Simplified edition decided (hard toggle, fresh component stories) and the VN toggle goes live
Simplified edition + VN toggle (2026-06-19)
Two dictionary threads the same day as the example-sentence work.
Simplified edition — decided
A Simplified edition of an encoding dictionary has a landmine: the stories are built on Traditional forms (開 = 門 + 开; 過 → 过 drops 咼). Resolved across a few corrections:
- Hard T/S toggle, like the Vietnamese language switch. Not side by side — a learner sees only Traditional or only Simplified, never both, no cross-learning. Two parallel self-contained editions.
- The Simplified encoding is authored FRESH from the Simplified components — not derived from Traditional etymology and not teaching the simplification. 过’s story is built from 辶 (walk) + 寸, never mentions 過. Judged by encoding power, folk/mnemonic allowed (many Simplified forms are arbitrary — 开 = 廾 + 一). “How Traditional became Simplified is linguist trivia” (Wedge).
- Glyph-only for v1 (OpenCC
t2s+tw2spfor free vocab swaps); curated mainland-vocabulary localization deferred to v2. - Scope: measured — 982 of 2,662 chars (36%) differ and need a Simplified story; the other 64% are identical and just convert. First phase = the 377 differing chars in A1+A2.
- Lanes: Opus authors the stories; Composer is offloaded the component work (Wedge’s idea) — build a deduplicated inventory of the distinct Simplified components (讠钅纟见贝车马门…, each identity settled once, arbitrary placeholders flagged) + per-char IDS-sourced decomposition cribs, then QA. Author ≠ checker preserved.
Docs: tsumugu/PRD-Simplified-Edition.md + work orders WORK-ORDER-simplified-components-composer.md (cribs) and WORK-ORDER-simplified-encoding-A1-A2-opus.md (stories).
VN toggle — reinstated and shipping
“Added a lot of Vietnamese lately; add an EN/VN toggle and push it public.” Discovery: the EN/VI machine was fully built and then retired — render (render_site.py emits EN/VI siblings under html[data-gloss]), CSS (.glosstog rules still present), search (search.js reads dataset.gloss), UI strings — all there. The code even documented how to bring it back (“reinstate them here + in the nav”).
So this was a reinstatement, not a build: a nav button (.glosstog, EN | VN) + the persisted handler in site.js (ted-gloss, default EN, sets html[data-gloss]). Missing-vi markers changed from a … leak to blank (Wedge: gaps fill soon, don’t clutter). Coverage ~75–76% → ~24% blank in VN mode, accepted. Deploy is host-side (sandbox .git is mount-blocked for deletes; push runs via PUSH-DICTIONARY.command).
Open / parked
scholarNoterenderer still unbuilt (the 苛 illness-loan aside is data-only).- Simplified surface pipeline (OpenCC
t2s+ the toggle wiring) is engineering still to build — the work orders produce the stories, not the toggle.