Language
36 notes
- A money-mindsets section, and Tsumugu's reading layer Two threads moved today: the public site grew a money section, and Tsumugu picked up the reading layer that makes it feel like the tools it's meant to replace.
- AI Mandarin Reader — graded reader meets LLM-wiki An idea logged for later: a personal AI reader for Traditional Mandarin — a graded reader crossed with the LLM-wiki pattern, built to teach through real input rather than textbook drills.
- Aim-Shoot-Skin for Language Learning BHS for languages: aim at the just-not-understandable band, shoot the gaps a live passage exposes, skin to compress what accumulates — words learned to solve a live problem outlast words memorized in advance.
- Attention is Important Attention is the difference between real immersion and background noise.
- Character Primer Character-based writing systems become easier to enter when the learner has an orientation layer for components, recognition, and repeated exposure.
- Chinese Characters, Condensed The character cluster as doctrine: one item with three separately-trained links, components classified by function per character, senses rooted in original meanings, and the two practice loops — IME recall and session-staggered pipelining — that run on that structure.
- Empty Components Two historical accidents leave shapes inside characters that do no work. A scribe adds an arbitrary stroke so one glyph writing two words can split in two; the stroke says only "this one is different.
- Form Components Early Chinese characters drew the things they named: 犬 (quǎn, "dog") sketches a dog in profile, 山 (shān, "mountain") peaks, 耳 (ěr, "ear") an ear, 网 (wǎng, "net") a mesh. A form component serves its ho
- Freeflow Immersion Target-language exposure needs a lower-friction mode where attention stays alive without constant tool use.
- Hacking Comprehension Menu Target-language content becomes more useful when the learner can adjust difficulty without abandoning real input.
- How Chinese Characters Work Cluster hub: writing records speech, so characters bind form to sound and meaning; components classified by function per character make the script predictive instead of arbitrary. Reading order, one-line map, and where the cluster plugs into Refold.
- Immersion Metalayers Real input becomes easier to inspect when tools are layered on top of content without replacing immersion.
- Interactive Immersion Real input becomes a learning session when tools and attention are used deliberately instead of passively consuming content.
- Language Isn't Math Language is a pattern system learned through meaningful exposure, repeated examples, and usable phrases in context.
- Mandarin Chinese Language Learning Resources The goal is to create a personal resource catalog, not to reproduce the course. Links and brief resource notes are preserved because the value is in the link list; the course flow is summarized into b
- Meaning Components A drawing of a plow lends power to 勢 (shì, "power"), effort to 功 (gōng, "achievement"), and hard work to 勞 (láo, "toil") — characters with no farming in them. That transfer defines the meaning compone
- Meaning Trees and Original Meanings Every sense a Chinese character carries descends by derivation from an older sense — sometimes one that itself arrived through sound alone — or is such a sound graft itself. The sense the form was inv
- Noticing Game Early immersion becomes less opaque when the learner deliberately hunts for recognizable pieces.
- Preparation Immersion becomes more useful when the learner has enough orientation to notice sounds, words, characters, grammar, and patterns.
- Refold Grammar Primers Grammar helps most when it orients attention during immersion instead of becoming a production rulebook.
- Refold Language Learning System Fluency grows when the language becomes increasingly comprehensible experience until words, sounds, grammar, and phrasing become instinctive.
- Sound Components An unfamiliar character yields two predictions before the dictionary opens: a meaning domain from its semantic component and a syllable range from its sound component. The sound component records what
- Sound Series Centuries of sound change separate modern Mandarin readings from the Old Chinese pronunciations that sound components originally recorded. The drift makes valid links look broken: a learner who meets
- Surface vs Deep Structure Cutting a character into parts stops paying off at the functional level — the components doing a job in that specific character, expressing its sound or its meaning. 部 (bù, "area") splits into 咅 (pǒu)
- The Hán-Việt Bridge A Vietnamese speaker learning Mandarin already owns about three thousand of its morphemes: the Sino-Vietnamese (Hán-Việt) readings buried in everyday Vietnamese words. The bridge's work is to surface
- The IME Method Typing Chinese demands only recognition: feed a pronunciation to a pinyin input method editor (IME) — the software behind every Chinese keyboard — and a candidate window pops up; you pick the form tha
- The Pipelining Strategy Attaching a written form to a word already known by sound and meaning takes a fraction of the effort of learning all three attributes cold. Native literacy runs on this asymmetry: a child learning to
- Three Attributes of a Character A written shape does one job: evoke a spoken word — a sound fused to a meaning — in the reader's mind. Speech therefore carries two attributes, sound and meaning; writing carries exactly three, the fo
- Three Pillars of Language Learning Language learning stays balanced when preparation, interactive immersion, and freeflow immersion each do their job.
- Tsumugu — overnight build, first real read Overnight, yesterday's idea went from PRD to a working build, and today turns it from a scaffold into something to actually read with.
- Tsumugu grows a voice — local Qwen3-TTS, per-sentence Phase 8 stopped being deferred today. What started as "research open-source Traditional-Chinese TTS options" ended with an engine locked, the PRD rewritten, and real audio rendering on this machine.
- Two agents, one reader — Tsumugu's study layer lands Yesterday Tsumugu grew a voice. Today the reader grew into an instrument you can study with — and it was built by two agents working the same repo at the same time.
- Vietnamese calibration brief (for native review) For the native reviewer. Goal: teach the AI what natural modern Vietnamese sounds like, by example — not to hand-translate the whole dictionary.
- Vietnamese Grammar Primer Vietnamese grammar should make immersion easier to notice and understand before it becomes something to memorize for output.
- Vietnamese Language Learning Resources The goal is to create a personal resource catalog, not to reproduce the course. Links and brief resource notes are preserved because the value is in the link list; the course flow is summarized into b
- YouTube Immersion Account YouTube becomes more useful for language learning when the recommendation environment is trained around target-language content.