아직 오지 않은 것들에 대하여 — 한 줄 시적 문구를 시그니처로 삼는 미니멀 에세이 블로그.https://jacobhan.me/kohttps://jacobhan.me/blog/2026-06-03-what-a-subscription-actually-buys-en/https://jacobhan.me/blog/2026-06-03-what-a-subscription-actually-buys-en/A subscription isn't a product you buy once but a standing influence that reshapes your habits, and the AI subscription is no exception.Wed, 03 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-06-03-what-a-subscription-actually-buys/https://jacobhan.me/blog/2026-06-03-what-a-subscription-actually-buys/구독은 한 번 사고 마는 물건이 아니라 습관을 바꾸는 상시적 영향이다. AI 구독도 예외가 아니다.Wed, 03 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-06-03-ai-burns-electricity-en/https://jacobhan.me/blog/2026-06-03-ai-burns-electricity-en/Software copies for free; an AI burns electricity with every word it generates. Why the era of zero marginal cost is drawing to a close, read off the income statement and the power grid.Wed, 03 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-06-03-ai-burns-electricity/https://jacobhan.me/blog/2026-06-03-ai-burns-electricity/소프트웨어는 복제가 공짜지만 AI는 한 마디마다 전기를 태운다. 복제 비용 0의 시대가 저무는 이유를 손익과 전력 수요로 짚는다.Wed, 03 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-06-03-why-veterans-struggle-with-ai-agents-en/https://jacobhan.me/blog/2026-06-03-why-veterans-struggle-with-ai-agents-en/Engineers trained for decades to remove ambiguity now face probabilistic AI agents. Google DeepMind's Philipp Schmid on the shift in mindset that trips up the most experienced.Wed, 03 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-06-03-why-veterans-struggle-with-ai-agents/https://jacobhan.me/blog/2026-06-03-why-veterans-struggle-with-ai-agents/결정론적 소프트웨어에 단련된 개발자가 확률적 AI 에이전트 앞에서 마주치는 사고방식의 전환. 구글 딥마인드 필립 슈미트의 진단.Wed, 03 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-06-02-what-ai-made-scarce/https://jacobhan.me/blog/2026-06-02-what-ai-made-scarce/AI는 산출량을 폭증시켰지만 사람이 검토할 수 있는 양은 그대로다. 진짜 희소해진 자원은 무엇인가.Tue, 02 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-06-02-what-ai-made-scarce-en/https://jacobhan.me/blog/2026-06-02-what-ai-made-scarce-en/AI exploded the volume of output, but not the amount a person can review. A look at what has truly become scarce.Tue, 02 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-06-02-when-ai-writes-the-code-en/https://jacobhan.me/blog/2026-06-02-when-ai-writes-the-code-en/As code gets written with barely a human hand, what disappears and what survives. Dark factories, the cost of verification, and taste as the real bottleneck.Tue, 02 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-06-02-when-ai-writes-the-code/https://jacobhan.me/blog/2026-06-02-when-ai-writes-the-code/사람이 코드를 거의 쓰지 않는 개발 현장에서 무엇이 사라지고 무엇이 남는가. 다크 팩토리, 검증 비용, 그리고 '안목'이라는 병목.Tue, 02 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-06-02-what-ai-cannot-make-for-you-en/https://jacobhan.me/blog/2026-06-02-what-ai-cannot-make-for-you-en/AI lowered the cost of producing the output, but not the cost of knowing whether it is right. When the bottleneck shifts to verification, the scarce thing is an eye for a domain.Tue, 02 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-06-02-what-ai-cannot-make-for-you/https://jacobhan.me/blog/2026-06-02-what-ai-cannot-make-for-you/AI는 결과물을 만드는 비용은 낮췄지만, 그것이 옳은지 가려내는 비용은 낮추지 못했다. 병목이 검증으로 옮겨간 시대에 희소해진 것은 한 분야를 아는 눈이다.Tue, 02 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-06-02-how-ai-assistants-connect-en/https://jacobhan.me/blog/2026-06-02-how-ai-assistants-connect-en/How AI assistants link to outside services, the arrival of a connection standard called MCP, the early costs it brought, and how those costs are being trimmed.Tue, 02 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-06-02-how-ai-assistants-connect/https://jacobhan.me/blog/2026-06-02-how-ai-assistants-connect/AI 비서가 외부 서비스와 연결되는 방식, 표준 규약 MCP의 등장과 초기 문제, 그리고 그 문제가 손질되어 온 과정을 일반 독자의 눈높이에서 정리한다.Tue, 02 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-06-02-who-buys-when-ai-takes-jobs-en/https://jacobhan.me/blog/2026-06-02-who-buys-when-ai-takes-jobs-en/The sky-high valuations of AI firms presuppose replacing the labor market. The 'AI layoff trap,' its rebuttals, and what a shrinking-population country like Korea changes.Tue, 02 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-06-02-who-buys-when-ai-takes-jobs/https://jacobhan.me/blog/2026-06-02-who-buys-when-ai-takes-jobs/천문학적 몸값의 AI 기업은 노동시장 대체를 전제한다. 'AI 해고 함정'과 그 반론, 그리고 일손이 줄어드는 한국이라는 변수.Tue, 02 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-06-02-who-buys-when-ai-cuts-jobs-en/https://jacobhan.me/blog/2026-06-02-who-buys-when-ai-cuts-jobs-en/When AI replaces workers, the real question isn't whether demand disappears but who ends up with the income — and why the stakes lie in distribution, not raw capability.Tue, 02 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-06-02-who-buys-when-ai-cuts-jobs/https://jacobhan.me/blog/2026-06-02-who-buys-when-ai-cuts-jobs/인공지능이 일자리를 대체할 때 진짜 문제는 수요가 사라지느냐가 아니라 그 소득이 누구에게 가느냐다. 관건은 기술의 성능이 아니라 분배다.Tue, 02 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-06-02-ai-jobs-two-questions-en/https://jacobhan.me/blog/2026-06-02-ai-jobs-two-questions-en/The claim that AI multiplies an expert's skill holds up, but it answers a different question than the one people are actually afraid of.Tue, 02 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-06-02-ai-jobs-two-questions/https://jacobhan.me/blog/2026-06-02-ai-jobs-two-questions/AI가 전문가의 실력을 곱한다는 관찰은 옳다. 그러나 그것은 일자리 수에 대한 답이 아니다. 뒤섞이기 쉬운 두 질문을 갈라 보기.Tue, 02 Jun 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-05-31-choose-boring-tech-en/https://jacobhan.me/blog/2026-05-31-choose-boring-tech-en/AI lowered the cost of producing code but not of verifying it. What "boring technology" means once the bottleneck moves to verification.Sun, 31 May 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-05-31-choose-boring-tech/https://jacobhan.me/blog/2026-05-31-choose-boring-tech/AI는 코드 생성 비용은 낮췄지만 검증 비용은 낮추지 못했다. 병목이 검증으로 옮겨간 시대에 '지루한 기술'이 갖는 새로운 의미.Sun, 31 May 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-05-31-what-you-can-just-say-en/https://jacobhan.me/blog/2026-05-31-what-you-can-just-say-en/How should we talk about the value of humans in the age of AI? Caleb Gross's essay is the starting point; the real question is where one word, 'value', quietly slides between moral dignity and functional usefulness.Sun, 31 May 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-05-31-what-you-can-just-say/https://jacobhan.me/blog/2026-05-31-what-you-can-just-say/AI 시대에 인간의 가치를 어떻게 말할 것인가. 케일럽 그로스의 글을 실마리로, '가치'라는 한 단어가 도덕적 존엄과 기능적 효용 사이에서 미끄러지는 지점을 따져본다.Sun, 31 May 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-05-23-ai-more-work-en/https://jacobhan.me/blog/2026-05-23-ai-more-work-en/Token prices fell 280x, but usage grew 1,000x. AI didn't take work off our hands — it created a new kind of labor: using more, and checking more.Sat, 23 May 2026 11:01:00 GMThttps://jacobhan.me/blog/2026-05-23-ai-more-work/https://jacobhan.me/blog/2026-05-23-ai-more-work/토큰 값은 280배 떨어졌는데 사용량은 1,000배 늘었다. AI는 일을 덜어 주는 대신, '더 많이 쓰고 더 많이 점검하는' 새 노동을 만들어 냈다.Sat, 23 May 2026 11:00:00 GMThttps://jacobhan.me/blog/2026-05-23-apple-vs-intel-en/https://jacobhan.me/blog/2026-05-23-apple-vs-intel-en/Apple and Adobe stayed out of the frontier-model arms race. Instead they bet on renting the engine — and owning distribution, workflow, and the place where value actually moves.Sat, 23 May 2026 10:01:00 GMThttps://jacobhan.me/blog/2026-05-23-apple-vs-intel/https://jacobhan.me/blog/2026-05-23-apple-vs-intel/애플과 어도비는 가장 큰 AI 모델을 직접 만드는 군비경쟁을 피했다. 대신 빌려 쓰고, 가장 잘하는 자리—유통과 작업흐름—를 지키는 쪽에 베팅했다.Sat, 23 May 2026 10:00:00 GMThttps://jacobhan.me/blog/2026-05-07-token-maxing-en/https://jacobhan.me/blog/2026-05-07-token-maxing-en/Token prices fell 280x, and a bragging culture called 'token-maxing' emerged. Tokens are becoming both the currency and the electricity bill of a new era.Thu, 07 May 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-05-07-token-maxing/https://jacobhan.me/blog/2026-05-07-token-maxing/토큰 가격이 280배 떨어지는 동안 '토큰 맥싱'이라는 자랑 문화가 생기고, 토큰이 새 시대의 통화이자 전기요금 청구서가 되어가는 풍경.Thu, 07 May 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-05-06-data-center-load-drop-en/https://jacobhan.me/blog/2026-05-06-data-center-load-drop-en/1,500 MW of data center load disappearing in a single moment in 2024 Eastern U.S. — and NERC's rule change that, after a century, redefined load as an active participant in the grid.Wed, 06 May 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-05-06-data-center-load-drop/https://jacobhan.me/blog/2026-05-06-data-center-load-drop/2024년 미국 동부에서 데이터센터 1,500 MW가 한꺼번에 사라진 사건과, 100년 만에 부하를 능동적 참여자로 다시 정의한 NERC의 규칙 변경.Wed, 06 May 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-04-28-mythos-leak-en/https://jacobhan.me/blog/2026-04-28-mythos-leak-en/Anthropic's Project Capybara and an automated hacking AI called Mythos. What the security academy calls the asymmetry of time, and a vulnerability unique to Korea's air-gap regime.Tue, 28 Apr 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-04-28-mythos-leak/https://jacobhan.me/blog/2026-04-28-mythos-leak/앤트로픽 카피바라 프로젝트와 자동 해킹 AI '미토스'. 보안 학계가 말하는 시간의 비대칭성과 한국 망분리 환경의 고유한 약점.Tue, 28 Apr 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-04-27-what-did-we-buy-en/https://jacobhan.me/blog/2026-04-27-what-did-we-buy-en/AI made color analysis free, and in doing so revealed what we were actually paying for. What AI killed was not the business, but the era when you didn't have to ask what your work was really selling.Mon, 27 Apr 2026 10:01:00 GMThttps://jacobhan.me/blog/2026-04-27-what-did-we-buy/https://jacobhan.me/blog/2026-04-27-what-did-we-buy/AI는 색상 진단을 무료로 풀면서, 우리가 사실 색상이 아닌 무엇을 사고 있었는지를 드러냈다. AI가 죽인 것은 사업이 아니라 자기 일의 정체를 묻지 않아도 되던 시절.Mon, 27 Apr 2026 10:00:00 GMThttps://jacobhan.me/blog/2026-04-27-beyond-the-blanket-en/https://jacobhan.me/blog/2026-04-27-beyond-the-blanket-en/Every time an AI model upgrades, another category of startups gets 'Sherlocked.' But the pattern of who survives is already clear — go deep and narrow, where AI cannot reach.Mon, 27 Apr 2026 09:01:00 GMThttps://jacobhan.me/blog/2026-04-27-beyond-the-blanket/https://jacobhan.me/blog/2026-04-27-beyond-the-blanket/AI 모델이 업그레이드될 때마다 '셜록당하는' 스타트업들. 그럼에도 살아남는 패턴은 분명하다 — AI가 손댈 수 없는 깊고 좁은 영역.Mon, 27 Apr 2026 09:00:00 GMThttps://jacobhan.me/blog/2026-04-27-illusion-of-knowledge-en/https://jacobhan.me/blog/2026-04-27-illusion-of-knowledge-en/The illusion of knowledge created by search and AI. When the border between external information and your own memory dissolves, the danger of 'knowing' what you don't grows quietly.Mon, 27 Apr 2026 08:01:00 GMThttps://jacobhan.me/blog/2026-04-27-illusion-of-knowledge/https://jacobhan.me/blog/2026-04-27-illusion-of-knowledge/검색·AI가 만드는 지식의 착각. 외부 정보와 자기 기억의 경계가 무너질 때, 모르면서 안다고 착각하는 위험이 어디서 자라는가.Mon, 27 Apr 2026 08:00:00 GMThttps://jacobhan.me/blog/2026-04-01-grid-and-data-centers-en/https://jacobhan.me/blog/2026-04-01-grid-and-data-centers-en/AI data center load fluctuations are pressing on the transmission and distribution grids. The grid that holds up our lives is being quietly, but seriously, redesigned.Wed, 01 Apr 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-04-01-grid-and-data-centers/https://jacobhan.me/blog/2026-04-01-grid-and-data-centers/AI 데이터센터의 부하 변동이 송·배전망에 가하는 압박. 우리 삶을 지탱하는 전력망이 조용히, 그러나 진지하게 재설계되는 풍경.Wed, 01 Apr 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-03-28-ancillary-services/https://jacobhan.me/blog/2026-03-28-ancillary-services/2025년 스페인·포르투갈 정전이 일깨운 보조 서비스의 존재. 재생에너지 시대 전력망의 숨은 주역이 어떻게 시장의 모든 참여자에게 열리고 있는가.Sat, 28 Mar 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-03-28-ancillary-services-en/https://jacobhan.me/blog/2026-03-28-ancillary-services-en/The 2025 Iberian blackout made the invisible work of ancillary services visible. In the renewable era, the hidden backbone of the grid is opening up to everyone.Sat, 28 Mar 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-03-23-what-is-skill-in-ai-era-en/https://jacobhan.me/blog/2026-03-23-what-is-skill-in-ai-era-en/Is what you make with AI your skill? When the tool disappears, what's left? The essence of skill lies not in the output, but in the direction.Mon, 23 Mar 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-03-23-what-is-skill-in-ai-era/https://jacobhan.me/blog/2026-03-23-what-is-skill-in-ai-era/AI를 이용한 결과물은 나의 실력인가. 도구가 사라지면 무엇이 남는가. 실력의 본질은 결과물이 아니라 방향에 있다는 이야기.Mon, 23 Mar 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-03-06-the-power-of-doubt-en/https://jacobhan.me/blog/2026-03-06-the-power-of-doubt-en/'Is this really true?' In an age where the line between fake and real has collapsed, critical thinking is the most powerful weapon you can carry.Fri, 06 Mar 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-03-06-the-power-of-doubt/https://jacobhan.me/blog/2026-03-06-the-power-of-doubt/"이것이 정말 사실일까?" 가짜와 진짜의 경계가 무너진 시대, 비판적 사고가 왜 가장 강력한 무기가 되는가.Fri, 06 Mar 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-02-22-neuro-symbolic-ai-en/https://jacobhan.me/blog/2026-02-22-neuro-symbolic-ai-en/Connectionism and symbolism split in 1957. The two currents that looked settled by one side's victory are meeting again at a junction called neuro-symbolic AI.Sun, 22 Feb 2026 09:01:00 GMThttps://jacobhan.me/blog/2026-02-22-neuro-symbolic-ai/https://jacobhan.me/blog/2026-02-22-neuro-symbolic-ai/1957년 갈라진 연결주의와 기호주의. 한쪽의 승리로 끝난 줄 알았던 두 흐름이 뉴로심볼릭 AI라는 합류 지점에서 다시 만나는 풍경.Sun, 22 Feb 2026 09:00:00 GMThttps://jacobhan.me/blog/2026-02-22-does-open-source-pay-en/https://jacobhan.me/blog/2026-02-22-does-open-source-pay-en/How do you make money giving the code away? A recipe-and-restaurant lens on the commercial paradox of open source.Sun, 22 Feb 2026 08:01:00 GMThttps://jacobhan.me/blog/2026-02-22-does-open-source-pay/https://jacobhan.me/blog/2026-02-22-does-open-source-pay/코드를 공짜로 풀어놓고 어떻게 돈을 버는가. 음식 레시피와 식당의 비유로 풀어본 오픈소스의 상업적 패러독스.Sun, 22 Feb 2026 08:00:00 GMThttps://jacobhan.me/blog/2026-02-21-tesla-raw-pixels-en/https://jacobhan.me/blog/2026-02-21-tesla-raw-pixels-en/Tesla trains its self-driving AI on raw, unprocessed image data. In a world where 13 milliseconds of reaction time equals 36 cm of stopping distance, that choice matters.Sat, 21 Feb 2026 11:01:00 GMThttps://jacobhan.me/blog/2026-02-21-tesla-raw-pixels/https://jacobhan.me/blog/2026-02-21-tesla-raw-pixels/가공 전 원시 이미지로 자율주행 AI를 학습하는 테슬라의 전략. 13밀리초의 반응 지연이 36cm 주행 거리가 되는 자율주행 기술 경쟁의 풍경.Sat, 21 Feb 2026 11:00:00 GMThttps://jacobhan.me/blog/2026-02-21-revenue-per-employee-en/https://jacobhan.me/blog/2026-02-21-revenue-per-employee-en/A firm earning $100M with 50 people is fundamentally different from one earning the same with 5,000. Revenue per employee is becoming the new measure of competitive strength.Sat, 21 Feb 2026 10:01:00 GMThttps://jacobhan.me/blog/2026-02-21-revenue-per-employee/https://jacobhan.me/blog/2026-02-21-revenue-per-employee/50명으로 1억 달러를 버는 기업과 5,000명으로 같은 매출을 올리는 기업은 본질적으로 다르다. 1인당 매출이 새로운 경쟁력 잣대로 떠오르는 풍경.Sat, 21 Feb 2026 10:00:00 GMThttps://jacobhan.me/blog/2026-02-21-its-the-data-stupid-en/https://jacobhan.me/blog/2026-02-21-its-the-data-stupid-en/The question isn't 'which model is best?' It's 'do you have data you can trust?' On the real competitive edge in the data-centric AI era.Sat, 21 Feb 2026 09:01:00 GMThttps://jacobhan.me/blog/2026-02-21-its-the-data-stupid/https://jacobhan.me/blog/2026-02-21-its-the-data-stupid/"어떤 모델이 1등이냐"가 아니라 "믿을 만한 데이터가 있는가". 데이터-센트릭 AI 시대의 진짜 경쟁력에 관하여.Sat, 21 Feb 2026 09:00:00 GMThttps://jacobhan.me/blog/2026-02-21-identity-theft-from-trash-en/https://jacobhan.me/blog/2026-02-21-identity-theft-from-trash-en/A British YouTuber's identity-theft experiment showed how digital and physical attacks combine. The sum of what we carelessly leak is the level of our security.Sat, 21 Feb 2026 08:01:00 GMThttps://jacobhan.me/blog/2026-02-21-identity-theft-from-trash/https://jacobhan.me/blog/2026-02-21-identity-theft-from-trash/영국 유튜버 신원 도용 실험이 보여준 디지털·물리 결합 공격. 우리가 무심코 흘리는 정보의 총합이 곧 우리의 보안 수준이다.Sat, 21 Feb 2026 08:00:00 GMThttps://jacobhan.me/blog/2026-02-17-ai-cant-read-korea-en/https://jacobhan.me/blog/2026-02-17-ai-cant-read-korea-en/91% of Korean public documents are in a format AI cannot read. The bottleneck runs deeper than HWP — it's a culture of decoration. America's answer separates authoring from publication.Tue, 17 Feb 2026 00:00:00 GMThttps://jacobhan.me/blog/2026-02-17-ai-cant-read-korea/https://jacobhan.me/blog/2026-02-17-ai-cant-read-korea/공공문서 91%가 AI가 읽을 수 없는 형식. HWP보다 더 깊은 한국의 '치장 문화' 병목과, 작성·배포를 분리하는 미국의 해법.Tue, 17 Feb 2026 00:00:00 GMT