Small enough to inspect. Built to grow.

Research-dossier coverage.

24 of this pilot’s records are members of the 3,500-character first level of the 2013 Table of General Standard Chinese Characters. That is list membership, not a measured frequency rank. The full table has 8,105 characters across three levels. Traditional-form coverage is a separate question; an absent mainland list membership does not imply an uncommon traditional character.

Data uses deliberately pinned Unihan 17.0.0. Property definitions were also checked in UAX #38 revision 41. No HSK coverage or corpus frequency measurements are claimed.

What works in this foundation.

• 8,243 source character entries and 125,400 word records; a split collection with worker search and individual entry loading.
• 32 static, independently addressable portraits, 135 distinct word-reading records and 64 original example sentences.
• Six synchronized atlas lenses, source-backed relation inspection, comparison of up to three records, and shareable exploration links.
• Radical and component indexes, exact and contextual component lookup, and labelled nested paths.
• Unicode-aware character, tone-number pinyin (including zhong1wen2), meaning, word and code-point search; local theme preference and saved entries.
• Modern stroke order with play, pause, restart and step controls for 6,995 characters (Make Me a Hanzi; the 32 dossiers keep their own writing panel).
• Cantonese from CC-Canto (34,335 entries with their Jyutping) and German meanings from HanDeDict (160,731 entries), shown as each source gives them; zimai does not translate one dictionary into another language.

What remains open.

• Single directly consulted secondary historical treatment for most dossiers; named primary scholarship within entries has not generally been read.
• No specimen-level historical image licenses or dated artifact records. History is an attributed evidence view, not a chronological timeline.
• No independently reviewed content or human pronunciation audio.
• Screen-reader and real-device checks (iPhones, Android phones) remain unperformed. Layouts, hover states and animations are checked headless in Chromium and in WebKit, the engine behind Safari.
• 晴’s older spellings and the precise historical transition into the modern 青-containing form require closer comparison. The sound-family graph therefore asserts fewer edges than the modern component view.
• Synced collections, OCR and recorded human audio are future work.

The next 100 research dossiers.

Prioritize common function characters, people and body forms, productive semantic components, and family-building examples. The proposal uses the first-level standard list as a broad common-use proxy, not a fabricated frequency ranking. Family value and current coverage gaps then determine batch order.