The Reality of AX, From Korea's First FDE Company

We finally started a Bloom podcast. Our first guest was Hyuntae Hwang, CEO of SpaceY, who has been testing and spreading the FDE (forward deployed engineer) model in Korea.

Working with companies like LG Electronics and Kiwoom Securities, he pushed this unfamiliar job title into the Korean market. SpaceY was recently acquired by Day1 Company (a Korean online education company), and he's starting his next chapter. We sat down on a Friday morning and talked for 50 minutes. Here's a summary of that conversation.

SpaceY, because we like SpaceX and Y Combinator

When I asked where the name came from, he said they liked SpaceX and they liked Y Combinator, so they went with SpaceY. The company was founded in February 2022, which makes this its fifth year. They picked the name first and then kept throwing out ideas.

What got him to start a company was a person, not an idea. He wanted to work with his co-founder, CTO Songhak Nam, and to work in that kind of environment. He said twice that if Songhak had said no, he would never have started the company.

From Middle Eastern gems to a Seoul R Real Estate

The hot method at the time was Alberto Savoia's pretotyping, and he got hooked on it and kept throwing ideas at the wall. He had his own framework that split problems into painful but rare ones and mild but frequent ones. With that, he went through selling gems from the Middle East, building a social dining platform and opening a counseling service. He even got as far as reading "Tokyo R Real Estate" (a Japanese book about an offbeat property agency) and wanting to build a Seoul version.

Around the sixth or seventh try came DIO, a hiring platform for senior developers. He described the whole process as finding founder product fit: however good an idea is, it has to be something you like and are good at.

A mechanical engineering major in a coding lab

His background isn't the usual path either. He majored in mechanical engineering at Yonsei University, but his lab worked on knowledge-based design and did a lot of coding. He coded all through undergrad and went straight into a PhD, right around when AlphaGo came out. He told his professor he wanted to focus on AI, and he ended up researching model evaluation and safety with a Dutch researcher who is now at DeepMind.

People who do a PhD usually become researchers or professors, but he didn't have a particular dream back then. Military service was part of it (in Korea, graduate research can count toward it), and he also just enjoyed the lab culture. An engineering lab can mean more all-nighters than a startup, he said, but you can do it because there's a deadline.

An older friend twenty years his senior suggested a tax startup

When he was about 23, an older friend from grad school suggested starting a company. This friend was twenty years older, and his pitch was that AI could change tax work. So he became a co-founder of Hyeum (a Korean AI tax and accounting startup) and spent four or five years there.

He looked back and said he didn't really do the co-founder job well. He followed his friend around, and his friend carried all the heavy responsibility, money included. Even so, he said he learned everything there: how to work, and the mental models and frameworks for running meetings. He thought he'd stay longer, but as he grew up, he wanted to build something of his own.

The way we define FDE starts with finding our client's customers

When I asked him to explain FDE in his own words, the scope was wider than I expected. Forward deployed engineer literally means an engineer sent out to the front line. At SpaceY, an FDE does all of it: finding the client's own customers, working with the client to pin down their vague problems, finding a solution, building it and deploying it at the client.

He said the most important thing isn't technical. It's a phrase the team has had since the DIO days.

Betters. The desire to work with people who are better than you. FDEs have to tap into that desire in the client.

That's why he said it doesn't work if developers just change their job title. Clients are cold, so they can never be left wondering why they're bothering with you. So I asked how SpaceY does it.

We just work hungry. Sometimes we're not good. We make a lot of mistakes. But we work hungry.

Big companies reach out first

SpaceY's clients range from LG Electronics and Kiwoom Securities to insurance companies. There's no fixed domain. He laughed and said they're still a hungry startup, so they'll eat anything. What I found interesting was that large companies you'd expect to be conservative were the first to open the door. Because the FDE role is still so new in Korea, the people looking for SpaceY right now are early adopters, whatever their industry.

All of their sales are inbound. Some come from YouTube, some from LinkedIn, some just by email. That's also how we got to know Hyuntae. I knew DIO but not the person behind it, until a colleague watched his YouTube videos and kept sharing them in our Slack because the insights were good. I saw how he could pull logical answers straight out of his head, sent him a DM on LinkedIn, and he talked with us for almost an hour. That led to us running an event together.

Honestly, I don't always know what the client really wants

When a company reaches out, the first thing he asks is why they came to SpaceY. Most companies that come to them have more manpower, more scale and more domain knowledge than SpaceY does, so he wants to find the core reason they're going outside anyway. Maybe they need a product. Maybe they need to shake up the mood.

When I asked how he figures that out, he said he doesn't really know. He keeps asking, meets them often, and sometimes only realizes in the middle of the work, "Oh, so this is what it was about." So he settled on the straightforward approach: dig up the problem, focus on what the core problem is, and study the domain hard.

Their approach changed within six months. In the first half of the year they'd prototype fast and show the client, "Is this the picture you had in mind?" Now clients have gotten more sophisticated and tell them up front what they want, so the team focuses on sharpening that problem.

Startups have no enterprise development experience

He was the one who brought up mistakes first. He named two things: fumbling because they can't absorb a client's domain knowledge fast enough, and the fact that startups have almost no experience building enterprise software.

His take on the second one stayed with me. The generation now working at startups grew up hearing that if you join an SI company (Korea's big IT systems integrators that build custom software for corporations) as a developer, your career gets derailed. So they've never coded inside a company's internal or air-gapped network. They never had that experience passed down to them, and now that they're trying to do enterprise work, they fumbled at first.

They're learning two ways at once. One is by running into things. The other came from a client who told them to sync up with an established large SI firm, so they worked alongside one at a securities company and learned from it. At first they didn't even know what they didn't know, so they didn't know what to ask. Now they ask as they go.

He was matter-of-fact about the client relationship too. The client is allowed to make mistakes, and SpaceY isn't. Even so, clients communicate quickly and tell them a lot to help the project go well. Most clients are at least five years older than him, so apart from engineering skills, he gets an enormous amount of help from them.

Korea's AX is already in its fourth generation

When I asked why big companies work with SpaceY, he answered with a generational theory of Korean AX (AI transformation). In the first generation, each affiliate company had its own AX team, with a core AX team solving problems for the business units. The old IT teams only had to solve a few problems a year, but AI was different.

AI moves so fast that if you really set out to solve things, you end up with hundreds of tasks, and that's where the overload hits.

So the second generation shifted toward education, so the business units could build those skills themselves. People say, "Can't everyone just get good at Claude?" But learning outcomes vary a lot, so that alone wasn't enough. The third generation tried bringing in startups, regardless of domain, that had small teams, decent revenue and were well optimized, to get hints for the enterprise.

The fourth generation, where things are now, is about leaders needing to become AI native. It hasn't been long at all, and it's already the fourth generation. That shows you how fast this market moves.

Build 4,000 agents and maintenance becomes impossible

What led him to the fourth-generation diagnosis was specific. He saw a group company that had run 4,000 PoCs and built 4,000 agents. When you trace why that many agents get built and why they become impossible to maintain, he said, you arrive at leaders who have no philosophy about being AI native.

So he said you have to reduce the maintenance burden and have dedicated maintenance roles. He mentioned the maintainers, builders and sweepers idea that came out of Anthropic. Most organizations focus on prototypers and builders, but org design needs to put weight on sweepers and maintainers too.

Along with org design, you need a philosophy for how you treat agents.

You have to go with the philosophy of making good use of the general-purpose agents you already have, instead of constantly trying to build new agents.

Power the agent, instead of build the agent. If you take the mindset of attaching one agent to every task, like a gas hazard assessment automation agent, a freelancer matching agent and a public bidding agent, he said, you get a catastrophe. He added that this isn't someone else's problem: he runs a freelancer platform, and the host that day (me) was building a public bidding solution. The structure he has in mind puts a universal agent in the middle and manages each task as a spec. The reason is simple. The work keeps changing.

Why you still need FDEs as models get better

I asked what happens to the FDE role once people on the ground can do everything in a couple of clicks. His first rough thought was that there would always be work only humans can do. Then he corrected himself and said that was the wrong answer.

Enterprises will always need an outside force. I think that's the real answer.

Even in an organization with optimized business processes, there are things internal motivation alone can't produce. Whatever version Opus is on, the FDE's job is to keep finding those gaps and filling them.

Trends really have shifted in six months. In the first half of the year, everyone was interested in work automation. Bring up work automation now and it might sound like old news. As clients get more sophisticated and models get better, he said he keeps feeling that SpaceY has to keep changing too.

Some companies have replaced their AX department with an FDE department

I'd seen an Instagram ad from a company saying they had shut down their AX department and replaced it with an FDE department, so I asked why big companies are announcing plans to train 100 FDEs. He said there's definitely a branding element. In Korea, "dispatched developer" comes with a certain image. FDE gives people a bigger role and a vision as a career. If you want to hire 100 people, you need at least 100 applicants, so it was probably a way to widen the pool.

He also touched on how the role was received in the US. There, it felt like engineers were finally getting recognized for their value. In Korea it has been localized a bit, and every company has a different idea of what an FDE does. At an FDE policy roundtable he attended recently, every company gave a different definition. He wondered whether it even needs to be unified, since that's where the edge and the differentiation come from.

Up at 6:30, starting with what's piled up

When I asked about his daily routine, he said he gets up between 6:00 and 6:30, washes up, and turns on his phone right away. Since starting the company, he can't stay asleep in the morning, so he always wakes up early. He starts the day by checking whether anything big happened while he was sleeping. He said he often goes to bed with a vague, nagging feeling that he didn't finish everything.

After he's cleaned up his email inbox and turned it into material to work with, he heads to the office at Orange Planet (a startup space) in Yeoksam, Seoul. Sometimes he goes to client sites. Mostly he checks in with and talks to the FDEs. The quality of FDEs can't vary, so he communicates a lot with the core team. Out of a team of a dozen or so, about six are FDEs and the rest are people who don't go out. These days job categories are blurry, so they just call them the people who go out and the people who don't.

I don't have scale-up experience

When I asked about the story behind the Day1 Company acquisition, the first thing he talked about was his team. It was a big decision, so he discussed it a lot with them. They welcomed it hugely, even more than he did. He explained why in his own words.

I have my limits. I don't have scale-up experience. I'm someone who builds a great little restaurant.

The FDE model is still emerging, so they needed insight all the more. They might need money, scale and a brand, and when a good offer came in, they gained a strong backer. He said he was certain that if AX, or AX through FDEs, is a megatrend, now is the time to move hard. Exactly what role he'll play once the two companies fully merge is something he's still figuring out.

Korea is built to make 99% very fast

I asked what he thinks about being called "the Palantir of Korea."

Korea is a Palantir country. It's really specialized in making 99% extremely fast, rather than making 100%. That's why we're getting so hooked on FDE, and on AI as a technology.

So SpaceY and other FDE companies might not scale this up hugely, he said, but a lot of boutique FDE companies will appear. Young people today are wary of getting a regular job because they feel they won't be fully recognized for what they're worth. His prediction was that the people capable enough to earn that recognition will go outside and do FDE work, and he added that he's one of them.

That's also why one of the conditions he weighs heavily when hiring FDEs is startup experience. Having started a company means you know the pain of it. In B2C, starting a company means building a platform right away. In B2B, it can take this form instead, and he sees that as another kind of startup culture that takes on challenges.

Enterprise and the public sector still need SI, because building a whole engineering team every time a pharma or finance company wants one thing done is inefficient. But that lineage got cut off once, because people believed joining an SI company would derail your career. His picture is that by carrying it on under the name FDE, you can create much more value far more efficiently than in the past.

I want city hall websites to be as good as Toss

The last thing I asked about was his vision for the public sector, and it started from a sense of generational responsibility. If you're from the generation that learned engineering and IT on top of the high-speed internet of Korea's IT-powerhouse days, he thinks you should feel responsible for remodeling and rebuilding public service IT, which is now old and often mocked. Companies in the market survive on their own, but the public sector isn't competitive, so it can't help but develop slowly.

When he started SpaceY, this was exactly the order he talked through with the FDEs: we want to make city hall websites as good as Toss (Korea's most popular fintech app), but nobody knows our name yet, so let's make a name for ourselves in Seoul first. That's how they started doing enterprise work.

He mentioned how an earlier generation digitized Korea's real estate registry. They spent four years traveling around the country setting up kiosks, printing out every registry record and labeling them. He said he'd like to do something like that once. He doesn't know if he can, but if he has the capacity, he really wants to try.

One goal for the second half: the renewal rate

When I asked what would make him proud by the end of the year, he said he has exactly one KPI: keeping the renewal rate on current projects at 90 to 100 percent. He also said he'd like to be able to take on all the work coming in, which would take more capability and a bigger team.

But the more enterprise the work, the more quality matters, so he has set a rule not to take on work they can't handle. He always keeps in mind that one person does one thing.

Speed is the strength and the trap

The line about Korea making 99% very fast stayed with me. That speed is a strength right now, but it was probably the same speed that produced 4,000 agents.

The next step for AX looks less like finding ways to build faster and more like deciding what not to build. The fourth-generation diagnosis that leaders need to become AI native and the suggestion to put a general-purpose agent in the middle were pointing at the same place.

We wrapped up by agreeing that I'd ask him again in three or six months how SpaceY has changed.

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