From Claude Blue to Claude Bloom

Conversations with an IBM developer in the UK and a reporter at a Korean daily newspaper

After I published my Claude Blue post, coffee chat requests started pouring in from places I never expected. (I have no idea what people think I am. I'm just a regular office worker.) The very first request came, out of nowhere, from a developer at IBM's office in the UK. This developer works in an AI department and told me that since the company's focus recently shifted from AI to quantum, a lot has changed internally, and they've been thinking hard about their career. These days they and their non-Korean coworkers complain so much that it feels like life is split into before GPT and after GPT, and my post hit home. They even shared my Brunch post with an American coworker and said to read it through a translator. (I hope the translation held up.)

That same afternoon, a reporter from a Korean daily newspaper came to my office for a coffee chat interview. (The reporter messaged me on KakaoTalk, Korea's main messaging app, over the weekend, and we set it up right away.) This is a senior reporter with 23 years in the business who leads the paper's in-house AI team. It was a topic they'd been thinking about for a while, so they wanted to know how people actually feel about it on the job, and especially how developers really use AI well. When I described a few companies, the reporter kept asking me to go over things again to be sure they understood. "Wait, people really work like that?" "That actually works?" Inside the paper, they've been pushing for AI pretty aggressively, but it hasn't caught on, and AI use there is still tightly restricted. (Of course, everyone is probably using it on the sly.)

What struck me is how far apart developers and non-developers are. Non-developers already lean on AI for a lot of their judgment and thinking, too. Most office work is research, cleaning up data and turning it into reports, and hardly anyone these days goes to Naver or Google, looks everything up by hand and writes the whole report themselves. We already depend heavily on how GPT or Gemini thinks, and whether we notice it or not, we're turning AI-native. So I keep wondering why the gap between developers and non-developers is so big. Developers have gone past awe and into fear, while non-developers talk about AI as if it were a different world, still stuck on the novelty of the early days. Maybe it's because they haven't felt for themselves, or even heard, how much AI can really do.

Later, I got on the phone with a friend who works as a PM at a US startup, and my friend had an interesting take. Developers weren't hit first just because AI is good at writing code. The way developers work, putting up scaffolding first and then filling in the details, setting the big structure and then layering on specifics over and over, is basically the same way an LLM works. That's why developers were the first to feel how powerful AI is, and also the first to feel threatened. Office work is slower to adopt AI because it hasn't been structured that way yet. AI can do it. The moment it gets structured, the same thing will happen.

Anyway, the point of today's post goes beyond the stories of people who've already run into Claude Blue or are living through it. It's about how each of them is actually responding, in very different situations.

1. How the UK Feels About AI Compared to the US

The conversation started naturally with the tech scene in the UK. Finance is London's core industry, so the UK is strong in fintech, but the developer said the US is clearly the center of frontier AI. DeepMind started in the UK, and so did the Industrial Revolution, but even when the UK starts something, it always moves to the US to scale. Since all the Big Tech headquarters are in the US, the important work and the frontier projects go to the US teams.

I once met another Korean developer working in the UK who said something similar. The UK's visa system is relatively open, so a fair number of people go there as a gateway into the English-speaking world, but everyone agreed that if you want to be at the frontier of a tech career, you eventually have to go to the US. Even in an age of remote work, meeting people inside that community, connecting casually with the inner circle, and the information that gets shared there are things money can't buy. Korea is a fast follower, so we study US trends closely and keep up with them. Because of that, there are probably many things we know about America better than Americans do. (We always learn it from books.) But putting our heads together doesn't always produce the answer, and often the experience of being on the ground counts for more.

As much as people dream of going to the US, you only have to watch the news to know life there isn't easy right now. Immigration policy is very unstable, there are the ICE issues, and living there as a foreigner on a visa is hard. The IBM developer is looking for a way to join a US startup on an O-1 visa and said the hardships ahead are painfully clear. Still, in the AI era, where you physically live is becoming a survival strategy, so they said it's worth going through.

2. A World That No Longer Needs App Interfaces

The developer's vision of the future was interesting. In their view, the very idea of building software for humans to use is already outdated.

About a year and a half ago, when Cursor first came out, there was hope: "Oh, I could build software all by myself." Now things are completely different. If we get to a world where a personal agent takes your data, visualizes it for you, and suggests what you need before you ask, the app as an interface stops being necessary. Once security is solved and the agent fully understands your context, you could mutter "I should work out," and the agent would suggest a workout program on its own.

Like Jarvis. An age when everyone walks around with their own Jarvis. Later we won't even need phones. Just one earbud in.

Their sense of threat as a developer was more concrete. If an agent can visualize data and surface insights whenever the user wants, doing all of that on the user's behalf, those interfaces aren't needed anymore. A world where you only need one device, like in the movie "Her". Maybe not even a device.

My company is wrestling with something similar. People have started calling it the SaaSpocalypse. As LLMs began delivering directly the value that software services used to provide, existing clients started saying, "We'll try building this in-house." At the end of last year, I took that as an excuse not to renew the contract. But if they don't plan to build a product to sell and only need something that fits their own needs, they really might try it. So I've started wondering whether software companies should dig deep into fields like law, finance and healthcare, where the legal risk is high enough that LLM companies would rather not touch them. It's like we all have a nuclear bomb, but someone still has to press the button.

Another founder I met the same day showed me an interesting case. This founder is building Sajuping, a Korean saju AI service. (Saju is traditional Korean fortune-telling based on the date and time of your birth.) The founder's logic was fun: you should look for areas LLM companies "won't" do, as opposed to areas they "can't" do. If an AI predicts the future, tells you "you should choose this path," and turns out to be wrong, that's exactly the kind of legal risk LLM companies never want to take on. But if you do it through the frame of saju, it's a different story, because there's a filter: that's what the saju said. It's also something that can go global on the wave of K-culture. According to the founder, a Korean saju variety show recently came out on Disney+, and overseas users jumped sharply. We're in an era where solving big problems has become the risky move. The LLM companies will solve the big problems, so startups have to find the areas those companies avoid because of risk.

3. What Can't Be Replaced

The IBM developer told me that wherever they go these days, they run the numbers in their head on whether each job can be replaced. At a coffee shop, they look at the barista and think, "Will that job get replaced?" Walking past a church, they think, "The pastor probably won't be."

But their conclusion was that even that isn't simple. If AI replaces everyone, people will flock to physical work, and if they pile into jobs like yoga instructor or plumber, competition will eventually drive prices way down. In Korea right now, young people aren't going into plumbing, so plumbers reportedly make 350,000 won (about $250) for 40 minutes of work. But what if AI pushed all the young people into plumbing? The scarcity of physical labor we count on can only be temporary.

I added a hypothesis of my own. I once wrote on my blog about this: technology replaced physical labor, which raised the value of knowledge work, so if LLMs are now replacing knowledge work, what comes next? My hypothesis was "the age of mass persuasion." Leadership that can move a lot of people, the ability to set a direction, people who carry some kind of authority. I thought people who can gather a crowd, like celebrities or influencers, might be in the only field left standing.

The developer's reaction was funny. "I agree. Lately I've been thinking, 'I need to get famous.'" They had gone to a talk in London the day before by Sora Unni, an influencer known for TikTok and a dating reality show. It was about women's empowerment and personal branding, and the conclusion was the same: whatever you do, you have to become known.

Everyone thinks they should become known, but very few people actually do anything about it. This developer said they can't even bring themselves to write a single LinkedIn post because they worry what people around them will think. To be honest, I didn't post on LinkedIn at all when I worked at a big company. I didn't want to talk publicly about what I did at work, and frankly there was no need. Then I joined a startup, and suddenly I had to get the word out about the company myself. As I kept writing, I started meeting people and my network grew. This coffee chat only happened because I wrote a blog post. Writing brings people, and people bring insight. I remember something my younger sister said: "Everyone says any random person can be a YouTuber, but do any of your friends actually have a YouTube channel?" None of them do. Maybe it's still wide open.

What's interesting is how differently people react to the same thing. My friend who's a PM at a US startup had a completely different tone.

Yesterday I read an article saying even Mark Zuckerberg is running OpenClaw. So he's replacing himself, right? If Zuckerberg can be replaced, why not me?

My friend said it as if it were just the natural course of things. Even if AI automates everything, what's left in the end is judgment and taste. Citing Jeff Bezos's principle of investing in the things that won't change, my friend said they're spending their time over the long term on fundamental areas like food, clothing and housing.

This PM uses AI deeply at work. A favorite tool is Perplexity's Model Council feature, which runs the same question through GPT, Claude and Gemini at once so you can cross-check them. Instead of taking one model's answer at face value, my friend compares three answers and decides where they agree and where they differ. No blind reliance on AI. It's a thinking partner. When my friend makes a report, AI lays down the first step, and then my friend fills in and polishes it. The starting point of the thinking goes to AI, and the judgment stays human.

My friend also brought up the example of someone who studied music and then moved into business, and said, "Life was never linear." The creativity and sense you develop through music can help later when you build products. Nothing from your past is wasted. It all connects somewhere. The point was that there's no need to fear jobs changing and careers shifting in the AI era, because what you've done so far can create value in a completely different form.

Nobody here is right or wrong. But facing the same wave, some people feel an existential crisis and others calmly pull out a surfboard. Some say they'll go learn plumbing, and others say they need to get famous. That responses differ this much depending on each person's experience and temperament is one of the more interesting sides of this era. (Honestly, that last line sounds a bit too much like AI wrote it, but I liked the metaphor, so I'm keeping it.)

4. People Move Only When Something Triggers Them

For a startup, survival itself is the trigger, every day. When a new tool comes out, you try it, whatever it is. That's how it went at my company. One person used AI to knock out, alone, a large enterprise project that would normally take ten people. The AI ran 24 hours a day and the output kept coming, so even when this colleague wanted to rest, there was no chance. After four weeks my colleague burned out and had to take a week off. The AI could keep going, but the human was the bottleneck. Running things solo with AI is possible in theory, but this showed a real transitional problem: human stamina and mental strength can't keep up with that pace. I think we'll be working on ways to compensate for that for a long time.

Big companies, on the other hand, have no trigger. They have plenty of resources, they already have lots of human agents, and they have no reason to change. But I recently heard an interesting story. A reporter who had worked at a news outlet for a long time saw a trusted editor leave the company and afterward asked whether someone could build an AI agent trained on that editor's editing style and standards. Once the filter that used to oversee their work was gone, the reporter finally felt the need for an AI agent. That's a trigger. People only move once they run into it personally.

Developers already feel the trigger every day. They watch AI write code, the core of their job, better than they do. But most of the office workers I'd talked to didn't seem to feel that trigger yet, which made the gap feel even stranger. The real question is whether we'll be ready if that trigger hits us harder, and at a moment we didn't see coming.

5. Choices After a Great Hardship

Near the end of our conversation, the developer carefully brought something up, saying, "It's something I can say because you're a stranger." A year and a half earlier came a cancer diagnosis, the hardest thing they had faced. After difficult treatment they're doing better now, but there's no knowing when it might come back. During treatment they wrote openly about it on Brunch, and connecting with people who'd been through something similar gave them a lot of comfort and strength.

There's a lot that scares me. Even after getting through that, other things still scare me.

That line stayed with me. Someone who has felt in their own body that life can change at any moment is bound to feel the uncertainty of the AI era differently. They beat something bigger and can still say honestly that they're scared. I think what matters is whether you find a way to keep moving forward inside the fear, more than whether you feel fear at all.

Looking back on the day after these conversations, one thing became clear. Everyone who related to Claude Blue and reached out to me was either a developer or an early adopter already using AI heavily. Sadly, I heard almost nothing from office workers. Even the reporter from the daily paper, who leads its AI team and came out to interview me in person, didn't seem to fully grasp the AI workflows developers describe.

This developer from the UK, a PM at a big company in New York, an engineer at Meta in Silicon Valley: in the end they were all saying the same thing. Across borders and across job types, everyone is feeling the same anxiety. And the gap between people who feel that anxiety and people who don't is becoming the most dangerous divide of this era.

I plan to keep meeting people on the ground. In particular, I want to hear from the developers who use AI most deeply and translate their insights into language non-developers can use.

I don't intend to stop at Claude Blue. We have to move past Blue and into Bloom.

I'm not an AI expert, but I'm always up for a coffee chat. If you'd like to share your story or hear more about other people's experiences, reach out on LinkedIn.

Claude Bloom means moving past Claude Blue: each of us growing and adapting to the AI era in our own way.

Originally published in Korean on Brunch, March 24, 2026.