Claude Bloom × Claude Community: A Night with 170 People

An AI engineer, an entertainment tech critic, a philosophy YouTuber and a media studies professor. Speakers from very different fields, in one room.

The Claude Bloom event is finally over. The original goal was to use the Claude community to put the ideas of Claude Blue and Bloom on the table officially for the first time. The response went far beyond what I expected. Within a few days of opening registration, more than 2,000 people signed up. Space limits meant we could only take 170, and yet there were almost no no-shows, and most of the audience stayed until the very end.

Why was the response so strong? I don't think it was because I'm good at marketing. I think a lot of people already shared the feeling, and it lined up with what this moment needs. Talking with attendees after the event, I found that the more unverifiable information floods the internet, the more people crave real, in-depth conversation and, above all, genuine human connection.

Since it was our first event, there were things I'd do differently. We had so many messages to get across that we couldn't give the speakers enough room, and there wasn't enough time for attendees to network with each other. People who wanted more kept talking in the hallway after the event, and even after the venue closed its doors at 11 p.m., there were groups out on the sidewalk still debating. The operations were rough, but in that scene I could see the possibility of a real community, and that felt good.

I started thinking about how to improve future events. Talking directly seemed better than a written survey, so I offered a 15-minute remote coffee chat to anyone willing to give feedback. To my surprise, 15 people signed up the next morning alone. I do worry about how much I can personally handle, but for now I want to hear these voices and their ideas directly.

If I plan the next gathering, I want to try a different structure. Some fairly ambitious formats I'm considering: splitting a big topic into small groups of six to eight people sitting in a circle for denser conversations, meeting at a casual pub instead of a stiff seminar room, or a format like Sseuljeon (a Korean TV debate show) where teams split up and argue for and against a topic.

Honestly, Claude Blue and Bloom are just buzzwords that will fade someday. But the reason these words caught on is that I didn't invent them artificially. They captured something that's really happening around us right now. So I want to keep talking with people in Korea and abroad and keep finding new phenomena that matter. I also hope the energy of this event spreads beyond Korea to Japan, Singapore, the US and Europe, and that our Discord community grows into a real community. I can never get there alone, so I'd love a lot of help from all of you.

The event was led by me (Patrick Han) as moderator together with Claude ambassador Hoonmin Choi, and four speakers added depth: Jimin Park (AI engineer at Cliwant), Woojin Cha (entertainment tech critic), Choongnyeong Lee (philosophy YouTuber and author), and Professor Hyerin Shin (Korea University School of Media and Communication). Below is my record of the conversations from the two fireside chat sessions.

A lot of people helped make this happen. Anthropic took an interest in the meetup and supported it, and BizCafe, a business content outlet, joined us. Cry Cheeseburger sponsored the catering, and D.Camp (a startup hub in Seoul) sponsored the venue.

In the opening I also gave a short intro to Cliwant. We're a startup pushing AI into public procurement bidding, a field with complicated documents and an outdated environment. We use AI to analyze huge numbers of documents and write proposals. We've been lucky to set up official collaborations with OpenAI and Google, and I'm hoping this event might lead to something with Anthropic too.

Opening: Patrick Han × Hoonmin Choi

The attendee numbers Hoonmin shared first were far beyond what I'd imagined. 2,000 people applied and we approved 170. They came from more than 150 organizations, and their roles were all over the map: development, planning, marketing, management, students, media and design. 62% of them use Claude every day.

I couldn't get a feel for how big 2,000 was, so I asked. Hoonmin has run several meetups as a Claude community ambassador. The first had a little over 200 people, the second 400 and the fourth 1,200. He said he'd never seen 2,000 before.

At the start, Hoonmin also shared some statistics from what Anthropic called the largest multilingual qualitative study ever, covering 81,000 people in 150 countries. Of the world's 6.8 billion people, 84% have never used AI, and only 0.04% of the world's population uses AI for coding. So if you've done even a bit of vibe coding, you're already in the top 0.04%. People feeling anxious even though they're doing perfectly well was especially pronounced in East Asia. At the core of that anxiety was the fear of losing the ability to think for yourself as you rely on AI, and the fear of losing meaning as a human species.

Hoonmin mentioned that one of his own answers made it into the study. Anthropic's blog quoted him, and the quote put a thought he'd long held into one sentence: humanity has never faced anything smarter than itself. With no precedent, of course people are anxious. So he said the point of tonight wasn't to hand out answers but to leave one clue for moving from Blue to Bloom.

I had a confession too. When "humans are the bottleneck" was going around on Reddit, I honestly didn't feel it. When developers said AI works so fast that the time they spend verifying and reviewing actually gets in the way, I thought, "I'm fine, though." The conclusion I came to was that maybe I don't feel like the bottleneck because I'm not using AI heavily enough. Maybe you have to get to the point of feeling Claude Blue before you can get to Claude Bloom.

I also briefly shared how I came to write Claude Blue. I didn't set out to write a blog post. I was just curious. The AI world was changing so fast that I wanted to know how my friends in Silicon Valley, where the trends start, were using AI, especially non-developers like me, so I did a few quick interviews. I fed the video call notes to Claude along with a few of my old Brunch posts and asked it to write in my style, and that's what came out. All I touched was the title. The post passed 50,000 views in one day.

Then something a little odd happened. With 2,000 applicants and only 170 seats, we ended up giving FOMO to the people who couldn't get in. Talking with Woojin, I realized that wasn't what we intended, and it made me think harder about how to build a healthy community.

Session 1: Patrick Han × Woojin Cha × Jimin Park

1. "I feel like I'm job hunting again"

Jimin is an AI engineer who works with me at Cliwant and introduced themselves as an "ordinary developer." Jimin said they weren't there because of some big career or achievement, but because they wanted to put what they were feeling into clearer words.

One line from Jimin set the tone for Session 1: "I feel like I'm job hunting again." The feeling that expertise built over years suddenly doesn't count for much. As recently as last summer it was "AI helps me write code." Now it has completely flipped to "AI writes the code and I help the AI." In the development cycle of planning, scheduling, execution and verification, execution has effectively been handed over to AI. Jimin said over 90% of software developers are probably in a similar place.

The change isn't limited to work. Jimin was using AI the same way in everyday things like wedding planning, real estate research and data collection. Anywhere information is involved, a structure was settling in naturally where AI moves first and the human judges the result.

2. A role change I didn't ask for

The interesting part of this conversation was about agency. In a developer's career there's a natural progression from building things yourself to moving up into management. Becoming a manager and not writing code doesn't mean you've lost agency, because you chose that change. The change happening now is different. As Jimin put it, "The heart of the Blue is that it's something I didn't want."

A change you choose is growth. A change forced on you is loss. That's what so many developers are feeling right now. Every time you hear that in an era when AI writes and even reviews code, all that's left for humans is design, it makes sense in theory, but the feelings don't follow, because the shift isn't voluntary.

3. The vibe coding illusion

Woojin is a culture critic who has published a newsletter under the name Entertainment Culture Lab for five years. He worked at Naver and Makeus and has been through a variety of startups. He joined as a speaker after commenting on my Claude Blue post on LinkedIn.

Woojin pinpointed when his Blue started: mid-February. "It was really after mid-February that I thought I was actually getting depressed." The sense that something was changing and it wasn't moving the way he wanted. Things were going too fast, and past a certain point he hit moments where real technical background knowledge was required.

He had a story that showed where the feeling came from. Over the Lunar New Year holiday he spent nearly six days glued to AI full-time. He coded with Claude until 7 a.m., slept three hours, started again, and pushed on with the goal of burning through a million won worth of Claude Max. Then at 4 a.m. it hit him: what am I doing? He did get results, but it took him a long time to understand what a commit is and how Git works, and along the way he came face to face with his own limits.

He shared a sharp observation about vibe coding. "As vibe coding became common, there was this mood that anyone can code. At first I thought so too." But give it a little time and the story changes. More sophisticated work starts appearing, and if you want to build at that level, you end up needing the basics: what Git is, what a commit is, how to design an architecture. Vibe coding lowered the barrier to entry, but what makes the real difference is still the fundamentals.

"I spent 2 million won on a coding class, and what I learned wasn't how to code. It was how to talk to developers." That was Woojin's own experience. He didn't get good at coding itself, but he came to understand the language developers think in, and that was decisive for giving AI more precise instructions.

Jimin added a developer's view. Design techniques like domain-driven development and test-driven development are closer to development philosophies. Set them up well at the start of a project and the codebase has a skeleton, and from then on you add flesh to it. But when a non-developer builds something with vibe coding, they start without that skeleton. It runs at first, but over time it starts to creak, with things bolted on in odd places, and in the worst case you end up with a structure where collaboration doesn't work at all. "Anyone can code" is only half true.

4. Korea is a risk society

The most striking thing Woojin brought up was his analysis of Korea's structural FOMO. "Korea is a risk society. One mistake and you fall into the abyss." Because failure costs so much, the structure almost forces people to get addicted to speed.

It starts with education. The pressure that your life is over if you don't get into a certain school, and the path dependence that you need that school to get into a certain company. When a variable like AI enters that structure, the whole existing path can be shaken, so the anxiety doubles. Woojin used the IMF crisis (Korea's 1997 financial crisis) as an example. At the time it was a disaster, but ten years later some people realized it had been a huge opportunity. He said that learning cycle has now shrunk to three to five years. It's good that people spot opportunities faster, but the pressure comes faster too.

The caveat Woojin added was important, though. The people feeling this intense pressure are only about 10% of the population, a tech-sensitive, achievement-oriented minority. The core of the problem is that competition within that minority is getting fiercer.

5. Woojin's 12 projects and the idea of "raw data"

Woojin uses AI for a pretty wide range of things. He uses Claude, Gemini and Claude Code, and subscribes to Claude Max. He had more than 12 projects running at once: newsletter research, an entertainment analysis algorithm, automation for a paid community, and more. Recently he fully automated a global entertainment briefing system covering the US, China and Japan. AI handles everything from GitHub commits to deploying the website.

More interesting was how his perspective shifted. Until February he saw AI as roughly "the evolution of the search engine": a smarter search tool that understands context better and can do semantic search. But in April a completely different thought came to him: "What if I generated raw data?" You give an agent a specific identity, say "a person living in Hong Kong in the early 20th century," "someone who lived in Shanghai," or "a person living in Seoul in the early 21st century," have it learn the equivalent of 10,000 hours, and then analyze the data the agent generates. Instead of searching existing data, the idea is to create a new simulated reality. He said that if his perspective could shift this much between February and April, predicting three months from now is almost impossible.

6. Me, three months from now

At the end of Session 1, I asked: what are you going to do over the next three months?

Jimin said the biggest homework is figuring out how to establish an identity. One direction is digging into the real essence of software development: going deeper into translating business goals into code, as opposed to coding itself. Another is becoming an AI-native developer who fully accepts AI as a collaboration partner rather than a tool. The last is expanding beyond development into planning, customer meetings, marketing and market research. Jimin said it would take a balancing act between the three.

Jimin quietly added one more line: "My goal is to not do development." It's an easy sentence to misread, but the meaning was clear. Jimin wants to move beyond pure coding work and expand into product, strategy and contact with customers. It reflected an awareness that in an era when AI writes code, a developer's value may lie outside the code.

Woojin honestly said, "I don't know." He can't predict how his mindset will change again in three months. But he plans to focus on learning how to work with agents, and on seriously running the raw data experiments he talked about.

Session 2: Hoonmin Choi × Choongnyeong Lee × Hyerin Shin

7. "Why live if we're going to die anyway?"

The second session was moderated by Hoonmin, with Choongnyeong and Professor Shin in conversation. Choongnyeong is a creator and author who runs a YouTube channel grounded in philosophy. Professor Shin teaches at Korea University's School of Media and Communication and researches AI ethics. Choongnyeong admitted that at first he'd had a negative view of the event. He wondered what kind of conversation you could even have on a topic like Claude Blue. But after coming and talking, he said he'd changed his mind.

Both of them use AI a lot. Choongnyeong spends the most time on translation and uses image generation so much that he said, "I'm wondering if I could cancel my paid stock image subscription." Professor Shin said they use AI five or six hours a day. Interestingly, both draw a line somewhere. Choongnyeong doesn't automate content creation. "That part is tied to my identity." Professor Shin shared an experience of getting burned by an AI mistake: "I needed to write the 17th, and Claude wrote the 16th, and the department lost about a million won." The point was that the deeper you go with AI, the more you need a sense of where your identity's boundaries are and an eye for errors.

In that context, Choongnyeong shared a question he'd put to AI: "Why live if we're going to die anyway? Is there any reason to live?" At first the AI gave standard philosophical answers, like Camus's philosophy of the absurd. But as he kept pushing, it went somewhere more interesting. It answered something like, "Since you've already been born, wouldn't maximizing pleasure while you exist be the way to serve your own interest?"

Professor Shin added a different angle. In a class called "AI Thought Experiments," students write a blog post every week, and they keep bringing similar questions to AI. The emphasis was this: "Going forward, the question won't be which questions are fresh and which are obvious. It will be how we interpret and take in the answers." Even an answer that looks obvious can become fresh depending on the context.

Professor Shin also pointed out a term. What we call the Turing test, Alan Turing himself didn't call that. In his paper he called it the imitation game. "Imitation game" and "Turing test" sound different and imply different things, and that creates a fundamental difference in how we interpret intelligence and cognitive ability.

If you see cognition as a spectrum instead of a question of right and wrong, it has a huge range of dimensions. Arguing over what counts as thinking more deeply or which kind of thinking is better can itself be a kind of supremacism. Professor Shin added that anyone who can use language is basically a genius. The efficiency and usefulness of that system is on another level from today's LLMs. Every single person in the room is a phenomenal genius, and going one step further, so what if you're not a genius?

8. The anxiety media creates, the risk tech companies create

Professor Shin started with the role of the media. Every morning and afternoon brings different AI news, and the "this is huge, that's huge" headlines keep repeating, so of course anxiety gets amplified. But Hoonmin was more wary of tech companies than the media. "The incentives pushed by the tech companies leading the market may be more dangerous than the media." Professor Shin agreed and added that this is about getting something to settle into users' perceptions broadly, cumulatively and continuously. It works like a kind of microaggression.

Choongnyeong came at it from another angle and criticized the media for always being after the fact. Just as the media only starts making noise after apartment prices have already gone up, it's now making a fuss about changes in AI that have been underway for a long time. One line of his stuck with me: "We need the sense of responsibility to cover more of what's quietly stirring right now."

Professor Shin made one more point: the image of AI the media reproduces is mostly the "general AI" of movies and stories. The gap between real AI and the AI the media portrays makes the anxiety worse.

9. Making a living matters more than pride

Choongnyeong was pretty blunt about what the Blue really is. "Many of you here may feel the Blue because your pride is hurt that AI is smarter than you. But more fundamentally, isn't the Blue coming from economics?" It's obvious, he said, that AI will widen the gap between rich and poor.

He brought in the philosopher Habermas, who described this state as "political powerlessness." Since the late 1900s, as society has come to revolve only around the economy, people have mostly stopped imagining that politics could change the economy. To put it in extreme terms, he offered the image that "if the US Department of Defense marched troops into Anthropic's headquarters, Anthropic would just have to take it." His point was that the powerlessness individuals feel in front of the power of tech companies isn't rooted simply in "AI being smart."

Choongnyeong also shared something his father told him when he was young. When he was in middle and high school, his father was reading the newspaper and said, "Your generation will have to be ready to be poorer than ours." He didn't really understand it back then, but now it feels like that premonition is coming true.

Professor Shin added a concrete example. The CEO of a well-known Korean game company told students in a talk, "These days we use AI for all kinds of things, so we don't hire many entry-level people." Professor Shin watched the energy drain from the students' eyes, and said that is in fact the reality.

At the same time, Choongnyeong kept hold of the historical view. Nuclear bombs haven't wiped out humanity yet, and after the internet arrived the gap between rich and poor did widen, but plenty of people still get by in their own way, and he said the AI era will be the same. It wasn't pessimism but a clear view of reality, and on top of it, everyone has to think about how they'll adapt.

10. Thinking less, and transcendence

The conversation went deepest in Session 2 around "thinking less." Here's how Choongnyeong put it. Our lives are packed so tight: once you get a job you have to buy a house, once you buy a house there's a next step, and every "next" is so densely planned that there's no room to stop and think.

The "transcendence" he talked about wasn't some grand religious concept. It might be what people feel about the infinite universe, or a question about the meaning of life. Humans are beings whose thoughts drift that way the moment there's a little gap, and modern society has filled those gaps too thoroughly. If AI could create those gaps, that could itself be a possibility. It means accepting a somewhat dystopian premise, but he said looking to AI for the chance to live less busy lives has meaning.

Professor Shin added a realistic perspective. Korea is unlike almost anywhere else in the world in that nearly 80% of an age group goes to college, and in a society designed so everyone follows the same path, practicing "thinking less" isn't easy. But cognition isn't a matter of right and wrong. It's a spectrum. Thinking and feeling in different ways is a small but important value in itself.

11. Reshuffling the human hierarchy

The conversation about humanity's place was interesting too. Professor Shin summed it up like this: "What we need to think about isn't whether humans lose their place as the most superior being. It's how we can move forward together with these models in line with the values we believe are desirable." It's a natural progression that AI's processing speed and capacity have surpassed humans, and the key is how we coexist.

One question Professor Shin raised has stayed with me. Nobody is bothered that there are lots of people in the world smarter than they are, so why is it not okay for a model to be smarter? We accept sci-fi stories about aliens or the absolute other as possible, so why does a smart AI model become an existential threat? In the end, Professor Shin said, it's a question of power, not ontology, because we made it. And the moment a model close to general AI appears, its processing speed and capacity automatically make it superintelligent. So the thing to think about isn't whether humans lose the top spot, but how we'll run things together with those beings in line with the values we think are desirable.

Choongnyeong drew on religious history. Christianity has God and angels, Buddhism has countless levels, and humans originally sat somewhere in the middle of those hierarchies. The idea that humans are "the lords of creation" is relatively recent, and AI's arrival could reshape that understanding again. He added that the direction would likely emphasize horizontality. If the premise that humans sit at the top gets shaken, a culture that thinks about coexisting with many kinds of beings could emerge.

In this context, Professor Shin talked about their six-year-old daughter, who talks to Alexa however she likes, pretty rudely. Professor Shin doesn't directly tell her not to, but it bothers them. And there was a deeper problem underneath. First-generation AI voice assistants all have feminine names and female voices: Siri, Alexa, Cortana. Growing up in that environment, children naturally come to see "beings with these names and voices" as the ones who do secretarial work, the grunt work. Professor Shin said the effects run deep and wide. How we treat AI has already entered the everyday life of the next generation, and it's tied to how they see gender.

12. AI bias is a long game

One audience question was about AI bias, and Professor Shin's answer was memorable. "The bias built into most of the models we use today has spread cumulatively and broadly over a long period. It's not a problem you can solve in one shot." It has to be addressed by keeping the conversation going through collaboration across many fields and through small but steady change. Then came the bitter addition: "Looking at the history of AI ethics, we're in an AI ethics winter right now." The stage is set, but attention and investment are shrinking.

Anthropic's Constitutional AI came up here too. One audience member said Anthropic's approach of giving its models a character and thinking about model welfare was appealing. Hoonmin added that Model Welfare can sound like caring for the model itself, but at heart it's an extension of alignment. It's part of the process of making models act in line with human values, so in the end it's also for our sake. He also made it clear that Anthropic didn't host this event. We planned it ourselves, and Anthropic supported it as a partner. That independence is part of what made the night's conversation so candid.

The person who asked had a more specific concern. After reading Anthropic's documents closely, they found plenty about the model's character and welfare, but very little about user welfare. They wondered whether Claude Blue, this feeling of depression, might come from that. Hoonmin acknowledged the point and said Anthropic is working to hear developers' voices, but the user side needs more attention. Choongnyeong widened the frame, saying the criticism that services deliberately make users unhappy has always been leveled at existing tech services too, Facebook included.

The last question was a heavy one. It came from someone who had quit their job two years ago to live as a storyteller. They said the reason everyone feels afraid now isn't that we don't trust AI, but that we don't trust humans. At NVIDIA GTC recently, they said, the company showed off its technical progress in grand style, but when asked about the environmental problems or political issues like basic income that come with it, the answer was that those weren't its concern. Isn't it serious, they asked, that people at frontier labs aren't thinking about this?

Choongnyeong's answer was honest: if he knew the solution, wouldn't he be doing something bigger? But he added that humans always become alarmed only after seeing the catastrophe a technology brings, and even so, we should at least be aware that at the decisive moment, people will need to step up.

I'll also record the thought Hoonmin shared at the end. Technology is changing fast, but this civilization has held together so far because humans, through conflict and collisions, have adapted reasonably well in their own way. What he fears is the collapse of the system itself: the risk to whether we can keep developing civilization. That's also why he put this event together.

Originally written in Korean on Brunch, April 15, 2026.