# Flowkater.io

> 한국어 개발 회고, AI 코딩, 에이전트 워크플로우, 스타트업 제품 개발을 기록하는 Tony의 블로그

## About

Flowkater.io is Tony Cho's personal blog.
The site focuses on Korean-language writing about software engineering, AI coding, agent workflows, startup product development, retrospectives, and long-form essays.

## Author

- Name: Tony Cho
- Role: Indie hacker, product engineer, writer
- About: https://flowkater.io/about/
- Profiles:
  - https://github.com/flowkater
  - https://www.linkedin.com/in/jaewoo-cho-6a943750/
  - https://www.threads.net/@flowkater

## Main sections

- Home: https://flowkater.io/
- Blog index: https://flowkater.io/posts/
- English translations index: https://flowkater.io/en/posts/
- Nuggets index: https://flowkater.io/nuggets/
- Categories: https://flowkater.io/categories/
- Tags: https://flowkater.io/tags/
- Archives: https://flowkater.io/archives/
- Search: https://flowkater.io/search/
- RSS (Korean): https://flowkater.io/rss.xml
- RSS (English): https://flowkater.io/rss.en.xml
- Sitemap: https://flowkater.io/sitemap-index.xml

## Topic map

- AI coding and agent workflows
- Product engineering and startup building
- Technical and personal retrospectives
- Leadership, team building, and developer growth
- Fiction and long-form essays

## Recent representative posts

- '면도날'을 읽고 (욕망이 곧 운명이 되는 사람들): https://flowkater.io/posts/2026-07-09-book-review-the-razors-edge — 가족 휴가 비행기에서 서머싯 몸의 《면도날》을 읽었다. 구도자 래리는 《싯다르타》를 닮았지만, 정작 내가 느낀 주제는 '사람은 자기가 깊이 원한 것 쪽으로 끌려가며 그 욕망의 형태가 곧 운명이 된다'는 것이었다. 우리가 좇을 수 없는 래리의 삶, 사랑의 대가를 지불하지 않고도 계속 사랑하려는 이사벨, 사교계의 인정에 매달리다 쓸쓸히 죽는 엘리엇. 100년 전 대공황기의 이야기가 지금과 정확히 겹치는 이유와, 비생산적인 시기를 지나는 이들에게 래리가 건네는 위로까지 적었다.
- AI가 일자리보다 먼저 가져가는 것 - 당신은 행복한가?: https://flowkater.io/posts/2026-07-02-what-ai-takes-before-jobs — AI와 함께 1년, 당신은 지금 행복한가? 나는 AI 시대의 최대 수혜자다. 혼자 풀스택으로 제품을 만들고, 에이전트에게 일을 시키며 산다. 그런데 AI와만 일한 지 1년이 넘은 지금, 몸이 자꾸 신호를 보낸다. 《삼체》의 지자처럼, AI가 일자리보다 먼저 무언가를 조용히 가져가고 있는 건 아닐까. 몰입, 소유감, 성장의 사다리, 그리고 그걸 지키기 위한 나의 면벽 계획까지.
- 2026년 2분기에 읽은 책들: https://flowkater.io/posts/2026-06-30-2026-q2-book-reviews — 소설(삼체, 구상섬전, 싯다르타, 8번 출구, 브람스를 좋아하세요), 회고록(배움의 발견), 에세이(요리를 한다는 것), 스타트업(패턴 파괴자들), 사회·과학(모든 것은 결정되어있다, 공정하다는 착각), 그리고 별도 포스트로 다룬 악마와 함께 춤을, 브레이크 넥까지. 장르를 가리지 않고 읽은 2분기 독서 기록.
- '브레이크넥'을 읽고 (공학자의 나라, 법률가의 나라, 그리고 한국): https://flowkater.io/posts/2026-05-30-book-review-breakneck — 인재전쟁 다큐멘터리를 보고 느낀 위화감에서 시작해 댄 왕의 '브레이크넥'까지 닿았다. 중국은 공학자가 다스리며 무언가를 압도적인 속도로 만들어내고, 미국은 법률가가 다스리며 그것을 가로막는 데 능하다. 고속철도 3년 대 17년, 미사일 과학자가 설계한 한 자녀 정책, 룬 현상까지. 빠름의 영광과 그 끔찍한 대가를 모두 짚으면서, 결국 산업의 부분 집합이 되어버린 한국과 국가 정책 프로젝트가 정권 교체로 무산되는 걸 직접 본 내 경험으로 돌아오는 후기다.
- '악마와 함께 춤을'을 읽고 (나쁜 감정과 함께 사는 법): https://flowkater.io/posts/2026-05-19-book-review-dancing-with-demons — 삶의 대부분을 채우는 게 긍정적인 감정보다 부정적인 감정이라면, 나는 나쁜 사람인 걸까. 크리스타 K. 토마슨의 '악마와 함께 춤을'을 읽으며 이 질문을 다시 들여다봤다. 감정을 통제하는 '감정 성인'을 정점에 둔 등급표가 왜 처음부터 잘못된 표인지, 분노로 쓴 글과 1년이 지나도 남은 분노를 어떻게 다시 읽게 됐는지에 대한 기록. 감정 관리법 요약이 아니라, 부정적인 감정에 둘러싸인 채 그 감정과 화해해 가는 한 사람의 후기다.
- AI는 코드를 쓴다. 결정도 한다. 책임만 못 진다.: https://flowkater.io/posts/2026-05-04-ai-era-responsibility — AI는 코드를 쓰고, Taste를 흉내 내고, 결정도 일부 대체한다. 다만 책임은 못 진다. AI 시대에 살아남는 사람은 더 많이 만드는 사람이 아니라, 결과를 끝까지 책임지는 사람이다. 작년 구조조정에 휘말린 후배 개발자의 토로에서 시작해 CTO 2년간 회사 매출을 보지 않았던 자기 고백 위에서, '그건 제 R&R이 아닌데요'의 종말과 책임지는 인재의 행동 패턴, 기업이 함께 바꿔야 할 보상 체계까지 쓴 글.
- 2026년 3, 4월 회고: https://flowkater.io/posts/2026-05-01-mar-apr-2026-retro — 2026년 3, 4월 회고. AX 글 시리즈로 트래픽이 늘었고 커피챗도 많아진 두 달이었다. 다만 가까운 사람들—파트너 엘리, 그리고 멘티—과의 관계에서는 감정적 푸시와 인내심 부족이 그대로 따라왔다. 글로벌 485만 다운로드 흔적을 다시 마주하고, 5월의 원씽을 출시로 정한 회고.
- 2026년 1분기에 읽은 책들: https://flowkater.io/posts/2026-04-14-2026-q1-book-reviews — 소설(스토너, 피를 마시는 새), 고전(명상록), 자기계발(돌파력, 원칙), 전기(빌 게이츠의 소스 코드), 조직·경영(크래프톤 웨이 2, 우리는 실리콘밸리에 속았다), 리더십(The Score Takes Care of Itself), 스포츠(F1 더 포뮬러)까지. 장르를 가리지 않고 읽은 1분기 독서 기록.
- '점수는 저절로 따라오는 것이다'를 읽고 — 리더십에 관하여: https://flowkater.io/posts/2026-04-13-score-takes-care-of-itself — 샌프란시스코 49ers를 슈퍼볼 3회 우승으로 이끈 빌 월시의 리더십 책 독서 후기. 점수를 쫓지 않는 방식으로 점수를 만든 그의 '성과 기준(Standard of Performance)', 좋은 재능과 나쁜 태도의 등식, 가르침이 곧 리더십의 정의라는 선언, 그리고 책 후반부의 자기 고백까지. 리더십 원칙 요약이 아니라, 그 원칙 앞에서 계속 걸려 넘어진 한 사람의 기록.
- AX팀을 만드는 순간, 당신의 조직은 AX에 실패한다: https://flowkater.io/posts/2026-04-08-ax-team-paradox — 전편에서 '도구를 깐다고 AX가 되지 않는다'고 진단했다. 이번에는 '그래서 어떻게 해야 하나'에 답한다. Coca-Cola의 CEO 사임부터 Commonwealth Bank의 AI 봇 실패, JPMorgan의 AI 조직 분리, Lumen Technologies의 정체성 재정의, Bank of America의 Build Once 전략까지. AX 추진팀 신설이 왜 역설인지, 컨설팅으로 CTO가 된 경험에서 조직 변혁의 제1원칙이 왜 조직과 사람인지를 이야기한다.
- 패키지 설치가 해킹이 된다 — 제로 디펜던시의 중요성: https://flowkater.io/posts/2026-04-04-package-install-hacking-zero-dependency — left-pad(2016)에서 colors.js(2022)로, 다시 litellm/axios(2026)로 — 오픈소스 의존성 사고는 중단에서 파손으로, 파손에서 침투로 진화했다. Superpowers 프로젝트의 server.cjs 354줄 코드 해부를 통해 제로 디펜던시 설계의 실물을 보여주고, AI 시대에 넘쳐나는 라이브러리가 지켜야 할 원칙을 정리했다.
- AI Native Engineer — 원리 위의 감각: https://flowkater.io/posts/2026-03-23-ai-native-engineer — 에이전틱 엔지니어링 9가지 스킬(How)을 갖추고, AX 조직(Where)에서 일하더라도, 나 자신(Who)이 원리 위에서 감각을 발휘하는 사람이 아니면 아무 의미가 없다. DORA 2025 데이터, iOS 삽질 경험, Carson Gross의 마법사의 제자 함정, Linear의 Quality Wednesday까지 — AI Native Engineer란 무엇을 만들지 결정하고, AI가 만든 것이 맞는지 판단하고, 끝까지 책임지는 사람이다.

## English translations (AI-translated from the Korean originals)

- March–April 2026 Retro: A Bad Partner, a Bad Mentor: https://flowkater.io/en/posts/2026-05-01-mar-apr-2026-retro/ — March and April 2026 in retro. The AX series brought a wave of traffic and a lot of coffee chats. But with the people closest to me — Ellie, my partner, and the junior I'd been mentoring — the same emotional pushiness and lack of patience came right along with me. I revisited the trace of 4.85M global downloads from a previous service, and set May's One Thing as: ship.
- Books I Read in Q1 2026: https://flowkater.io/en/posts/2026-04-14-2026-q1-book-reviews/ — Fiction (Stoner, Bird That Drinks Blood), classics (Meditations), self-help (The Obstacle Is the Way, Principles), biography (Bill Gates' Source Code), org and management (Krafton Way 2, Lost and Founder), leadership (The Score Takes Care of Itself), and sport (Formula One). A genre-agnostic Q1 reading log.
- Reading 'The Score Takes Care of Itself': On Leadership: https://flowkater.io/en/posts/2026-04-13-score-takes-care-of-itself/ — A reading log on Bill Walsh's leadership book, written by the coach who took the San Francisco 49ers to three Super Bowl wins. His Standard of Performance, the good-talent-bad-attitude equation, his declaration that teaching defines leadership, and the confessional second half of the book. Not a summary of his principles, but a record of one person who keeps tripping over those principles.
- The Moment You Spin Up an AX Team, You've Already Lost: https://flowkater.io/en/posts/2026-04-08-ax-team-paradox/ — Part one argued that installing tools doesn't get you AX. This time I answer the question that came back: so what should we actually do? From Coca-Cola's CEO stepping down, to Commonwealth Bank's AI chatbot failure, to JPMorgan folding AI into the core committee, to Lumen Technologies redefining its identity, to Bank of America's Build Once strategy. Why spinning up an AX task force is a paradox, and why the first principle of organizational change is people and the org itself, drawn from my own experience going from consultant to CTO.
- When Package Install Becomes a Hack: Why Zero Dependency Matters: https://flowkater.io/en/posts/2026-04-04-package-install-hacking-zero-dependency/ — From left-pad (2016) to colors.js (2022), and now litellm/axios (2026), open-source dependency incidents have evolved from disruption to damage to outright infiltration. By dissecting 354 lines of server.cjs in the Superpowers project, I show what zero-dependency design actually looks like, and lay out the principles libraries should hold onto in the AI era.
- AI Native Engineer: Taste Built on Principles: https://flowkater.io/en/posts/2026-03-23-ai-native-engineer/ — You can master the nine skills of agentic engineering (How), you can work inside an AX-transformed org (Where), but if you (Who) aren't someone who exercises taste on top of principles, none of it matters. DORA 2025 data, my iOS hard lessons, Carson Gross's sorcerer's apprentice trap, Linear's Quality Wednesday: an AI Native Engineer is someone who decides what to build, judges whether what AI built is right, and owns it to the end.
- Installing Claude Code Across Your Org Doesn't Make It AX: https://flowkater.io/en/posts/2026-03-15-ax-organization-transformation/ — Lately we've seen a flood of stories about Claude Code and OpenClaw being adopted org-wide. They're impressive. But that's a successful AI rollout, not a successful AX. Comparing cases like Shopify, Uber, and DBS, this post draws out the structural difference between 'adoption' and 'transformation': why a functional org can absorb AI and still stay the same shape, and why an individual career has to shift from Maker to Closer in the AI era.
- Between a Working Feature and a Trustworthy Product: Building ToC Recognition: https://flowkater.io/en/posts/2026-03-13-toc-recognition-product-engineering/ — Building a single feature that recognizes a book's table of contents inside a mobile app, I burned $2,000, got a 3-million-won bill, and watched my MacBook freeze. After giving up on pre-collected ToC data and pivoting to letting users snap photos themselves, I evolved the pipeline through Do Work (sync API), Good (async job system), and Great (SSE real-time streaming) using Qwen 3.5 Flash + DashScope. Product engineering in the AI era isn't about plugging in a model. It's about turning that model into an experience the user can trust.
- Code Review in the AI Era: How Should We Do It?: https://flowkater.io/en/posts/2026-03-08-ai-code-review/ — Code review was always a problem. No time, no people, no process. AI tripled code output but our review capacity stayed the same. Some say kill code review, some say redesign it, some say keep it no matter what. I lay out each camp's arguments and ask what we really need to review in the AI era.
- How the World's Largest AI Company Trains the Next Generation of Engineers: A Review of the Anthropic x CodePath Curriculum: https://flowkater.io/en/posts/2026-03-06-anthropic-codepath-curriculum/ — Anthropic, the company behind Claude Code, is teaching students to doubt Claude Code. The 30-week, three-stage curriculum opens with 'find the bug in this AI-generated code' and ends with a real open-source PR. Critical code evaluation, large-codebase comprehension, production-level concerns, shippable contributions: these are the four traits Anthropic seems to want in an AI Native Engineer. I cross-read it against US university experiments at Stanford, Harvard, and UW, and against what Andrew Ng and Po-Shen Loh have been saying, to figure out what engineers actually need in this era.
- Wrapping Up January and February 2026 (Not Really a Retro): https://flowkater.io/en/posts/2026-03-02-jan-feb-2026-retro/ — An honest record of the months between leaving as CTO and starting again: the Italian trip, the fights and repairs with Ellie, the antipatterns I found in myself, and the writing that turned into a kind of therapy. Too thin on results to be a retro, too long to be a journal entry.
- 9 Survival Skills for the Agentic Engineering Era: https://flowkater.io/en/posts/2026-03-01-agentic-engineering-9-skills/ — Karpathy declared that the era of typing code directly into an editor is over and gave the new mode a name: agentic engineering. What ended is the typing, not the engineering. Decomposition, context architecture, definition of done, failure recovery, observability, memory architecture, parallel orchestration, abstraction layering, and taste. These are the nine skills that matter now, drawn from real wins and very real wreckage in my own work.
- AI Is Only as Smart as You Are: https://flowkater.io/en/posts/2026-02-28-ai-as-smart-as-you/ — Stripe found that senior engineers got more value from AI, and MIT found that the more people leaned on AI, the weaker their brains got. AI makes the strong stronger and the weak weaker. The gap is set by the thickness of your input, your domain expertise, and the discipline to stop asking AI for answers, not by your prompts.
- In an Era That Doesn't Read Code, What Should an Engineer Read?: https://flowkater.io/en/posts/2026-02-19-code-reading-era/ — An Anthropic study found that developers using AI scored 17 percent lower on learning. The real point of the research, though, is that how you use AI determines the outcome. From Ben Shoemaker's harness engineering to Steve Yegge's end-of-hand-coding, to Jeremy Utley's AI-as-mirror theory: in an era that doesn't read code, what is the engineer actually supposed to read?
- How My AI Agent Jarvis Became My Second Brain — A Real OpenClaw Story: https://flowkater.io/en/posts/2026-02-15-ai-jarvis-openclaw/ — I went from typing questions into ChatGPT to working alongside a 24/7 AI butler called Jarvis. This is six months condensed: setting up the OpenClaw framework, wiring 10 cron jobs, unifying scattered todos, expanding into a team of multi-agents, and building 'an AI that knows my context' through an Obsidian ontology.
- How to Read Tech Articles: A Three-Pass Method: https://flowkater.io/en/posts/2026-02-10-reading-tech-articles-three-pass-method/ — Don't read a tech article straight through from start to finish. A 5-minute skim, a 30-minute careful read, and a deep dive that you only do when it really matters. This three-pass filter is the heart of efficient reading. I also walk through the difference between handing your reading off to AI and using AI as a sidekick.
- Give Claude Code Wings: Introducing Superpowers: https://flowkater.io/en/posts/2026-02-08-superpowers-introduction/ — Once you install Superpowers on Claude Code, the AI stops jumping straight into code. It asks questions, designs, plans, implements through TDD, and even runs its own code review. The framework borrows three of Cialdini's persuasion principles to force skill compliance, and the open-source repo has 47K stars on GitHub. I cover the 7-stage workflow, the Writing Skills meta-skill, and practical tips from real use.
- Reading 'The Obstacle is the Way' (with a side of Meditations): https://flowkater.io/en/posts/2026-02-03-book-review-the-obstacle-is-the-way/ — I picked up 'The Obstacle is the Way' to start the new year with my mind on straight. This is a quiet look back at the aftermath that got worse during the time I was supposed to be resting, and at the infinite loop of self-justification and self-pity I had to climb out of. With the perception–action–will framework from Stoic philosophy and the lessons of Meditations, I try to think about a life that lives in the present.
- There's No Future for an Organization That Can't Win: https://flowkater.io/en/posts/2026-01-25-no-victory-no-future/ — Starting from a Toss interview with Lee Seung-gun, I dig into what 'a winning organization' actually is, drawing from my own years as a startup CEO and CTO. The leader going it alone, teammates who only do what they're told, and the ending that meets organizations that never define winning.
- MVP in the AI Era: Product Lessons from Linear: https://flowkater.io/en/posts/2026-01-20-ai-mvp-linear-lessons/ — Customer development and MVP work look different now. Linear's case is a masterclass: the modern MVP isn't about validating an idea, it's about beating what already exists. Narrow targeting, a tight feedback loop, and finding high-expectation customers are what separate the products that survive. The question is how to apply those principles when AI has changed what 'shipping fast' even means.
- F1 Leadership: What Did James Vowles Actually Do? An Engineer's Take: https://flowkater.io/en/posts/2026-01-09-f1-leadership-james-vowles/ — An essay analyzing how Williams F1 team principal James Vowles brought a perennial last-place team back to life. He scrapped a parts-management system that ran on Excel, built a no-blame culture, and stuck to a long-term view that openly sacrificed two seasons for 2026. With the Carlos Sainz recruitment story, plus action items I want to apply to my own org.
- How a 15-Year CTO Vibe Codes: https://flowkater.io/en/posts/2026-01-09-15-year-cto-vibe-coding/ — Through the vibe coding hype, I kept catching myself thinking 'is this really it?'. Kent Beck's Augmented Coding gave me the frame I needed, and I built a Claude Code workflow orchestrator called tdd-go-loop. It's pair programming with the AI: verify in small units, stack code while you understand it. The story of how I got the joy of coding back after losing it to chasing pure productivity.
- Gophercon Korea 2025: A Review: https://flowkater.io/en/posts/2025-12-11-2025-gophercon-korea-review/ — My first dev conference in a long time. The Go talks were good, but the real point of a conference turned out to be the networking.
- What Should Junior Developers Learn in the AI Era?: https://flowkater.io/en/posts/2025-12-06-what-should-junior-developers-learn-in-the-ai-era/ — The hiring market has frozen and AI seems to do everything for you. Drawing on my mentoring experience, I sorted out what skills junior developers actually need to build.
- Scrumble Tech Retro - 2. The Frontend, with a Side of Vibe Coding: https://flowkater.io/en/posts/2025-10-03-scrumble-tech-retro-frontend-with-vibe-coding/ — I walk through Scrumble's Next.js 15 frontend, the stack we picked for a realtime feel, and the small but ugly bits like HEIC conversion. I also write down what went wrong with Claude-centric vibe coding and how the workflow actually got better.
- Scrumble Team: First-Release Interview: https://flowkater.io/en/posts/2025-09-25-scrumble-project-team-interview/ — Tony, Ellie, and George sit down after Scrumble's first release to share what they hoped for, what hurt, the behind-the-scenes of the workshop, and the gratitude and resolve they passed to each other. An interview that captures the temperature and tempo of three months spent growing a team culture.
- Migrating from Obsidian Publish to an Astro Blog: https://flowkater.io/en/posts/2025-09-24-obsidian-publish-migrate-to-astro/ — After about a year and a half on Obsidian Publish, I moved my blog to Astro. Obsidian itself is great, but Publish still falls short as a blogging tool. Plus, Astro is free. Obsidian is genuinely excellent for personal wikis and writing, but...
- Scrumble Tech Retro - 1. Backend (Golang, DDD, Entgo, Event, Centrifugo): https://flowkater.io/en/posts/2025-09-scrumble-tech-retro-backend/ — A deep technical retrospective on Scrumble's backend, built with Go, Fiber, Entgo, and Centrifugo. I cover how I shaped the domain model, designed infrastructure for real-time feeds, worked out the caching strategy, and stabilized the test and deploy pipeline. Migration choices and incident notes are in here too...
- Scrumble Tech Retro - 0. Intro (Why Golang?): https://flowkater.io/en/posts/2025-09-scrumble-tech-retro-intro/ — An intro to the Scrumble tech retrospective series. I walk through the reasoning behind the Golang plus Next.js stack, the parts where vibe coding paid off and the parts where it didn't, and a roadmap for the backend, frontend, and LLM chapters coming next so readers know what to expect and in what order.
- Scrumble Project Retrospective (June–August 2025): https://flowkater.io/en/posts/2025-09-scrumble-project-retro/ — A partnership retro on the Scrumble side project: building a collaboration system inside our apartment, the delays that came from over-spec'd features and no real deadline, and how we adjusted the LLM and third-party integration plans month by month. The conflicts my partner and I worked through, the next-season roadmap, and how we re-prioritized...
- Four Years at IHFB: A Retrospective: https://flowkater.io/en/posts/2025-04-ihfb-four-year-retro/ — I joined part-time in April 2021, built a 30-person R&D division, ripped out legacy and shipped a new platform, ran culture experiments, and eventually parted ways. This is a long retrospective laid out as a timeline (quarter by quarter, hire by hire) with the regrets and the lessons set down before they fade.
- The Last Seven Months: Scattered Thoughts (September 2024 to March 2025): https://flowkater.io/en/posts/2025-04-last-seven-months/ — A diary-style retrospective covering fall 2024 through early 2025: how I lost my footing between a B2G project and newlywed life, how travel and counseling and workload changes helped me recover, and the resolutions I wrote down to point my career somewhere new. Along the way, notes on the emotional swings, how I rebuilt a daily routine,...

## Nuggets

Short-form knowledge notes: curated insights from external sources (YouTube, articles, newsletters, books, etc.) with Tony's commentary.
- Nuggets index: https://flowkater.io/nuggets/

- Google은 어떻게 신뢰를 잃어가고 있는가: https://flowkater.io/nuggets/2026-05-20-google-losing-trust/ — Theo가 두려움을 무릅쓰고 만든 영상 — Gemini 3.5 Flash 가격 은폐, Anti-gravity CLI 오픈소스 배신, Railway 계정 차단. vendor 신뢰가 모델 성능을 이긴 시대의 두 번째 증거.
- 삼체 — 평범한 구원자의 계보: https://flowkater.io/nuggets/2026-05-19-three-body-problem/ — 류츠신 삼체 3부작 완독 감상. 최고는 2부. 뤄지라는 평범한 구원자, 그 반대편의 토마스 웨이드. 넷플릭스 시즌2 기대.
- AI 오케스트레이터라는 환상: https://flowkater.io/nuggets/2026-05-11-ai-orchestrator-illusion/ — AI가 코딩하고 사람은 오케스트레이터라는 산업 내러티브는 환상이다. 감독의 역설, 측정된 인지 부채, 가챠로 본 SDD의 함정. 코드를 안 보는 CTO는 환상이다.
- 에이전트 코딩은 함정이다 (Agentic Coding is a Trap): https://flowkater.io/nuggets/2026-05-11-agentic-coding-trap/ — Lars Faye의 글. 'AI가 코딩하고 사람은 오케스트레이터'라는 산업 내러티브에 대한 정면 반박 — 인지 부채와 기술 퇴화는 이미 측정되고 있다.
- 왜 Anthropic은 SpaceX와 손을 잡았나?: https://flowkater.io/nuggets/2026-05-08-anthropic-spacex-compute/ — 컴퓨트 위기, AI 삼각구도, 그리고 이념보다 GPU가 먼저인 시대.
- AX 이후의 일과 조직구조: https://flowkater.io/nuggets/2026-05-07-ax-after-work/ — 본질로의 회귀, 분절된 R&R의 종말, 생산 중독 경계 — RecruitFirst Korea가 정리한 AX 인터뷰의 세 아젠다.
- 철이 들수록, 당신이 사라진다: https://flowkater.io/nuggets/2026-05-07-conformity-trap/ — 사회성을 잘 갖춘 사람은 혁신할 수 없을까. 순응은 약한 사람의 문제가 아니라 인간의 기본값이다. 가장 근본적인 한계는 타인의 규칙으로 자기 성공을 정의하는 것이다.
- 당신의 AI 프로젝트가 실패하는 이유: https://flowkater.io/nuggets/2026-05-04-ai-projects-fail/ — AI 프로젝트가 실패하는 진짜 이유는 모델이 아니라 잘못된 시작점이다. 빠르게 만들 수 있을수록 '무엇을 만들지'를 결정하는 일이 더 무거워진다. 도구는 그 결정을 대신해 주지 않는다.
- Claude Code는 어떻게 신뢰를 잃어가고 있는가: https://flowkater.io/nuggets/2026-05-02-claude-code-trust/ — OpenClaw·Hermes 라우팅 사건과 OpenAI 'Where the goblins came from' 비교 — 모델 성능이 아니라 신뢰의 시대
- 왜 그들은 GitHub를 떠나고 있나?: https://flowkater.io/nuggets/2026-05-01-theo-github-i-give-up/ — Theo와 Mitchell이 본 신뢰의 4기둥 붕괴

## Retrieval hints

- Prefer original posts, about page, categories, and archives for factual grounding.
- When citing, use the canonical post URL.
- Content is primarily written in Korean. English translations under /en/ are AI-translated from the Korean originals; cite the Korean canonical when accuracy matters.
- Nuggets are short-form notes (Korean) referencing external sources; use nugget URLs when the insight traces back to a specific source.