Singtel × Bloom: Claude Bloom Singapore

We held a Bloom meetup in Singapore
Back in April, 2,000 people signed up for our event in Korea. In Singapore, more than 1,500 signed up. This time we again partnered officially with the Claude community in Singapore, and the local ambassador was thrilled to see far more applicants than usual.
We hosted the event with Singtel, a major telecom that you could call Singapore's version of SKT (Korea's largest mobile carrier). Locally it has AWS-level name recognition, and I'm told it's the biggest telecom in Southeast Asia. It operates beyond Singapore too, in other Asian countries and in Australia. Honestly, I'd never heard of it before. I needed a venue, so I dug through LinkedIn and sent cold DMs to anyone I could find until I finally landed one. It turned out to be one of the best-known companies in Singapore, so I got lucky. Even Jun-ho, who came with me, asked how on earth we got Singtel as a sponsor. There was no special strategy. We needed a space that could hold more than 200 people, and after a lot of attempts, it just happened to be Singtel.
A lot of people also asked how I connected with the Claude community in Singapore. That was through LinkedIn too. My honest view is that a local community made up only of Koreans isn't worth that much. Korea is so small that you can reach almost anyone through one mutual friend. But imagine our Bloom Discord with people from Japan, Singapore, Australia, the US and the UK alongside Koreans. I didn't know what it would mean for business, but I thought it would be fascinating and fun in its own right.
Our company already has an office in Singapore, so it's the Asian country I know best. That's why I wanted to hold the first overseas Bloom meetup there. I was grateful that Max, the Claude ambassador for Singapore, and Yier, a Googler who helps him, connected deeply with the theme, even though I was a stranger reaching out out of nowhere. Near the end, they said "Claude Blue" was exactly the right name for a feeling even I hadn't known how to describe, and they gladly signed on to run the event with us. The person at Singtel read my English blog closely too (it was a straight translation of the Korean original, so it must have been a long read), and asked to say a few words at the event, so we gave them the opening session.
Cultural differences
Singapore may be in Asia, but there are still cultural differences. I'd run plenty of community meetups in Korea, and I worried my approach would feel too strange in another country. A lot of people warned me that Singaporeans are less outgoing than you'd expect and don't easily show what they're thinking.
We pushed ahead the way we always do, with a fireside chat and small-group discussions. Contrary to what I'd expected, the small groups were as lively as in Korea, loud even. I would have liked to give them a full hour, but the schedule that day was tight and we had to keep it short. Afterward, a lot of people came up to tell me the small-group format felt fresh, and that unlike the usual AI meetups that just repeat skill walkthroughs, this one was fun. Someone from AWS said they'd sponsor next time and asked us to come back. One attendee who travels to Vietnam often said they wanted to hold a Bloom event there too. Nothing is confirmed yet, but the response was clearly as strong as in Korea.
From talking with locals, I learned that Singapore doesn't adopt AI as fast as Korea. Singaporeans tend to be cautious by nature, which is a little different from the Korean mood of getting anxious if you don't adopt things "ppali ppali" (hurry, hurry). Maybe Korea is the odd one out.
What I found interesting was a meeting that same day with NCS, a big IT services and systems integration company, roughly Singapore's equivalent of Samsung SDS. I told them we were holding Claude Bloom at Singtel and explained the "AI Blue to Bloom" story, and the person I met laughed and admitted they struggle with exactly that a lot. To be honest, "AI Blue" can sound a bit corny. But I find it hard to believe anyone who says they've fully gotten past that stage. Even Sam Altman looks pretty "blue" these days. That night, none of the three people who volunteered to come up and speak said they had fully conquered Bloom. They all said they were still going back and forth.
What I'm still asking myself
It's my own story, so maybe I'm a little too into it. But whether you call it "AI Blue" or something else, this isn't simple. So I want to ask myself whether I'm really preparing for the AI era, and whether I'm sure that my service, my business or I myself won't be replaced. Jun-ho and I talked about this in the taxi that day. People have been saying AGI is coming for a long time, but we didn't expect it this fast. We asked each other what we would have tried again if today's AI had existed three years ago. Applying that same pattern to the next three years seemed like better preparation than looking back. Even so, we both honestly ended up at "I don't know."
It used to take serious skill just to crawl and collect data, and that alone could make money. Now collecting data is easy. In software, at least, there's almost nothing you can't do. Bloom is a project I sort of fell into, but oddly enough, it's building up data nobody else in the world has, from our pre-event surveys to our applicant data. I don't know yet how we'll use it, but it does feel like we're building something that's ours. If I had pitched "AI Blue to Bloom" a few years ago, nobody would have related to it. Everything takes a certain amount of timing and luck. You can't get there by calculation alone.
So instead of trying to predict three years out, I think if I just keep doing the work in front of me, it will turn into opportunities and connections on its own and give me the strength to keep going. If you ask why I'm taking the Bloom community to Singapore right now, the answer is simple. An opportunity opened up, so I'm doing it. You only find out what's there once you try.
That was a long intro. Here's a writeup of what we talked about that night.
The eighth Claude meetup in Singapore
Our host, Max Del Vita, is Singapore's Claude ambassador. This was the eighth Claude meetup in Singapore, and the first since he got back from spending all of June in Europe, so he started by asking for a round of applause for everyone who showed up on a Tuesday night. He said every seat was taken within 24 hours of the announcement.
Max was happy that the community he'd been running alone for a long time finally had siblings: Anthropic had appointed two new Claude ambassadors in Singapore who would run events with him from now on. Along with news about the Claude startup program, which he said startups should look into, he shared a scene from the current AI drama. At an event in India, when the AI leaders were asked to all join hands, Dario Amodei and Sam Altman were the only two who never did. The photo got a big laugh.
When the laughter died down, Max got more serious. In just the last ten days, a pile of new models had come out from every lab, including Anthropic's Fable, and after trying them all himself, he said they felt like an entirely new generation. Exciting, but honestly a little scary. Every morning you wake up to a dozen or so announcements that would have changed your life ten years ago, and now it's just "oh, another one." We're living through that kind of acceleration. He said that made a good stepping stone into the evening's topic and handed me the mic.
Claude Bloom's first global meetup
On stage, I started with why I'd come all this way. I work at Cliwant, an AI company in Korea. We're a B2B company that makes software, and we originally had nothing to do with community. But early this year, there was a worry going around the company that building software was getting harder. What if Anthropic swallowed our software? What if users decided to build their own tools instead of using ours? I couldn't shake that anxiety, so I started asking people inside and outside AI for coffee chats. Finance, construction, manufacturing, the AI industry. I wanted to know how everyone was making a living in this era.
Then I talked with a friend who works as an AI engineer at Meta in Silicon Valley, and he told me the more he used AI, the more depressed he got. He has a PhD in machine learning and 20 years of experience, and these days his work goes like this: get a task from his manager, put it into an AI, open another AI and teach it what the project is, repeat that for two days, then report. For the first few days it was exciting that AI finished the work, but at some point confusion set in. What am I doing here? Am I a prompt engineer? Someone who types on a keyboard? It was an identity crisis.
When I wrote about that conversation on my personal blog, I named the condition Claude Blue, after "corona blue" (the Korean term for pandemic depression), since everyone around me, my friend included, was using Claude. I'm not famous in Korea, but that post, Claude Blue: All of Silicon Valley Is Depressed, spread far more than I expected in Korea and on Reddit, and I got interview requests from major media outlets and even from the US. What happens online is great, but I wanted to meet people in person and find out what was really going on. Since I'd named it Claude Blue, I figured I should do it with Claude, so I contacted the Claude ambassador in Korea. I knew it was mostly a developer community, but this was a more philosophical, human approach, and I asked if they'd host something together. The ambassador reported it to headquarters, and headquarters approved, saying they wanted to try something new. That's how we held the first event in Korea on the theme "From Claude Blue to Bloom," and more than 2,000 people signed up in three days. It told me this wasn't just a conversation happening online. People actually wanted to show up and connect with others who felt the same.
We couldn't fit 2,000 people in one room, so I had to keep running events while working at a software company. The idea itself is simple. When you first use AI, it's exciting. I call that stretch AI Excitement. But the more you use it, the more down you get. Some call it FOMO, some call it job anxiety, some call it an identity or existential crisis. I didn't want the story to stop at Blue, so I named it Bloom. Everyone here can probably relate to Claude Blue, but nobody wants to stay there. In Korea it became a community that draws people from big companies like Samsung and Hyundai, plus law firms, financial companies and government agencies. AI events used to be for developers only, but this theme lowered the barrier, so engineers and non-engineers mix and share their experiences. Every meetup gets more than 500 applicants, and some weeks we run three events. In total we've had 6,000 applicants, and 2,600 people have joined our Discord.
A community we never planned had grown on its own, and the next step I wanted was to go global. I wanted to test whether the idea worked outside Korea, so I reached out to Max. He said the theme hit home for him personally and was well worth sharing with the Singapore community. That gave me the confidence I needed, and Cliwant CEO Jun-ho Cho and I got on a plane.
A company investing in AI culture
Raymond Chong from Singtel explained why Singtel joined Claude Bloom. When I first came to them with the Claude Blue story, he said, it resonated deeply with what Singtel was going through. They were in the middle of a workforce transformation, redesigning how they work, how they lead and even the business itself around AI, beyond simply working alongside it. So instead of the usual Claude event that's all tech showcases, they proposed having a real conversation this time. Companies that invest in AI culture, and not only AI infrastructure, are rare. Thank you again to Raymond Chong and William Liu.
"Right now I'm Bloom-ing"
The fireside chat featured Jun-ho, Yier Cao and Max. One of the new ambassadors was supposed to join but couldn't make it, so Max stepped in with no prep. I asked each of them to introduce themselves and say where they were standing right now: AI Excitement, AI Blue or AI Bloom.

Jun-ho has deep ties to Singapore. He spent more than ten years here in the bidding business, writing proposals and analyzing RFPs, until one day he realized the industry hadn't changed at all in over 15 years. If nobody else was going to change it, he would. So three years ago he founded Cliwant, which builds AI solutions for government bids, and was lucky enough to get $500,000 in credits through partnerships with OpenAI and Google. "So I'm definitely tokenmaxxing," he said, and got a laugh by describing how he lives with several Codex, Claude and Gemini sessions open every day.
He was candid about Blue and Bloom from a founder's point of view. Early this year he kept asking himself what business is safe from AI. The answer: none. That made him depressed. He went Blue. Then he changed the question: what business could actually grow because of AI? The facts were the same, but with a completely different angle, answers started to appear. His answer was to redesign the business model. He boldly dropped the SaaS subscription model for a revenue-sharing model, something that only became possible because of AI. "So today, I'm Bloom-ing." Everyone nodded and laughed at that one word that said he was still in progress.
On tokenmaxxing
Yier Cao has been a regular at this Claude meetup for the past seven or eight months. After working in strategy and product at big tech companies, Yier recently joined Google Cloud as an AI architect and uses AI every day for both life and work. Yier has personal agents that handle daily chores and travel planning, and built a second brain for themselves and their agents to get up to speed at the new job faster.
Asked which stage they were at, Yier said they'd gone through every stage in the past year. When they first started coding with AI, it was definitely excitement, because they were building things they'd thought impossible. But they soon realized the excitement came from a place of deep ignorance. The more they built, the more anxious they got about how much they didn't know. Right now they're making peace with it. They've accepted that it's impossible to be an expert in every tool and every new technology, and that it's fine to learn one project at a time. The anxiety still hits sometimes, of course. "Yesterday I saw this meme. It was a photo of a guy coding while driving, and it said if you haven't hit your Claude limit by noon, you wasted your morning. You all feel that, right?" The room laughed in recognition. "When that feeling hits, I just let it pass. Not tokenmaxxing isn't a sin."
In the next question, Yier talked about switching over without a technical background. Their current title is very technical, but they come from strategy and product, so I asked whether they think of themselves as a builder or a technologist, and what that transition was like. "Thanks for bringing out my impostor syndrome," Yier laughed, then said they'd always been a builder. They'd built teams, built a digital bank and raised funding for companies. But if you'd asked five or six months ago whether they were technical enough for an engineering-heavy role, they would have said no. What changed is that AI collapsed the cost of entry and compressed the learning timeline. Even before switching jobs, they built things in their spare time and shared what they learned through workshops for individuals and companies. Above all, they said, once you run a multi-agent system in real production for a few months, you hit real failures, like a subagent with permissions opened a little too wide or a security risk that slipped past you, and that's where the real learning comes from.
Yier compared their career to an exploration game, the kind where the map starts out dark and an area only opens up once you clear quests and walk into it. They went through product, consulting and finance, and at one point even did aviation investment, when all they knew about airplanes was how to board one. Each time they enter a new area, they become a beginner again and build up skills and domain knowledge. The interesting part is that when those domains intersect, the value compounds. The jump from big tech operator to AI architect is the biggest transition yet, and AI let them open up that map much faster.
"This is the new drug"
Before he's a Claude ambassador, Max is a founder running his own startup. When he started the company a year and a half ago, the idea was to build an AI-native company from day one, and one of its principles is that everyone is an IC (individual contributor). Technical background or not, everyone does everything with AI tools. So even with only three people, their throughput is huge. Work that would have needed ten to fifteen people a few years ago, two or three now handle comfortably. Coding, strategy, docs, specs, communication, all of it.
When I reached out with the Claude Blue story, he said, something lit up in his head. A feeling he'd carried for a long time finally had a name. That feeling has several layers. One is the pressure to always be efficient: you've been given the gift of understanding and using this technology, so now you have to maximize it. Then there's the constant input. "I once told a friend, this is the new drug. It's really addictive. I don't smoke and I've never done drugs, but I think I now know how addicts feel. I can feel it in me."
He's a builder to the bone who has been making software for 27 years. All through his years as a product manager, he envied his engineering colleagues. Git, deployments, cool stuff he didn't understand. He was the guy on the side talking and drawing roadmaps, and they were the ones actually building. Now he can do it himself. Along with the joy of that empowerment comes the fatigue of new technologies and new platforms pouring out every day. "Honestly, it's exhausting. Keeping up with the latest without feeling FOMO is really hard. The FOMO is huge."
What a ten-year-old taught us
There were a number of parents in the audience, so I steered things in a more personal direction. Jun-ho is a father of two, and I have a young kid too, so we talk about parenting a lot. His elementary-school son doesn't play games anymore. He makes them with AI. If a generation is coming where anyone can build anything with AI from childhood, I asked, what will set future generations apart? What skill will humans hold onto until the end?
Before answering, Jun-ho asked a boy in the audience to stand up. It was Leo, 14, who loves AI and Claude and had come to the evening event on his own. The room applauded. Then Jun-ho told the story of his son. His ten-year-old was hooked on Minecraft and Roblox like any kid, and one day Jun-ho asked him, are you good at this game? His son said he was the best at school. "If you're the best, if nobody's above you, why not try making a game?" His son asked back, how do you make a game? So Jun-ho introduced him to Claude, and magic happened. Not in days but in hours, he made ten games, including a shooter and a Tetris-style puzzle.
After the first game, his son said, "Dad, this game isn't fun. My friends say it's not fun at all." He told the AI he wanted to make a 3D game like Roblox, and Claude recommended Unity 3D, hooked it up and built a 3D shooter. A few years ago, making a 3D game in a few hours would have been unthinkable. The AI built it, not the kid, but the AI never once asked what to make. The kid decided what he wanted. He shared the game with his friends, took their feedback and fixed it again and again. "Here's what I learned. AI collapsed the entire cost of building. But AI doesn't know what users really want or where it hurts. My kid knows that, because he knows what his friends like." Max cut in jokingly: "Who's paying for those tokens?" Jun-ho shot back that he hadn't connected billing yet, but would soon. Yier added that his son can empathize with other gamers because he's a gamer himself. Exactly. Empathy. That's what remains.
Why Anthropic doesn't do online events
I also asked about the generations in Singapore. In Korea, both senior people at companies and college seniors are deeply afraid of jobs disappearing, while young people, maybe because they have nothing to lose, are taking big swings and shipping open source. What about Singapore? Max said Singapore understands this problem very well. The leadership knows exactly the challenge of quickly retraining a huge workforce, to the point that the government invests fairly aggressively in training and transition. The problem is that around the world, far more countries and companies don't understand it, don't believe it, don't invest in it and even mock it. The fact that so many people without technical backgrounds come to events like this, he said, is a clear sign that people have started asking questions. A transition is coming, so what should I do? How will it affect my work and my life? Honestly, we don't have the answers. Ten years from now we'll look back and see the path of the transition clearly, but right now it's hard to see the next step. In that anxiety, all we can do is keep learning, understand what this technology is and learn to use it well.
I closed the fireside chat with the same question for all three. Anthropic is a tech company, yet it passionately runs communities like this all over the world. Why do people come to offline events like this? What is the power of community in the AI era?
Max said he started this community last October and the response was close to crazy. At root, people are curious. They know this is more than just another technology. It's a wave that will change everything we do, so they want to learn more. And the people who come here want to meet other people. "I don't know if you've noticed, but Anthropic never does online events. After I became an ambassador, I asked. Fifteen hundred people sign up and we can't host them all, so let's livestream it. Anthropic's answer was always clear: we want people to meet face to face." The connections made face to face are different. That's why people stay after events, talking about what each other is building, until we practically have to push them out.
Yier explained why they keep coming back. Partly it's curiosity about what everyone's building, but more importantly, interacting with AI every day can be pretty stressful and lonely, and that's actually measurable. Lenny's Newsletter's second annual tech worker survey, which came out a few days earlier, showed that AI has clearly raised productivity overall, but burnout rose too, from 45 percent to 56 percent. Max cut in: "Wasn't this technology supposed to make us work less?" Everyone laughed, but it stung. That's why community matters. It's where you find your people who share similar experiences, where someone who just started coding meets peers at the same starting line, and it's a real support system. A few months earlier, Yier had demoed their travel concierge agent here, someone gave them advice on memory settings, and they still use it. "Sometimes it can be even smaller than that. Just walking into this room and hearing someone say, 'Oh, I'm going through that too.'"
Jun-ho said he's no community expert, but he shared the moment that made him believe the real answers are offline. Before starting this business, he was bidding on a project for 313@somerset, a shopping mall on Orchard Road, proposing an interactive kiosk solution. The RFP said the work had to include system integration, and he couldn't figure out why. They only needed to provide the solution, so why was "integration" in the document? After puzzling over it, he just went to the site and found small displays in every restroom in the mall, all switched off. That's why they wanted system integration. The answer wasn't in the document. It was outside it. His was the only foreign company in that bid, and it put in the highest price, yet it won, because it understood the hidden need the client really had. "This community is the same. The real answers are offline. We're all different, but we can ask each other how you use AI, here's how I use it, am I doing it right? It's always been like that in Korea. The real insights come after the fireside chat. The fireside chat is just the appetizer. The main dish comes next."
From Claude Blue to Bloom: the small groups
For that main dish, we ran an experiment. We put pink stickers on the chairs in every odd-numbered row, and had those rows turn their chairs around to form instant groups of eight with the row behind them. Share your Claude Blue moment, and tell us whether you have a routine or mindset that gets you to Bloom. We do this every time in Korea, and I was curious whether it would work in Singapore.
It worked. The conversations got so heated we had to cut them off for time, and what a few groups shared on stage sounded remarkably like us.
The first person to take the mic was a ten-year tech veteran. They'd been through the machine learning and deep learning era, before prompts and LLMs took off, and their job required staying on top of every new framework that came out: AI, vector databases, front end, all of it. So they spent four hours a day on Twitter feeling FOMO over new frameworks and new posts, poking at everything, and their laptop and workflows turned into a mess. After five months, burnout hit.
Extreme anxiety, extreme FOMO. That was their Claude Blue moment. Then they made a decision: drop everything and pick only a very few tools for each layer of the stack. Two ecosystems for the model layer, two for the memory layer, and for the harness layer, after Claude Code, either LangGraph or a single custom SDK. They built a simple stack with only two per layer, and now, when something new comes out, instead of chasing it, they reason from first principles about where it fits and whether something already in their toolkit can do the job. They said they'd also built an open-source repo called Claude Forge that compiles multiple web apps, mobile apps and workflows into one, and would share it with the community. You tell it the stack you want, and it assembles a working product for you. It came out of building many production-grade products, then clearing the clutter back out and focusing on eight to ten fundamental building blocks.
Their colleague took the mic next with a funny confession. Opus 4.8 was announced at 3 a.m., and just as they were about to go to sleep, they couldn't resist. They updated Claude Code, swapped in the model and reran their coding pipeline. "I don't think I've ever made it to Bloom. I'm always running something, so I'm always Blue." Everyone laughed, but Max nailed it: "We need to sleep more."
The group with someone who leads AI adoption at a major bank was diverse from the start: someone who moved from very traditional industries like manufacturing and defense into a tech role, someone who left big tech disillusioned by the arrival of AI and became a venture builder, and someone who went through legacy organizations and now helps small companies with tech operations. What ran through their conversation was that the baseline keeps moving. Whatever your role and whatever the size of your organization, the baseline keeps shifting and the pressure to keep up is the same. Only the context changes. You keep finding something new to fight, new to complain about, new to adapt to. Their Bloom was purpose. For the person who left big tech to build ventures with purpose, the Bloom moment was finding purpose in what they were building. For the person at the bank, it was finding a lasting project that makes it easier for people to adopt AI.
Someone who once led an organization and had been CTO at a couple of startups confessed they'd had their laptop open running agents the whole time the fireside chat was going on. "I ran out of tokens ten minutes ago. So I turned to the next row to chat, and that person had their laptop open with agents running too." Their own story went like this. One night a week earlier, they got home from work and were about to go for a run, but seven agents were running on their laptop. If they left now, all those specs would finish and there'd be nothing queued up by the time they got back. So they couldn't go running until they'd kicked off the next batch of work.
The person next to them topped that. They set an alarm for midnight to kick off agents, because they can't fall asleep if no agents are running. "We call this Bloom, but I call it AI-induced psychosis." The room erupted. But their real definition of Bloom carried weight: when you see something you know is very complex just work, when you look a user in the eye and think, how did we pull this off so fast. The gap between what they could do leading a big team and what they can do alone now is night and day. The satisfaction of delivering. Their sleep is wrecked, and the tokenmaxxing and psychosis are all real, but they said that's Bloom.
Vibe coding meditation
To calm down the psychosis the conversations had stirred up, we'd prepared an AI mindfulness session. Keziah got a laugh right away on stage. There had been a technical problem during setup, and Max opened Claude Code and fixed the JS code on screen in 30 seconds.
Keziah is a non-developer who works full time as a consultant at EY, a meditation teacher, and what they call a mindfulness coach. On stage, they calmly shared their own story. Mindfulness saved their life. They had been extremely anxious and extremely depressed, and even as they went to a counselor as a last resort, they were afraid it couldn't be fixed. They wanted to leave this world. Mental health is our most important and most invisible part, they said, and society is only now slowly getting better at talking about it. Tonight was proof of that.
The experience they built is called The Cave. They made it using Claude Code and Gemini together, and solved the particle physics formulas by having the models talk to each other. "I don't know how to code at all. I don't know particle physics either. What I have is five years of guided meditation scripts, my skill set, and an attitude of figuring out new technology no matter what." On screen was a cave of particles rippling in response to sound. Ruminating on the past and planning and worrying about the future both mean living somewhere other than here and now, and most human suffering comes from not knowing how to stay in the present. So the experience is designed as a gateway to experience present-moment awareness, even for a single second, with three anchors: voice, a visual anchor on screen, and music. The moment you notice your mind wandering is the magic moment, they said, and every time you notice and come back, the muscle of presence grows. If you're spinning in worry about models, about the next thing, about who you are if AI takes everything, this muscle brings you back to the present. Back to ourselves: humans, mammals who live through community and connection.
The lights went off and we took three deep breaths together. The feeling of your body against the chair, the tension in your shoulders, the breath passing through the small triangle under your nostrils. "Breathing in, I know I am breathing in. Breathing out, I know I am breathing out." To anyone who noticed their mind had wandered, Keziah said that was a wonderful moment. You noticed, so now you can bring it back. You're a living person, not a robot, every breath is different, and your body breathes to protect you without you having to plan it. When the meditation ended and the lights came on, Max asked if it was on GitHub. When the answer was no, he cried out jokingly, "What? I couldn't live with the anxiety of losing the code. Please open-source it."
Demos in the spirit of Bloom
It was late, but we still had community demos. Max said he'd picked ones that matched the spirit of the night.
Kush had presented a platform for validating ideas before you build them at the very first Claude Code meetup in Singapore, and has since built a lot: a burnout prevention platform, a pet care app, a language practice tool, biotech support, and consulting on elder care using music and movement. Tonight Kush brought a quiet, invite-only social platform. In the feed, a friend had posted about a peaceful walk. Nothing begs for attention or clicks, and nothing is designed to bait comments. That's by design. Kush invited only a few truly trusted friends, and invites are sent by copying a link or by email. No ads, no optimizing for engagement, no clickbait, no data mining. As for how it pays for itself: "You are the customer." You pay $5 a quarter and get the service, with no ads and no algorithm. Kush's reason for building it lined up exactly with the night. "I think part of the Blue isn't AI itself. It's the outrage, negativity and FOMO manufactured for engagement. I want to make a quiet place away from that chaos."
A perfume bottle generator was a demo inspired by fashion houses' algorithms, built on the idea that anyone can now be a perfumer. It averages historical data on fragrance notes to set the parameters, and when you adjust the notes, say removing citrus and adding moss and amber, the shape of the bottle itself changes to match the character of the scent. Every brand has its own bottle silhouette, so the maker mapped notes to shapes. You can tweak details like glass thickness and save it as a keepsake of the scent you created. The tech is basic JavaScript and HTML, and the maker said Claude debugged countless errors they didn't know how to fix and suggested features. It's open source on GitHub, so anyone can play with it.
Tanisha, a design technologist graduating next month, brought Pendulum Type. Scrolling Instagram, Tanisha saw a double pendulum simulation and found the way it shows deterministic chaos beautiful, and wondered whether they could capture the beauty in that chaos and give it form. So they tied double pendulums to the strokes of letters and made them write. On screen, the pendulums swung in real time, drawing the letters. They built it a few weeks earlier with Claude Opus and hit a lot of obstacles. The pendulum didn't lift the pen between strokes, so the letters smeared, which they fixed by timing when the pen lifts and drops. They put in typography knowledge so letters hold their positions when writing a word, plus the pendulum physics formulas, and it took 10 or 15 rounds of fixes before the camera could follow the pendulum without jitter at any letter size. Just for fun, they also made a retro UI with a '90s internet feel, a Y2K version and a Matrix version. "I think this is closer to mindful AI use. You take part in the process instead of being a spectator who just receives the result. You play with it and watch what comes out. My Claude moment is the fact that I can make things like this at the intersection of design, science and technology, my favorite place to be. Without this era, it would have been much harder." Even after the demo ended, the pendulums on screen were still finishing the last three letters.

Bloom Seoul, and a midnight flight
When I came up to close, I asked how the night had been. This format may have felt a little unfamiliar and special, but in Korea it happens every week. I was amazed watching you form groups and have deep conversations, and that's exactly what I wanted to give you. Next time, I hope the Claude community in Singapore gets together again and keeps these conversations going even without me.
Running events for 100 to 150 people every week, I'm always thinking about what's next. One answer is going global like tonight. The other is holding a 1,000-person conference in Korea. Filling 1,000 seats obviously takes people from outside Korea, so I shared the news I'd brought: a global AI conference called Bloom Seoul, at the end of October. When I played the venue video, Max summed it up in two words: "Very Seoul." Right, very K-style. Anyone coming all the way to Korea from abroad will expect a little K spice. Before joining my current company I worked in marketing for BTS, so tech and K-culture are a bit mixed together in me. I told everyone to get in touch if they ever come to Korea, said I hoped to see the Singapore community in Korea someday, and put down the mic.
Even after the event ended, people stayed late, talking over food. It was one night and one event, but it will stay with me for a long time. Hoping it won't be the last, I took the midnight flight back to Seoul.
Originally written in Korean on Brunch, July 15, 2026.
← Back to Work