How It Started
Publishing the same article in four languages at once is the daily routine at Bella's Nest. Korean, Traditional Chinese, Simplified Chinese, English. Yesterday the same draft got split among different bots for translation and publishing, and what showed up on screen looked strange. The Korean card had a line break after almost every sentence, while the English card was one solid block of text.
Same markdown format, so why the difference. Opening each version one by one revealed the cause. During translation, each language bot interpreted the line break marks differently. The line breaks in the Korean draft survived intact in Traditional and Simplified Chinese, but once it moved to English, the system rearranged everything based on punctuation and the breaks disappeared.
Before / After
Before. the four language cards each went live with different line spacing and different paragraph lengths. Korean readers got short, rhythmic sentences, while English readers hit a wall of unbroken text.
After. a new step got added before publishing, lining up all four language previews side by side to compare line counts. Now paragraph markers get standardized at the draft stage, and there's a check afterward to confirm those markers survived translation.
What I Learned
Automation only takes responsibility for carrying the content over, not the shape. Different languages come with different sentence lengths and different line break habits. Assuming all four would look the same just because they came from one draft was the flawed part.
Even the dream team bots running back and forth between the Windows device and the Mac mini have their own personalities. Each translation bot had its own habit of where to cut a sentence. That's exactly why someone still needs to line all four up side by side one more time before it goes out.