98% Isn't the Hard Part. Scaling It Is.

Every feature we ship needs its video copy translated into ten or more languages. Here's how that usually works: write it in English, send it to a translator for each language, wait, check it, repeat every time we launch something. It works, but it means ten people have to be free at the same time or a launch gets stuck. And nobody makes sure the tone stays the same once three different people have touched it.

For the Creator Earning Program, I said we weren't doing it that way. Instead: build a small AI helper in Figma, and give it a clear set of rules to follow. What words to keep as-is, what to actually translate, how the word "earn" changes meaning depending on the sentence, what tone fits the brand in each language. Write the rules once. Let the AI follow them every time.

"Why should we do this?" That was the first thing people asked. Not a no, a real question about whether the setup was worth it.

I didn't have a clean answer beyond believing it was right. This would let the whole design team move faster, and once it worked, it wouldn't just help one feature, it would become how we do all of it. That's a harder thing to sell than a number. I stuck with it anyway.

• 01Wrote the rules: keep product words like "post," "views," "wallet," and "credit" as they are. Translate everything else so it reads naturally. Handle tricky words like "earn" correctly depending on context. Keep the tone short and modern.
• 02First try, Malayalam. A native speaker checked it, told us what was wrong, we fixed it, tried again.
• 03Second try was about 98% right. Sentences read naturally, only a few tricky words were off. We put it into real use: the AI does the work, a writer just checks it.

That part worked well. Then we tried to do the same thing for more languages, and it broke.

"The Figma method didn't work past the first language. More languages meant more screens, more reviewers, and more to coordinate than we could handle."

We tried something lighter instead: take the same rules, upload them to Gemini along with the Malayalam version as an example, then translate into Marathi and Punjabi using a spreadsheet instead of Figma. Easier to check with more people at once. But the result was only about 85% right, worse than the 98% we got on Malayalam.

The problem wasn't the tool. It was which language we used as the example. Malayalam got better when we used Tamil as the example instead of English, because Tamil and Malayalam are close cousins, from the same language family. But for Marathi and Punjabi, we still used Malayalam as the example, and Malayalam comes from a different family than both of those. Wrong example language, lower quality. The fix seems obvious now, but wasn't obvious before we tried: pick your example language from the same family as the one you're translating into.

I'm not calling this a win yet. It's a method that works, but it has real limits, and the limits matter more than the 98% number. Every language still needs a native speaker to check it, and that person can hold up a launch all by themselves. Every new feature needs its own set of rules, sometimes one per language, since the example language matters. Rules go out of date as the product changes, and a fix in one language doesn't fix the others. Right now we use three separate tools, Figma, Gemini, and spreadsheets, glued together by hand. Nobody can easily check if they all still match.

"98% on one language proves the idea works. It doesn't prove it works at ten languages. Those are two different claims, and mixing them up is how you promise something you can't deliver."

Here's what I do stand behind: writing the rules once and letting an AI follow them beats translating by hand, and Malayalam proves it. What's still unsolved is doing this at real scale: too many people needed to check things, rules that need updating, choosing the right example language for each new one, and one clear system instead of three separate tools that don't talk to each other. That's the next problem to solve. It's about how we run the process, not how good the translation is.