Langdogu is a language-learning system built around a personal ontology: a map of what you know, what you don’t, and how words connect. It turns what you already read, watch and hear into study material, and keeps one record of your knowledge across all of it, built up over years.
One place for all of your language study.
An early-stage, solo project before incorporation. Built with Claude Code and Codex.
You have learned more than any one app can see.
Every language you have touched has left something in you: a word from a menu, a line from a song, a character you can read in one language and not pronounce in another. None of it is written down anywhere.
None of them keeps one record of what you know.
The missing piece is not another app. It is the record itself: one place where reading informs review, and review changes what you recognise on the next page. Langdogu exists to hold it: every word, every sense, every sentence you met it in, in every language, for as long as you keep learning.
The ambition goes further: sound, script, words and their senses, fixed expressions, grammar, sentences, conversation, speaking and writing, context, the links between languages, and your whole history. That is where Langdogu is going, not what exists today.
The bar every screen, every feature and every line of code is measured against.
Every session shows you something you could not do yesterday: the share of known words rising on your own text, a finish line you can see, a companion that grows only because you learned. No streak counter, no guilt.
A finish line, not an endless feed
SampleEach level’s finish line is the next one’s start.
Being rebuiltKnowing is tracked per sense, across languages, with fixed expressions as units and every word tied to the sentence where you met it.
One word, two senses
SampleYour vocabulary can be opened by Claude and ChatGPT, run on your own server, and taken with you: the full JSON backup keeps everything, and Anki files carry your expressions and their sentences.
Today only the first step is real: Lingua Layer laying Korean over Japanese. The rest is what the code is being built toward.
On the train home
LiveMinseo opens her rail line’s service notice in Firefox. Lingua Layer lays Korean over the Japanese, and one word stops her: 発車, departure.
A moment later
PlannedShe saves it with its sentence. Her map lights the same concept in three languages at once: 発車, 발차 and 发车. She half knew those characters already; now they have a sentence and a place.
In a manga, later
Being rebuiltShe meets 空気を読む. She marked 本を読む, to read a book, as known last month, so Langdogu shows her the same 読む in a sense she does not have yet: to read the room.
That evening
In developmentShe asks Claude what she learned today. Connected to her Langdogu over MCP, it quizzes her from her own sentences. When she gets 空気を読む right, Kkaret celebrates once. If she does not come back tomorrow, nothing scolds her.
Langdogu is built one piece at a time, and every piece carries an exact label. Nothing on this site claims more than exists.
One address for everything. Write about companion setup for Lingua Layer, privacy questions, collaboration, or to hear when the app is ready.
What this site is set in and built with.
The word board is not a font: its dot maps are made when the site is built.
#0A0A0A#3DFFA8#E5C385#146E64