A durable carrier
Latin was copied, taught, disputed, and transported across changing states and technologies. Its survival depended on networks of monasteries, schools, libraries, chanceries, scholars, and scribes, not on parchment alone.
Lingua · institutio · auctoritas
The decoder was.
For centuries, Latin crossed borders and generations as a shared protocol. Its roots, grammar, prose forms, coordinates, and specialist conventions could carry meaning in plain sight. Access to those layers belonged largely to readers trained by monasteries, churches, universities, courts, and states.
Enter the reading roomI · The transmitted surface
Verba patent omnibus; ratio autem sub verbis posita non omnibus aperitur. Qui formam novit, sensum alterum legit.
A reader sees Latin.
A trained reader sees structure.
II · The interpreted mechanism
Words lie open to everyone; the system placed beneath them is not opened to everyone. Whoever knows the form reads a second meaning.
The authority appears miraculous only when the mechanism remains invisible.
Track one · History
Latin was copied, taught, disputed, and transported across changing states and technologies. Its survival depended on networks of monasteries, schools, libraries, chanceries, scholars, and scribes, not on parchment alone.
Grammar governed relationships between words. Rhetoric governed recognizable forms. Scribal rules, tables, ciphers, legends, and coordinates carried additional layers. Language was the carrier for many protocols at once.
Reading letters was not the same as commanding roots, inflection, theology, law, cartography, and inherited commentary. Institutional training determined which layers a reader could recover and which interpretations society would accept.
When religious, scholarly, or state institutions control training and the accepted decoder, interpretation can arrive clothed as revelation or official truth. Hidden mechanics amplify authority.
The surviving evidence
The mechanism was not one invention owned by one order. These artifacts show different ways language, rules, and trained readers moved information.
c. 150 CE · Greek, later Latin
The Geography lists thousands of locations by latitude and longitude and supplies instructions for drawing the known world. Its coordinates survived as text and enabled later readers to construct maps. Jacopo d'Angelo translated the work into Latin in 1406.
1518 · Latin
In the Polygraphia, Johannes Trithemius published tables in which plaintext letters select religious Latin words. The visible result resembles pious language. A reader with the tables recovers the second message beneath it.
1599 · Transnational schools
The Jesuit Ratio Studiorum standardized teaching practices across a network of colleges. It is included here not as the origin of the idea, but as a clear example of an institution reproducing the training required to read, argue, and transmit within a shared intellectual system.
1601 · East Asian correspondence
Bernardo de Angelis composed the Cifra nova Indiae Orientalis for Jesuit superiors communicating during persecution and expulsion from Japan. It shows an actual administrative network adding a cryptographic protocol when ordinary learned language was not enough.
Track two · Function
Type a message. Proto hides its UTF-8 bytes in the lexical choices of devotional Latin prose. The result contains no Base64 and no visible block of ciphertext.
What the trained reader recovers
The manual fax
A drawing can be reduced to a grid and transmitted as ordered marks. The recipient does not receive the original sheet. The recipient reconstructs it from the protocol.
Draw a route
Tap cells to alter the map.
Reconstructed sheet
A verified copy, rebuilt from text.
Ptolemy's coordinate catalog is the historical anchor for this experiment: geographic form transmitted as text plus reconstruction rules. Proto uses a modern raster grid rather than Ptolemy's projection, making the same carrier-and-decoder relationship visible in the browser.
Inside the demonstration
Text becomes UTF-8 bytes.
Two-bit values select grammatically compatible Latin words.
The selections form ordinary-looking devotional sentences.
A checksum detects alteration or copying errors.
Proto I exposes the machinery by representing each byte as a conspicuous Latin token pair. Proto II conceals each byte through four lexical choices: its subject, modifier, object, and verb. The surrounding title, connective phrases, punctuation, and paragraph structure make the payload resemble a devotional meditation.
This is a teaching codec, not encryption: secrecy disappears once the specification is known. That is precisely the lesson. Apparent mystery can be manufactured by unequal access to method.
Both formats carry a two-byte marker, a two-byte payload length, the UTF-8 payload, and a 32-bit FNV-1a integrity witness. Proto I maps the high and low halves of every byte to a visible modifier and noun. Its appearance announces that a code is present.
Proto II divides each byte into four two-bit values. Those values select one of four compatible words for a sentence's subject, modifier, object, and verb. Fixed connective language carries the visible meditation while the variable words carry the concealed message.
The decoder recognizes either form, validates its frame, and verifies its witness. These are teaching codecs, not secure encryption. Their purpose is to compare a visible symbolic system with linguistic steganography that gives normal-looking prose a second reading.
The method returns
A language model repeatedly chooses the next token from several plausible candidates. A watermarking system can subtly favor choices selected by a hidden key. To a person, the passage reads normally. To a detector holding the method, the accumulated choices form a measurable signal.
This is not universal AI detection. It is keyed provenance embedded during generation. It works best across longer, varied passages; substantial rewriting or translation can weaken it. Proto II makes that otherwise invisible mechanism concrete.
Google DeepMind: SynthID text watermarkingResearch: A Watermark for Large Language Models
A note on evidence
The historical argument is broader than any order. Monks, clergy, scholars, diplomats, surveyors, universities, churches, and states all participated in systems of trained transmission. Named examples below document particular implementations; none owns the underlying idea.