I’m really into puzzles.modular.com and have made a few contributions to the repo. I pitched this idea to @Ehsan and he encouraged me to share it with the community, so here it is!
I’m planning to translate the puzzles into Korean and wanted to share my plan.
Why Korean translation?
Mojo/MAX is still under the radar in Korea, but there’s growing government-backed AI investment here. When people start discovering what MAX can do, they’ll look for resources. And honestly, even developers who can read English (like me) often understand things faster in their native language.
On a personal note, this is also about my own growth. Going through each puzzle carefully and translating it is one of the best ways to really learn the material and build deeper expertise.
My plan
Quality & Maintenance
Terminology guide for consistency
GitHub Issues for community feedback
Sync with release tags
Blog series alongside to drive traffic back to the official puzzles
I translated MIT’s Missing Semester into Korean before, but didn’t keep it updated - still bugs me. This time I want to stay in sync with upstream changes properly.
Timeline
Start with Part I to nail down the workflow
Cover all parts currently available
1-2 days per puzzle, approximately 2 months to finish
My initial thought is to use a separate repository to keep the Korean version cleanly separated from the original. But if anyone has a better approach, I’d love to hear it. I want to gather everyone’s input and go with whatever direction makes the most sense.
Updates
I’ll post progress updates in this thread. Feedback, suggestions, or questions are always welcome.
You will undoubtedly learn more as you translate the ideas. If/when fellow Koreans turn to you for clarification, that will also help sharpen your understanding. Whenever you do something a second or third time, you start to see things you didn’t the first time around. Sounds like a wonderful plan.
I like that you’ve scoped it specifically to GPU puzzles. There’s a lot of content at Modular and even taking on a single part of mojo or MAX could be overwhelming.
Thanks! I feel lucky that puzzles happen to be both what I’m most interested in and a scope I can actually finish. It is way better than being too ambitious and never crossing the finish line.
Using an LLM for the first pass. Then I read the original to understand it properly, and go through the translated version sentence by sentence. AI gets a lot right, but there’s always mistranslations or meaning that’s slightly off, those I fix by myself.
Why LLM? Honestly, the puzzles are massive and I didn’t trust myself to stay consistent from start to finish. So, I feed it previous translations as reference and keep a glossary file to make sure terminology doesn’t drift.
Tracking sync status
Simple approach. I track main commit vs translated main commit. If they’re different, then time to translate.
Progress
Done puzzles 1–8. These were familiar from last year so they went fast.
Translation markdown files are under book/i18n/ko/, built via pixi run build-book which chains both EN and KO builds. A language switcher in the nav bar lets readers toggle between English and Korean. A preprocessor tracks source commits and warns when translations fall behind original markdown files.
While translating I found broken docs links in the source, so I submitted PR #222 separately. Once that’s merged, I’ll do a full review pass and submit the translation PR.
The Korean translation PR is up! #225 — please check it out for the details.
Reflections
From Part IV, I started skimming through the docs first to see the big picture before reading carefully. Going one section at a time was too slow. And many times, my questions were answered later or a different approach came up in the next puzzle. Skimming ahead saved my time so much.
I finished faster than I expected. The holiday break helped, but the main reason was that I didn’t want to lose the context I had built up. For me, once I’m focused, stopping and coming back later costs more than just keeping going. The downside is I didn’t review as carefully as I wanted, but I’ll keep reading the puzzles anyway, so I can improve the quality over time.
It’s been a while since I felt this excited about a self-mission, and I’m happy to reach this first milestone. Big thanks to @Ehsan for supporting the idea, and to everyone who cheered me on!
What’s next
I want to go through the puzzles more deeply. Any suggestions or contributions to the Korean translation are always welcome!
! Looking forward to it. Please send PRs when ready.
