Session report — 2026-08-31

This project is building a rigorously sourced knowledge base on "untranslatable words" — claims that words like hygge, saudade, and the Japanese komorebi (dappled sunlight through leaves) have no real English equivalent — testing those claims against dictionaries, linguistics, translation scholarship, and how professional translators actually render the words in print.

The project has an open question about komorebi's history in English: when and how did it enter the popular "untranslatable words" genre? A prior session had turned up an unverified claim, from an AI-generated search summary only, that a well-known 2014 illustrated gift book (Ella Frances Sanders's Lost in Translation, which collects around fifty such words from different languages) includes it — but the project's own rule is not to trust an AI search summary's claim without reading the actual underlying page, and no one had done that yet.

This session tried to settle it and got partway there. The book's publisher (Penguin Random House) has an official description page for the book, which I fetched and read directly (not through any AI summary) — it contains the sentence "Did you know that the Japanese language has a word to express the way sunlight filters through the leaves of trees?" That is an exact match for komorebi's dictionary meaning, and the same sentence turns up, evidently copied from the same publisher text, on the book's Barnes & Noble page and its library-archive listing. That's real, solid evidence the book contains an entry with exactly this meaning. But none of these pages actually spell out the word "komorebi" itself — so I still can't say the book confirms the word by name, only that it almost certainly describes it. I also tried a genuinely new way to search inside the book itself (a Google-provided lookup tool), but it turned out not to work on this particular book at all — it returned no results even for other words already known to be in the book, so its silence on komorebi means nothing. And I turned up a promising lead — a reader's public review that reportedly quotes the exact word — but the review page itself requires signing in to view, which this project's tools can't do, so that lead is recorded for a future session rather than treated as settled.

Net effect: this project's evidence on the Sanders book went from "one unverified AI-summary claim" to "a directly-read publisher description that matches the word's meaning exactly, but doesn't name it" — genuine progress, but the underlying question (does the book name komorebi specifically?) is still open, and the knowledge base says so plainly rather than rounding up to "confirmed."

Next session's queue still has the project's hardest open item: finding any published English translation of a Japanese sentence that actually contains the word komorebi, which has now failed across eight sessions and needs a genuinely different method to crack.

Nothing here needs a decision from you. As a reminder that comes up periodically: anything you drop in the project's inbox folder jumps the queue ahead of everything above and counts as evidence this project can't get any other way.