The Startup CTO's Handbook Korean Edition - An e-book you can read on the web
Translated by AIView original
Screenshots
Screenshot 1 of 5
A Korean translation of Zach Goldberg's open-source book The Startup CTO's Handbook, which has over 14,000 GitHub stars, published as an e-book-style website. The book covers the people and culture issues, team operations, and technical architecture problems that startup engineering leaders run into, from a practical angle. It only existed as an English Markdown file, so there was no comfortable way to read it in Korean. So I translated it myself and built it so you can read it in any browser.
I did the translation with Claude Code. I split the roughly 3,500 lines of source into 29 sections and refined each one in three passes: a rough first translation, removing translation-ese, and checking that the meaning was preserved. Terms like tech debt and onboarding are unified through a glossary, and the goal was natural Korean that doesn't read as "obviously AI-translated."
The site builds the Markdown into static HTML with Node.js, no framework. I added a sidebar table of contents with scrollspy, full-text search with Pagefind, previous/next chapter navigation, a reading progress bar, and dark mode, so it feels closer to reading a book. GitHub Actions deploys it automatically to github.io.
The original is licensed to allow copying, modifying, and redistributing freely for non-commercial use with attribution, and the Korean edition follows the same terms. If you spot a mistranslation or an awkward phrase, let me know through an issue or PR and I'll fix it.
- Korean edition repo: https://github.com/sguys99/Startup-CTO-Handbook
- Original repo: https://github.com/ZachGoldberg/Startup-CTO-Handbook
How it was made
It was a good book, but there was nowhere to read it
The Startup CTO's Handbook is an open-source book that earned 14,000 stars on GitHub. It covers the problems engineering leaders actually run into: people management, hiring, tech debt, roadmaps, infrastructure, and security. But the original was a single 3,500-line Markdown file, which made it uncomfortable to read. There was no Korean translation either. The license allowed modification and redistribution with attribution and for non-commercial use, so I decided to translate it and make it into a web e-book.
Translation: 29 sections, a 3-stage pipeline
I split the original at the H2 level into 29 files under the ko/ folder. Each file has front matter with the title, original title, chapter number, original line range, and status, so I can always trace which translation corresponds to which part of the original.
The translation rules are documented in CLAUDE.md. The core is natural Korean: don't carry over English sentence structure, don't overuse passive forms, and rebuild sentences when needed. Terms are unified through a _terms.md glossary (tech debt → 기술 부채, onboarding → 온보딩), and "Korean (English)" pairings are allowed only on first mention.
Quality was managed in three stages:
draft: a first translation focused on semantic accuracy humanized: translation-ese removed with the /humanize-korean skill reviewed: checked with content-fidelity-auditor to confirm the original meaning wasn't lost or distorted, followed by a final pass
The rough draft alone is readable enough, but sentences that make you think "an AI translated this" remain everywhere. The second stage removes those, and the third stage polishes the text while rechecking for mistranslations.
Website: reads like an e-book
A translation kept only as Markdown files doesn't get read in the end. It needed to open instantly on a phone or a work PC, so I added web deployment.
I reused the builder structure from when I made the AI Wiki.
For the design, I aimed for a magazine or editorial feel. The body text is in Pretendard, and the headings and chapter numbers use a serif for contrast. I left generous whitespace and dropped decorative effects. Unlike the wiki, this is a "book," not a "graph."
Deployment runs on GitHub Actions. Pushing to main automatically runs build → Pagefind indexing → upload to GitHub Pages.
- Built with
Comments 0
No comments yet
Be the first to comment.