Chengyu, literally set phrases, are overwhelmingly four characters, and scholars can count their sources: the Book of Songs contributes 177, the Spring and Autumn Annals 255, the Analects 173, Mencius 136, the Zhuangzi 161, Sima Qian's Records 265. Most are fossils of a specific scene, a text, a historical embarrassment, frozen into four syllables. Dictionaries list tens of thousands; educated speakers carry a few thousand active. To speak Chinese well is to cite constantly, compressed.
Using the right chengyu signals education; using the wrong one exposes you instantly, because every phrase carries its story. A speaker who says burned books and buried scholars for censorship is quoting Qin Shi Huang's actual purges, knowingly or not. The form itself enforces the canon: four characters, classical grammar, no improvisation tolerated. It is the only major language whose everyday speech requires a working memory of two thousand years of literature.
The claim: chengyu are memes with version control, stories compressed to four-character hashes that verify themselves, because a corrupted idiom stops scanning, so the archive self-heals across centuries where proverbs in freer languages drift. Verdict: the integrity argument is sound, classical idioms change far less than equivalent folk sayings. But compression has a cost: the story fades until only the hash remains, and speakers quote purges, betrayals, and dead strategists without remembering who was betrayed. The checksum survives. The file corrupts silently.
What does it do to a language when a quarter of its wisdom arrives pre-packaged?