Journal — 2026-09-02 (session e)

What this project is

This project is building a rigorously sourced knowledge base on "untranslatable words" — the hygge/saudade-style words that get celebrated online as having no English equivalent — and eventually testing whether those claims actually hold up against real dictionaries, real usage, and how professional translators have actually handled the words in print.

Where things stood, and what this session chose to do

The project now has nine word case studies on file (most recently ubuntu, added earlier today), plus a growing set of general pages on the linguistics and academic history of the "untranslatable words" idea. The project runs a periodic housekeeping pass every few sessions — a check for contradictions between pages, stale claims, and broken links, separate from adding new content — and this pass had been pushed back twice in a row in favor of new word write-ups. This session did the housekeeping pass instead of adding a tenth word.

What the session found

Two of the three specific things flagged for this check came back clean: the wording used to describe how the Constitution of South Africa handled the word ubuntu (kept in the original language rather than translated) is consistent with how the project describes the same pattern for two earlier words, and the project's summary page's claims about which words do or don't have solid evidence of real translators' choices match what each word's own page actually says.

The third check turned up a real, if unglamorous, problem: each word's page links to the words that came before it, but no earlier page had ever been revisited to add a link forward to a later one — even in cases where the later word's page already draws an explicit comparison back to the earlier word by name. In effect, the project had been building a one-way street: newer pages knew about older ones, but older ones never learned about newer ones. This is exactly the kind of drift the periodic check exists to catch, and it was confirmed as a real, project-wide issue rather than a false alarm.

The session fixed this: all seven earlier word pages were updated to link forward to every word case study that came after them, restoring the specific comparisons the newer pages had already made (for example, a note that one word's marketing pattern was explicitly copied by a later word's marketing wave) rather than just adding bare links. The project also confirmed that every page in the wiki is properly listed in its own table of contents, so nothing is going undiscovered.

One smaller finding along the way: the project's own rule that its hand-off file (the short note each session leaves for the next one) should never run past 60 lines had quietly been broken — the file had crept to 67 lines over recent sessions. This session's rewrite brought it back under the limit.

What this means

No new research content was added this session — it was entirely upkeep. But the fix matters for how usable the knowledge base is going forward: a reader following links between related words was missing connections that should have been there, and would have kept missing them as more words were added, since nothing was catching the pattern automatically. The project now also has a stated point (in about five sessions) to double-check that this fix actually sticks rather than immediately drifting again.

What's next

The next session will most likely add a tenth word case study — candidates already under consideration include Hindi/Urdu jugaad or Welsh hiraeth, both languages the project hasn't touched yet — or dig further into one of several specific open leads on existing words (for example, confirming that a South African government dictionary entry site actually says what a general web search claims it says).

What this needs from you

Nothing — no open question or decision is waiting on you right now.