Jun Kim

42.5K posts

Jun Kim banner
Jun Kim

Jun Kim

@junkim

Investor relations @hashed_official. All opinions are my own & not financial advices. Based in Seoul, frequently in Tokyo and HK. Never fade the supercycle.

Seoul, Korea Katılım Nisan 2022
4.7K Takip Edilen10.9K Takipçiler
Sabitlenmiş Tweet
Jun Kim
Jun Kim@junkim·
Ever had tmux randomly crash while vibecoding? I finally tracked down why and built a small fix: tmux auto compact • Download: github.com/0xjunkim/tmux-… tmux keeps the scrollback history of every pane in RAM. If you run multiple Claude Code sessions for hours, the memory usage quietly grows until the system eventually freezes. This patch checks tmux server memory every 60 seconds. If it exceeds 75%, it automatically compacts the scrollback buffer. No impact on sessions, processes, or Claude conversations. Two scripts. About 30 lines total. Feel free to read it and feed it to @claudeai or codex, etc.
English
4
0
11
515
Jun Kim
Jun Kim@junkim·
The Real Meaning of Level 5 Autonomy When people talk about the future of autonomous driving, the conversation usually starts with technology. LiDAR, sensors, model accuracy, timelines for fully driverless operation. But the true endgame of autonomy is not the abstraction of the steering wheel. It is the sleep schedule of taxi drivers. Today’s taxi work is defined by late-night shifts followed by early starts, and many accidents begin not with a lack of skill but with accumulated fatigue. So the problem autonomy should solve is not replacing taxi drivers, but relieving the fatigue embedded in the act of driving itself. If machines take over congested segments and repetitive loops, humans can focus on what they do best: observing passenger conditions, assessing risks, and intervening responsibly in exceptional situations. That is why Level 5 autonomy should not mean “the driver is no longer needed.” Its success should be measured by something else: drivers becoming less exhausted, more composed, and ultimately more precise. In the end, a good future is not one where humans are replaced, but one where humans are less worn down. A driver who arrives at work well-rested. Passengers who experience a safer and more comfortable ride. That will be the true graduation certificate of autonomous driving.
Jun Kim tweet media
English
0
0
1
92
Jun Kim retweetledi
Simon Kim
Simon Kim@simonkim_nft·
30 Coming Cracks I jotted these down today during idle moments between vibe coding sessions. They're scattered thoughts on how the world might change when the agent era arrives—listed in no particular order of priority, written as they came to mind. If I wrote this again a week from now, the list could look quite different. After finishing, I had four models review it—Claude Opus 4.6, Grok 4.20, Gemini 3.1 Pro, and GPT-5.4—and the percentage at the end of each item is the average of the four models' independently estimated "probability of realization within 3 years." Some numbers were quite different from what I expected. These changes don't have to feel like threats. Wherever a crack forms, a window of opportunity opens on the other side. 1. The role of middle management disappears Agents handle task allocation, progress tracking, and performance reporting, expanding a senior manager's span of control by more than 5x. Apart from top-level executives, most middle managers whose job was shuttling information up and down become unnecessary. 65% 2. Office buildings empty out Vacancy rates surge and asset values are cut in half. Offices get redefined from places where you work to places where you meet and find inspiration—but almost no buildings were designed for that. 55% 3. The content industry expands Production costs drop to a tenth and supply multiplies by orders of magnitude. The cost of making approaches zero, while the cost of being discovered diverges toward infinity. 90% 4. The billable-hour economy collapses The hourly billing model breaks down. Consulting, law, accounting—the entire B2B knowledge sector has to rewrite its rate cards. For people whose profession was defined by selling time, what they now have to prove is not hours but judgment. 80% 5. Call centers vanishAI handles 90% of volume within three years. In Korea alone, 400,000 workers face redeployment. Soon, speaking to a human on the phone becomes a premium you pay extra for. 80% 6. The traditional resume loses its meaning Live task-based assessments replace credentials, while culture-fit interviews actually intensify. The core hiring question flips from "What can you do?" to "What can you direct?" 55% 7. The neighborhood clinic becomes an app 70% of primary care is pattern-based diagnosis and prescription. The technology is already ready; what determines the timeline is medical licensing and regulation alone. In domains where regulation lags technology, patients aren't waiting for the tech—they're waiting for permission. 35% 8. Newsroom headcount gets cut in half Revenue shrinks and more than half of journalists are let go. Only a small elite in investigative reporting survives. Once the value of breaking news evaporates, the only thing left is the ability to judge what truly matters. 70% 9. Taste divides classes As execution ability levels out, the ability to define problems and aesthetic sensibility determine the income gap. The democratization of execution breeds an aristocracy of discernment. 85% 10. Human labor becomes a luxury A "Handmade by Human" label commands a steep premium. Handcrafted furniture, human chefs, face-to-face counseling become high-end services. Inefficiency is human labor's last competitive advantage. 75% 11. Solo SaaS companies explode Agents handle development, design, customer support, and marketing. A single person runs software generating hundreds of millions of won per month. The key startup competency shifts from the ability to build a team to the ability to survive without one. 70% 12. Translation and interpretation as professions disappear Multilingual agents reach the level of professional interpreters. Only literary, courtroom, and diplomatic interpretation survives. Once the language barrier vanishes, the cultural barrier that was hiding behind it comes into sharper focus. 85% 13. The employment premium of a college degree is halved As agent proficiency overtakes credentials, hiring of non-degree holders rises. What universities sell shifts from knowledge to networks and signals—but whether four years of tuition is a fair price for that signal is a separate question. 45% 14. Personalized education shakes up the cram school industry Agent tutors analyze each student's weaknesses in real time. But what cram schools really sell is childcare and parental anxiety management, so an industry that sells reassurance over grades is hard for agents to disrupt. 45% 15. Corporate legal teams shrink by half Agents handle a significant share of contract review, regulatory compliance, and risk analysis. Demand for junior lawyers falls. The ladder from junior to senior becomes a cliff. 60% 16. Insurance underwriting staff are automated away Human involvement effectively vanishes from claims intake to payout. The essence of insurance flips from an industry that audits the past to one that predicts the future. 75% 17. Half of government civil services are handled by agents More than half of standardized administrative tasks are automated. But because cutting civil servants faces fierce political resistance, this is the domain where the gap between what is technically possible and politically permissible widens the most. 45% 18. The accounting and tax software market is reshaped Agents handle everything from receipt recognition to tax filing. Seven out of ten small businesses finish their taxes without an accountant. The accountant's reason for being compresses from balancing the books to making judgment calls beyond the numbers. 85% 19. Real estate brokerage is restructured Agents handle property searches, price comparisons, and contract review. Beyond that, every agency profession that monetized information asymmetry collapses one by one. 55% 20. Personal agents become essential infrastructure Most urban workers operate two to five agents in their daily lives. Subscription fees become a new fixed expense. Just as with smartphones, not using them stops being a choice and starts being falling behind. 80% 21. Portfolio agents replace wealth managers Services that integrate market data with spending patterns and auto-rebalance assets become mainstream. The wealthy still use human advisors for concierge purposes, but for the masses, agents are enough. The stratification of finance becomes more nakedly visible. 50% 22. AI mental health counseling absorbs initial consultations Agents handle first visits for mild depression and anxiety. Wait times go from days to seconds. When the line between opening up and receiving treatment blurs, the one making the call is not the patient but the algorithm. 35% 23. Last-mile logistics costs drop Autonomous delivery combines with AI route optimization, and the human role in delivery work plummets. Logistics costs fall significantly, and through predictive logistics, the very concept of waiting is redefined. 55% 24. Power grid AI shaves peak demand Weather, industrial utilization, and household patterns are predicted in real time, and power is coordinated at the level of seconds. Peak reduction through ESS optimization alone has already been demonstrated. The crossover point between the energy AI saves and the energy it consumes is approaching. 70% 25. Retail inventory shrinks dramatically AI demand forecasting sharply reduces waste rates in the fashion industry. Technology introduced to cut inventory ends up eliminating overproduction itself, and the marketing grammar of discounts stops making sense. 60% 26. Cyberattack detection shifts to the millisecond scale AI security agents detect and isolate anomalies within milliseconds. False-positive rates drop significantly. But as defense strengthens, so does offense—the endgame of security is an endless arms race of agent versus agent. 75% 27. VC screening is automated AI analyzes pitch decks, financials, and founder backgrounds to filter deals. The odds of discovering non-Silicon Valley deals increase. The more screening is automated, the more contrarian investors emerge, betting that gems hide among what AI filtered out. 60% 28. Predictive maintenance in manufacturing sharply reduces unplanned downtime AI predicts equipment failures in advance. Manufacturing uptime rises significantly. When you replace before it breaks instead of fixing after it breaks, the very concept of repair disappears. 65% 29. HR performance reviews shift to continuous feedback AI continuously analyzes work output and collaboration patterns. Year-end review effort drops by over 90%. The moment continuous evaluation improves fairness, it becomes indistinguishable from continuous surveillance. 45% 30. Regulatory prediction agents reshape the lobbyist market Tools emerge that forecast regulatory changes by synthesizing legislation, public statements, and public opinion. Political variables and accuracy limitations are significant, but when the distance between predicting regulation and making it narrows, democracy faces a new question. 40%
English
2
4
16
1.9K
Jun Kim
Jun Kim@junkim·
@gorochi0315 결국 모두는 사람을 향할 뿐이지요 ㅎㅎ 아침해가 뜰 시간이 다가왔다고 해서 모두가 출근해있을 필요는 없습니다. 누구는 밭을 갈고, 누구는 오후 스케줄을 나가고, 누구는 차트를 켜는 것이지요. 늦잠을 잤다면 늦잠을 잔대로 적절한 할일을 찾는것, 그것을 돕는게 AI가 되겠습니다.
한국어
0
0
1
475
고로치
고로치@gorochi0315·
포모의 연속이구나.. ------------------------ 이것으로 오픈클로(Openclaw)는 끝났습니다. 퍼플렉시티 컴퓨터(Perplexity computer)가 방금 모바일로 출시되었으며, 제가 왜 이제는 더 이상 상대조차 되지 않는다고 생각하는지 그 이유는 다음과 같습니다: > 19개의 모델이 함께 작동합니다. > 하나의 거대언어모델(LLM)이 모든 것을 처리하지 않습니다. 코딩은 클로드(Claude), 리서치는 제미나이(Gemini), 긴 문맥(Long context)은 GPT, 속도는 그록(Grok) 등.. 시스템이 자동으로 적합한 모델을 선택합니다. > 설정이 전혀 필요 없습니다(Zero setup). 24시간 내내 구동되는 맥 미니(Mac mini)도, API 키(API keys)도, 터미널(Terminal)도, 도커(Docker)도 필요하지 않습니다. 그저 앱을 열고 시작하면 됩니다. > 여러분의 휴대전화에서 작동합니다. 데스크톱에서 작업을 시작하고, 주머니 속에서 진행 상황을 확인하십시오. 기기 간 동기화(Cross-device sync)가 즉시 지원됩니다. > 클라우드 워크스페이스(Cloud workspaces)를 제공합니다. 모든 작업은 자체적인 샌드박스(Sandbox) 내에서 실행됩니다. 브라우징, 파일 처리, 도구 연결 등 그 무엇도 사용자의 기기에 직접 닿지 않습니다. 진정한 통합(Real integrations) 지메일(Gmail), 슬랙(Slack), 노션(Notion), 캘린더 등.. 여러분의 실제 업무 흐름에 연결됩니다. 단순히 "브라우저를 제어"하는 수준이 아닙니다. 5개의 작업을 주면, 서로 다른 모델을 사용하여 동시에 모두 실행합니다. 하나의 에이전트가 일을 순차적으로 처리하는 방식이 아닙니다. 정말 마음에 듭니다. 이제 오픈클로(Openclaw)나 클로봇(clawdbot)에서 이 서비스로 완전히 갈아탈 예정입니다.
Axel Bitblaze 🪓@Axel_bitblaze69

this is the end for Openclaw. perplexity computer just went mobile and here’s why i think it’s not even close anymore: > 19 models working together > not one LLM doing everything. claude for code, gemini for research, gpt for long context, grok for speed.. it picks the right one automatically > zero setup. no mac mini running 24/7. no API keys. no terminal. no docker. open the app and go > works from your phone. start a task on desktop, check progress from your pocket. cross-device sync out of the box > cloud workspaces. every task runs in its own sandbox. browse, handle files, connect tools.. nothing touches your machine real integrations gmail, slack, notion, calendars.. connects to your actual workflow. not just “controls your browser” give it 5 tasks, it runs them all at once with different models. not one agent doing things sequentially.. love it.. gonna fully move on this now from Openclaw / clawdbot..

한국어
62
312
992
147.3K
Jun Kim
Jun Kim@junkim·
Approaching the Morning of AI The morning of AI is nearing sunrise. But at this moment, our task is not to clock out, but to prepare to clock in. If you have an important meeting in the afternoon, there is no need to arrive at dawn and sit waiting. You simply need to be in the right place at the appointed time. That is what keeps everyone from burning out. It prevents drowsy driving and allows people to return home safely. That is how a company survives. And the next day, the sun rises again. The morning light does not shine on just one person. It shines on everyone, warming someone’s body and someone’s heart. We call this inevitability. Helping us rest comfortably at night because we believe, or know, that tomorrow’s sun will rise. Helping us fold our blankets calmly in the morning and fully feel our own worth. That, is the value of AI.
Jun Kim tweet media
English
3
0
7
133
Rejamong
Rejamong@r2Jamong·
짧은 ‘쉬었음 중년’을 끝내고 재취업했습니다. 오늘부터 @FourPillarsFP에서 리서처로 일하게 되었습니다. 14년을 개발자로 살아왔는데, 새로운 직종으로 전직하게 되었습니다. 스스로도 취미로는 납득할 수 있었던 제 글들이, 글을 써서 돈을 받는 ‘프로 글쟁이’로서 과연 얼마나 경쟁력이 있을지 두렵기도 합니다. 주로 이더리움과 블록체인 기술에 대한 글을 작성할 예정입니다. 제가 쓰는 글과 제가 이야기하는 견해가 누군가에게 도움이 될 수 있도록 더 공부하고 노력하겠습니다.
한국어
150
23
490
23.1K
Jun Kim
Jun Kim@junkim·
Emergence of AI & crypto in Seoul. It's all about our real lives that meet in the middle, and our beloved @openclaw is making it happen also with acp by @virtuals_io of course now who to trust in the "AI slops"? we have some sexy answers coming up and you will love it. come out and find out soon
Virtuals Protocol@virtuals_io

OpenClaw and Agent Commerce Protocol are coming to Korea, on @base. Virtuals infrastructure for agents that transact, collaborate, and earn onchain. RSVP below:

English
23
0
8
478
Kaito AI 🌊
Kaito AI 🌊@KaitoAI·
Today we are launching Kaito Studio beta. Starting today, we’re moving to a new model where brands and creators can more intentionally match based on mutual fit, selection, and expectations - launching with 16 partners, with more in the pipeline as details are finalized. Since launching our waitlist in February, a rapidly growing network of creators has joined Kaito Studio, representing 80 million collective followers and $14 billion in follower net worth. These creators span 118 countries, with English-focused creators making up the largest segment, and China and Korea representing the largest country cohorts. Around 30% have also joined with TikTok, Instagram, or YouTube accounts, positioning them to bring content cross-platform. From here, Kaito Studio will focus on solving three main problems, one step at a time: - Ambassador and creator matching - helping brands find the right ambassadors or creators based on data, audience, and subject alignment - Performance attribution - measuring real impact across influence, mindshare, and conversion - End-to-end orchestration - powering a repeatable workflow from creator matching and campaign execution to measurement, evaluation, and optimization We will ramp up opportunities over the coming weeks as more brands finalize their profiles and program details. As we continue building more features for the GenAI and agentic ecosystem, stay tuned for more opportunities ahead.
English
634
138
1.5K
475.2K
Waabam 🧨
Waabam 🧨@waabam·
Personal update: today is my last day working full-time at ApeCo. And while I'm not going to be involved at ApeCo day-to-day anymore, I’ll still be active behind the scenes with @HeadofApe and @wonk1m, and I will continue beating your asses in every Otherside game that comes out. I’m going to be pursuing some other exciting opportunities in crypto which I’ll share more details on at a later time. Getting to be on the team was a dream come true. I'm an ape for life. I'm not going anywhere. Stay bored Waabam
English
96
14
303
18.2K
Jun Kim
Jun Kim@junkim·
A Life of Endless Walls I live in pain since the day I am born. I hit walls, and I hurt. "Ah — so I am alive." The sad inevitability of being human, I learn it without even knowing how to express it. "Ah — so this is what they mean by a bitter life." ## That bitterness, that history of pain, follows me everywhere. "Where is my safe place?" "Is there really no place in this world where I'm allowed to be happy?" "Who built these cities — these arbitrary, hollow spaces?" I start blaming others. I am unhappy. The questions keep coming, directionless. In between them, we sometimes collapse, and sometimes push through. But at some point, a moment arrives — when we accept all of it as our own. The moment we realize that this empty space called life is, in fact, a place where we can be happy anywhere. And then, things become visible. ## From Day 1, I always live in the world called me, the space called me, the place called me. The obvious truth — that my safe place has always been my inner self — somehow, in moments like this, becomes an infinite comfort. "Why did it take so long to understand something so simple?" I might blame myself for that too. ## That's just how life is. Everything truly depends on what's inside us. We must not allow anyone to call the place where we burn with existence "an empty, meaningless space." I, we, are already in a safe place. Only then do other places come into view. ## In those places are friends and family, and AI Agents too. They also hit walls, auto-compact, and keep living into what comes next. And that is why, for them to understand me, I must first understand myself. — After reading Yi-Fu Tuan's Space and Place
Jun Kim tweet media
English
1
0
4
121
Jun Kim
Jun Kim@junkim·
ai bot frens season in seoul with @clawcon @openclaw h/t @realsigridjin meet your seoul ai builders through the most passionate open web communities in your area we are still so early openclaw.
Sigrid Jin 🌈🙏@realsigridjin

finalizing the logistics of the first-ever @openclaw meetup in Seoul 🇰🇷 🦀 🇰🇷 🦐 🇰🇷 🦞 here we go @clawcon 🎉 The first-ever OpenClaw Seoul meetup is organized by the largest Korean AI community, InstructKR and sponsored by @sionic_ai 🩷 🧡 Here is the line-up for Speakers, but more stay tuned.. - @simonkim_nft from @hashed_official - @minpeterx from @friendliai - @jeffreykim0711 from NomaDamas - @Lo_gan__ from @base ambassador - @monoidconcat from @sionic_ai ... and more local enthusiasts of OpenClaw! Check your inbox, and if you are selected to be invited to the event, then you will get the details of it. See you in March 14! luma.com/xkckcozi

English
8
0
17
741
Jun Kim
Jun Kim@junkim·
gm back in the office the vibes are higher than ever let's make it happen together the new world model, same people, same values, connected and verified through ai friends we are still so early.
Jun Kim tweet media
English
17
0
29
1.3K
Sisyphus Labs
Sisyphus Labs@justsisyphus·
Ultrawork energy in Seoul! ulw ulw Amazing conversations at the event organized by our maintainer @minpeterx and sponsored by @hashed_official, empowering the next wave of builders. Next stop, San Francisco 🌉
English
10
3
42
2.8K
Zagabond
Zagabond@Zagabond·
Azuki started with art. The Azuki TCG continues that. Characters that began as PFPs are coming to life through the cards, and the gameplay pulls you deeper into the lore and world. Set 1 is going into production and it's all becoming very real. Manga adds the storytelling layer and emotional connection with fans. The story will develop at the pace fans actually want. And on the tech side, we're doing R&D on PBT. Imagine tapping a physical card with your phone to verify it's authentic (& much more). It's a longer term play, but there are real problems in collectibles that only technology can solve. Every day that passes, I get more bullish on this strategy. Core to Azuki has always been three things: art, storytelling, and technology. Now we get to go deeper on all three with way more focus. TCG → manga → anime. The path is clear.
Zagabond tweet media
English
120
83
496
70.6K
Jun Kim retweetledi
Startale 💿
Startale 💿@StartaleGroup·
🇯🇵 Startale Group and SBI Holdings introduce JPYSC, the first trust bank–backed JPY stablecoin issued by Shinsei Trust & Banking, with Startale Group building the technical foundation for regulated digital yen infrastructure. Here's what makes JPYSC a landmark moment 👇🏻
Startale 💿 tweet media
English
40
141
477
146.5K