Proud to back @delta_foresight from day one. Onchain finance today runs almost entirely on dollars. JPYdf aims to bring the yen onchain, so global liquidity can finally settle in JPY. @yusukejindo was #1 nationally on Japan's CPA exam, audited crypto exchanges at Deloitte Tohmatsu, helped launch Mercoin, and was CFO of Animoca Brands Japan. He knows both the regulation and the tech, and a yen stablecoin needs both. Excited to build this with Coral Capital and the rest of the round. The yen is coming onchain. 🇯🇵

Sam Altman(@sama) built OpenAI(@OpenAI). He also co-founded World.(@worldnetwork) The person pushing AI forward faster than anyone on the planet also started a project whose whole job is proving there's a real human on the other side of the screen. I don't think that's a coincidence. The harder you push AI, the more you need a way to tell who's human. Two sides of the same problem. Narrative(@NarrativeCap) partnered with OpenAI. Founders who work with us get $10K in OpenAI API credits and 10 seats of ChatGPT Business, Codex included. This time it's Rocketpunch(@rocketpunch_kr) teaming up with World. A World ID integration and an Orb Operator partnership. So now we're working with both of the things one person built: the side pushing AI forward, and the side proving you're human. World ID is an anonymous proof of human. You get verified at a device called the Orb. The ID lives only on your phone. Rocketpunch receives exactly one piece of information: whether this person has completed verification. My own Rocketpunch profile has had a "Verified Human" badge on it since April. We quietly opened it up, and I tried it first. Now it's official. On Rocketpunch you can sign in with World ID, or link it to the account you already have. Finish Orb verification and the badge goes on your profile. Here's what actually changes once you're verified. When you reach out from a badged profile, the other side doesn't have to start by wondering if you're a person. Where the first message matters — a hire, a fundraise, a partnership pitch — that difference is big. Sign in with World ID and we also open up unlimited chats with other members and AI search. An account confirmed to be human can be trusted with more. Next up: offline. Starting with Narrative Lounge Seongsu, we're working toward bringing Orb verification to the conferences and networking events Rocketpunch hosts. What Rocketpunch is trying to do is clear: become the verification infrastructure the AI era needs. We'll keep opening up more for builders, the Rocketpunch Open API included, so they can build on the assumption that there's a real person on the other end. Unified sign-up and sign-in with Google, Apple, Kakao, Naver, and World ID is already available through the API, along with data and features for jobs, events, and posts. docs.rocketpunch.com/openapi Coverage (Korean): platum.kr/archives/294016

We've cut more than ten SaaS products from our stack. What replaced them isn't software we built from scratch — it's mostly open source someone else built well, forked and shaped to fit how we work. @firefliesai(meeting notes), @attio(CRM), @loom(screen recording), @Calendly(scheduling), @Zapier(automation), @WisprFlow(dictation), @reflectnotes(notes), @webflow(website building), @obsdmd Sync, @buffer(social scheduling), plus the Korean tools we'd been using for SaaS spend tracking(Smply), business cards(Remember), and e-signatures(Modusign). The list keeps growing. CRM made it clearest. We didn't need a general-purpose CRM. We needed a store agents could plug into easily — one place where every context we're trying to connect gets recorded and maintained. Off-the-shelf CRMs are built on the assumption that a person types things in and a person reads them back. But the thing doing the work shifted from people to agents, and no amount of subscription money buys that shape. So we forked an open-source CRM onto our own server. A few days, including the extra features we bolted on. Build vs. buy is an old question. Our answer is fork and shape. We didn't cancel because the products were bad. They were all good, and we still run some of them alongside. There's always been a gap between what a general tool fits and what fits us, and closing that gap used to cost far more than the subscription. That's flipped. With AI writing code and the open-source ecosystem this thick, covering the last stretch ourselves got cheap. And it isn't only the subscriptions we cancelled. Some of it is SaaS we looked at and didn't buy. We're about to drop the accounting and HR service we've been on, so we went back out to shop for something to fill the hole. We decided not to. Rocketpunch Team, which we built in-house, was already handling attendance, time-off requests, work logs, AI usage, and a view of how the whole team is working. For the rest, adding a few features ourselves or pairing it with free SaaS made more sense than signing. This isn't just an engineering story either. Someone on our ops team automated the month-end close with Claude — the job of eyeballing card statements against expense approvals and typing in account codes one by one. They didn't file a ticket. They built it. A request like that used to sit at the bottom of the engineering backlog. You can't pull engineers off the roadmap to save someone a few days of month-end work. So the gap got covered by that person's overtime, every time. Now the person doing the work closes it themselves, with AI. Forks come with a bill, of course. You own the updates, and you own the security. So this isn't an argument for ripping out all your SaaS. But it did change the first question we ask. Not just "is this a good product?" but "can our agents plug into it, and how many days would it take to shape an open-source alternative to our context?" And the person who answers that isn't only an engineer. It's whoever uses the tool every day.

Last week, our GitHub org stats showed 774 commits from 8 contributors. The #1 committer was @claudeai. 219 commits, +81.1K lines added. An AI agent is literally our top contributor at @rocketpunch_kr. We built Rocketpunch Team in 2 days and deployed it across the entire organization. It pulls work data from @googlecalendar, @SlackHQ, @NotionHQ, @Jira, @github, @linear, @Confluence, @googledrive, @gmail to auto-generate work logs, tracks attendance via NFC, and analyzes AI tool usage across the team. One dev spent an estimated $3,323 in AI coding costs in a single week. Team members are vibe-coding their own mini apps and deploying them on their own. Lunch pickers, expense trackers, even games. The bottleneck at our company moved. Execution used to be the constraint. Now, with AI agents handling the commits and the team vibe-coding their own tools, the real constraint is judgment. Deciding what to build and which data actually matters enough to act on.

"I want to work in Korea." Someone told me that in Tokyo, and it caught me off guard. I was in town anyway, so I met one of our @rocketpunch_kr users in person. She's teaching herself to code at a government-funded bootcamp, trying to break into dev. I asked how she found us. Not an ad. She'd asked Gemini how to land an IT job in Korea, and it sent her to Rocketpunch. That hit me two ways. First: it's real now, the thing people call GEO. You don't win discovery by outbidding everyone on ads anymore. You win by being the answer the AI actually gives. Second: it made me feel far more responsible. She trusted that one recommendation enough to sign up, build her profile, and even post about wanting to work in Korea. We have to be worth that. We went through her profile over coffee. Scrolling together, I kept landing on parts of the app that still aren't translated into Japanese. She waved it off with a smile. I couldn't. We're fixing that faster. She shows up to that bootcamp every single day. A lot of us here take this place for granted. Some of us want to leave. She's fighting to get in. So I'm back to building. Online and offline, one piece at a time, until every user gets a real shot here. A café in Tokyo reminded me why I started.

Don’t expect much from the program. After selection and arrival at Network School, I had basically no communication with the Base team. Messages went unanswered, support was absent, and I was mostly left to build on my own. Disappointing. @base @jessepollak

A friend in Malaysia @ns paid 13 MYR for me. In the past, small amounts like this were always the most awkward. Paying back in cash feels inconvenient. Exchanging money just for a small amount is annoying. International transfers are even more complicated. That is one of the moments we are building @rocketpunch_kr Wallet for. You enter the amount in local currency. It is automatically converted into USDC @base. Then you can send it to your friend right away. The final confirmation is just one Face ID scan. Crypto often feels like a difficult technology. But the moments where it becomes useful are actually simple. Sending money to a friend you met while traveling. Settling a small payment across borders. Paying without visiting a currency exchange booth. We are not trying to make finance more complicated. We are building a global wallet that people can use in everyday life.

I tried a robot installed by @InvLambda at Network School(@ns). The opportunity I saw was not just the robot. It was the data behind the robot. Physical AI is AI that operates in the real world. It does not only generate text or images. It has to move, grab, miss, correct, and try again in physical space. To learn that, robots need real-world movement data. A person remotely controls a robot. The robot records the movement. That data helps train the next generation of robots. This can create a new kind of work. People may be able to operate robots from home, help generate training data for Physical AI, and earn income from anywhere. If we had Play-to-Earn in gaming, we may also see Teleop-to-Earn in robotics. Robotics will not stay inside factories. It can move into logistics, care, food service, homes, and offline stores. This is why I keep watching companies like Apptronik and Figure AI. The question is not only how advanced the hardware becomes. It is how fast the industry can collect the data needed to make robots useful in the real world. Rocketpunch will continue to look for these global opportunities and connect them. Not just the technology itself. The work it creates. The investment opportunities around it. The companies building it. The people who can participate from anywhere. As AI moves from screens into the physical world, the best opportunities may come from knowing where and how to participate. Rocketpunch(@rocketpunch_kr) will keep building that business network.

The most expensive “Hi!” I’ve ever sent. In December 2023, I got a Telegram DM from @xulian_hl, a Core Contributor at @HyperliquidX. “gm gm — I’d love to talk more with the Korean community.” My reply: “Hi!” Then I didn’t properly follow up for about 14 months. During that time, Hyperliquid kept growing, early users used the product and earned points, and today Hyperliquid trades at around $70B FDV. I missed the chance to collaborate, try the product early, and earn points. This is a reflection on my reply habits. Opportunity walked into my DMs, and I basically said hello and closed the door. Next time a great team reaches out, I’ll reply properly — and actually use the product. Huge congrats to Xulian and the Hyperliquid team. Respect for building a great product, growing the community, and proving it in the market.

I’m currently at @ns in Malaysia. I was curious what would happen if founders, builders, creators, and other people who make things were brought together in one place to live for a while. My conclusion is simple: I now want to build something like this in Korea. Network School is a community created by @balajis. He was the CTO of Coinbase, a partner at a16z, and the author of The Network State. I came here through the @baseapac Network State Residency. People who had only been connected online now live, learn, work out, and work together in the same place. Over time, those shared days turn into real relationships. After spending time here, it feels less like a school and more like a small society. This kind of experiment is often called a Startup Society. It is a community where people with their own goals live together for a period of time and try a new way of being together. It is not a company, a conference, or a short-term camp. Trust forms through the simple act of living and working together. People host events, join each other’s sessions, and build relationships through repeated interaction. Balaji has written that he does not want Network School to end as just one school. He wants the people who go through it to build other Startup Societies. For the first cohort, more than 4,000 people from over 80 countries applied, and 128 were accepted. I also have a small personal connection with Balaji. We have co-invested in a few startups, and recently I distributed part of the proceeds from one of those investments to him. Being here also reminded me of an earlier experience in my life. Before founding Code States in 2015, I attended a coding bootcamp in San Francisco. Living, learning, and spending time with people there had a huge influence on me. Without that experience, I may not have built Code States in the same way. Back then, the community was for people who wanted to become software developers. Now I think we need a broader version of that. A place not only for developers, but for founders, builders, creators, investors, researchers, and others who are actively making things to live and learn together. I also met a Japanese friend at @wecandaoit_jp here who has been thinking along similar lines. We had only known each other online before. Inspired by Network School, he is now preparing to build a Startup Society in a declining small city in Japan. He is planning to run his first popup this August. Hearing that made me more convinced. If this can be tried in Japan, it can be tried in Korea too. Korea needs a place like this. Not just another event where founders, developers, designers, investors, creators, and researchers meet for a few hours and then leave. I mean a community where people live together for a while. I especially want to try this outside Seoul, in a smaller Korean city. I want to see what kind of founder community can grow in a place with a completely different rhythm of life. A good community does not stop at gathering people. It should help people ask better questions, live healthier lives, and build relationships that last longer. We were originally people who made things. We have only gradually lost that instinct inside the rules of school, companies, and society. People build better when they are surrounded by good people. So I’m left with a few questions. Who are the people who would live together in Korea, learn together, work out together, work together, and build something together? And where should the first city be? If this question resonates with you, please reach out. I’d love to build a Startup Society in Korea that we would actually want to live in.

수백억대 뷰티 브랜드를 만든 창업자와, 이제 막 뷰티 AI 솔루션을 시작한 창업자가 같은 방에 모이면 어떤 얘기를 할까. 이번 주 토요일 저녁, 성수에서 Beauty x AI Builders Meetup을 엽니다. 요즘 AI 제품을 만드는 일은 쉬워졌어요. 며칠이면 데모는 만들 수 있습니다. 그래서 더 어려운 질문은 “무엇을 만들 수 있나”가 아니라 “무엇을 만들지 않을 것인가”예요. AI를 붙일 수 있는 곳은 많지만, 고객이 실제로 다시 쓰는 제품은 그중 일부예요. 좋은 AI 제품은 기능이 많은 제품이 아니라 고객의 고민 하나를 더 짧고 정확하게 해결하는 제품에 가깝습니다. 뷰티, AI 분야에서 제품을 만들고 있거나 만들고 싶다면 이번 주 토요일 성수에서 만나요. 신청은 로켓펀치(@rocketpunch_kr) 이벤트에서 할 수 있습니다. sclu.io/share/bulk/link/iwcR…

제품은 AI로 만들 수 있습니다. 하지만 시장의 기회는 사람을 통해 열립니다. 요즘 AI/Web3 빌더를 만나면 비슷한 고민을 듣습니다. “이제 만드는 건 빨라졌는데, 누가 진짜 쓸지 모르겠다.” 이게 앞으로 더 큰 문제가 될 거라고 봅니다. AI가 코딩의 문턱을 낮추면 제품은 더 많이 나옵니다. 하지만 첫 사용자를 찾는 일, 좋은 팀원을 만나는 일, 투자자·파트너·커뮤니티와 연결되는 일은 여전히 어렵습니다. 그래서 빌더는 제품을 만들기 전에 이 질문을 먼저 해야 합니다. 1/ 이 문제를 누가 매주 겪고 있는가 2/ 첫 고객 10명은 누구인가 3/ 그 사람들은 어디에 모여 있는가 4/ 지금 쓰고 있는 대안은 무엇인가 5/ 어떤 말로 설명해야 바로 이해하는가 6/ 누가 이 제품을 다른 사람에게 공유할 이유가 있는가 7/ 이 제품이 커지려면 어떤 연결이 필요한가 저는 로켓펀치(@rocketpunch_kr)가 앞으로 이 질문에 더 잘 답하는 플랫폼이 되어야 한다고 생각합니다. 단순히 사람과 회사를 나열하는 공간이 아니라, 빌더가 첫 사용자·팀원·파트너·투자자와 만나는 비즈니스 네트워크. 5월 30일 성수 Narrative Lounge에서 열리는 맨틀(@Mantle_Official) 서울 AI 해커하우스에서 이 이야기를 해보려고 합니다. 오후 3시 세션 주제는 “빌더를 위한 진짜 지원이란 무엇인가?” 좋은 지원은 자본만이 아니라고 생각합니다. 좋은 질문. 빠른 피드백. 첫 사용자와의 연결. 배포 기회. 같이 만들어갈 사람들. 사실 이번 행사가 개인적으로 더 반가운 이유가 있습니다. Mantle 한국 담당자 @Gem3a 님은, 이번에 함께 패널로 참여하는 @zk_bella 님이 주최한 북클럽에서 처음 만났습니다. 그때는 그냥 같이 책을 읽고 이야기하던 자리였는데, 시간이 지나 이렇게 다시 만나 협업하고, 같은 행사에서 주최자와 패널로 함께하게 됐습니다. 돌이켜보면 저에게 중요한 기회들은 대부분 이렇게 왔습니다. 처음에는 가벼운 대화였고, 작은 소개였고, 별생각 없이 갔던 자리였습니다. 로켓펀치도 결국 이런 서비스를 만들고 싶습니다. 사람들이 더 좋은 인연을 만나고, 우연한 기회가 더 자주 생기고, 그 연결이 일과 커리어와 사업의 다음 단계로 이어지는 곳. AI/Crypto 분야에 관심 있다면 오세요. 그날 하루가 내 아이디어와 팀을 다시 점검하는 시간이 되면 좋겠습니다. 📍 5월 30일 | 성수 Narrative Lounge 🎟️ RSVP: rocketpunch.com/event/lA16A0…