what companies trust a single person with is education, events and advertising. Advertising is obvious, since it uses the person's own channel and a company can't do that itself. Events and lectures could be handled in house, but companies hand them over because giving them to someone influential is lower risk.

But projects are really hard. Unless there's a lot of trust, it doesn't really happen.

That's also why he teaches. Advertising doesn't lead to follow-up projects, but education and events have participants, and the conversation keeps going, which creates openings to build trust. He was using that as his first door into companies.

The number of applicants for freelance gigs is unreal

He thinks it's very hard for someone just starting out to land work in the same position he's in. Usually you end up on a freelance site like Wishket (a Korean freelance development marketplace), which also asks you to upload a portfolio. The development market is in such a slump that the number of applicants per posting is unreal. Developers who already did freelance work and newcomers all crowd into the same postings, so you probably won't even make the first cut. To get past that, your portfolio has to be seriously good.

The work that comes to him comes from the network and media presence he's built up, not his skills. Same market, completely different way in. One side applies to postings and competes. The other gets inbound calls. AI narrowed the gap in the ability to build, and the gap in front of that got even sharper.

No investment, no government grants

He hasn't taken investment and has never even applied for a government grant. His goal was to hold out, earn and survive purely on private money, and he said he was relieved to start earning sooner than expected.

His reason was that this is the only way to play the game he originally wanted to play. He wants to build one piece of software that has impact, and he's realizing very clearly that you can't do that if you're thinking about making money right away. So he plans to keep expanding features for free. But if you just say something is free, people don't feel any scarcity and don't use it, so figuring out how to get the word out is a whole other challenge. The picture he has in mind is that one day, something you're using turns out to be something he made.

He doesn't much care about the field as long as it works in the market, and right now he's operating fully for profit.

I thought software alone would fail

When I asked why he looks at so many overseas examples, he started with the conclusion. AI is so good at making software that he thought software by itself would fail. Nobody, company or individual, can grow by just doing software. That's when he started thinking much more about sales, about how to sell before thinking about what to build.

His case for YouTube as a channel also came from overseas. Watching Riley Brown and other YouTubers, he saw a structure that's already working: developers gaining influence by being good at YouTube, rather than selling courses on lecture platforms, and AI companies reaching out to them first because of it. Education is closed, so it's hard to get broad recognition from it, but it lets you build one-on-one relationships with companies. If you depend only on that, you're basically employed by them, so he figured he had to build public influence at the same time.

The real reason Koreans don't do YouTube

He told a story about when he said YouTube was a red ocean and his younger sister pushed back. "Do you know any YouTubers?" "No." "Then it's a blue ocean." So I asked him why he thinks people don't do YouTube.

His answer: it isn't the gear, and it isn't being too late. People just can't stand putting their face out there. People with day jobs especially have a lot holding them back, from having to report a side business to how they'll be seen as an employee.

Koreans are pretty shy, right? But there's also so much we want to do.

So right now, there's a part of YouTube where embarrassment is the bigger barrier, which actually makes it a good market to take a shot at. Later, when everyone is doing it, regret will outweigh embarrassment and competition will get much fiercer.

The real danger for a solo builder is loneliness

Working alone seemed like it would be hard mentally, so I asked who he uses as discussion partners. He thinks a lot on his own, talks with AI, and occasionally talks with other builders he works alongside. His default is to try something rather than spend time worrying.

The risk he pointed out for solo builders is that it's hard on people who get their energy from other people. He said it was strange to go from talking a lot to suddenly barely talking, and he hit an emptier stretch than he expected. He added that it's probably better to have something that eases that human emptiness and loneliness.

Tokens are becoming money

One trend he's interested in is tokens. He pointed out that the word "token" is the same as in web3, and he's been thinking of this as a version of web3 that's actually helping the world.

The example he gave was Stripe acquiring OpenRouter. Payments and AI are completely different fields, but the moment you define AI tokens as currency, it makes sense. Then an era where you invest in tokens or get paid your salary in tokens may seem unrealistic now, but it may not be impossible. The catch is volatility. Web3 had value when it was stable, and people are scared of it now because it isn't. Tokens could also get suddenly much cheaper or much more expensive.

He brought up the movie where time is currency ("In Time") and pointed out a difference. That movie has one currency, but there are many models, so you get exchange rate gaps between tokens. Just as manufacturing moved from the US, where labor is expensive, to China, the gap between Chinese and American models could create differences in what you're able to do.

Personal AI assistants are another area he's interested in. When I asked how far it has to go before you can call something a personal AI assistant, he said most people have never had an assistant, so they don't even know what to ask one to do. People who are good with assistants hand over everything from scheduling to all their work, sometimes even housework. Reaching that level is the bar.

The only plan until year end: don't belong anywhere

When I asked about his plans for the rest of the year, he said he made just one this time: don't belong to anything until the end of the year. Stay solo.

He also named the trade-off. Without a goal, he isn't really driving hard toward anything. He's handling what's in front of him well and building trust that way, but with no short-term plan, he's aware he isn't heading anywhere in particular. So he's trying to split his time without forgetting his original intent: building software everyone uses.

Is it hard to survive without a personal brand?

I asked what he thinks of the extreme claim that you can't get by these days without a personal brand. He said it depends on how you want to live. If you want to run your own business and already have a strong network, you can do well without promoting yourself. If you have no network at all, the important thing is to build something, and that's when using channels becomes really important.

By that logic, it's no different from before. Back then, if you were in an organization, your business card was your brand, and if you weren't, you had to build one yourself. The difference is that when jobs were plentiful, it was easy to build a network through your company. Now some people have no company and nobody building it for them. In exchange, the barriers are lower. Editing is easier, and you can even plan content with AI. Digital reaches everyone, and someone is watching YouTube right now, so if you're active there, you scale up naturally.

Stop hesitating and just do it

At the end, I asked what he wanted to say, and he said this is an era where if you have an idea, you should just do it. His evidence came from his wallet. To buy the same MacBook he bought when he quit his job, he'd now have to pay about 2 million won (about $1,400) more because RAM prices went up. He'd written down a dozen or so things under the label "2026 sunk costs," things he wouldn't have lost money on if he'd bought them back then. The list ran from AI subscription plans whose prices jumped to machines to travel.

Don't hesitate. The world is changing too fast. Efficiency isn't what matters anymore. Just doing it first might turn out to be the cheapest option.

The same goes for action. While you hesitate, someone else is doing it, so if you're going to do it anyway, do it now. It isn't easy and competition is fierce, but then you have two choices: join a good company and be truly happy living that life, or do it properly. Once you're unhappy, it comes down to whether you act. If you acted, that thought was worth having. If you didn't, it's a pointless thought that only wears down your mental strength.

When I asked about his AI subscription spending, he said the software he built tallies it at $1,800 a month. He corrected himself twice: that's at list price, and for various reasons he doesn't actually spend that much. He also gets discounts, so his real spending is lower.

We'll meet again when he hits 100 million won a month

We closed with a promise. He hasn't hit 100 million won (about $70,000) a month yet, so we'll talk again once he does. The money matters, but the number is symbolic because it's the point where his hypothesis comes close to being proven.

There are already plenty of people who can build. Now the question seems to be less what you can build and more who knows you before you build it.

Originally written in Korean on Brunch, September 17, 2026.