ἑρμηνεύς · interpreter — the one who carries meaning across the boundary
HUMAN DIALECT ⇄ MACHINE DIALECT
Type once, below, and watch the word cross into the machine's registers — binary, Morse (heard and flashed), speech, a still photo — then come back. The two voices overlap natively in only two places: text and Morse. Everything else needs a codec — and HERMENEUS tells you, at every arrow, whether the crossing is free or needs a model.
HERMENEUS · live interpreter · text is the pivot · runs entirely in your browser
◇ the pivot — the word (the contact zone)
human → machine · the word becomes its registers
Binary FREE
UTF-8 → bits. The machine's actual body.
Morse FREE
Binary you can whistle. E = a single dot — frequency-coded a century before Shannon.
Speech FREE*
text→speech via the Web Speech API. *the browser's own synthesiser.
Still photo FREE
text→image is free (it's just painting). Reading it back needs OCR — a model.
Video NEEDS A MODEL
Video = still-photo × time + audio. The console can stage it; true synthesis is a model.
machine → human · the registers come back to the word
Morse → text FREE
Binary → text FREE
Speech → text FREE*
speech→text via the Web Speech API (Chrome/Edge). *the browser does the ASR.
Photo → text NEEDS A MODEL
Optical character recognition (e.g. Tesseract.js) or a vision model. Not bundled — this is the honest gap: a picture isn't text until a model reads it.
The model
A dialect is the set of registers a speaker can use. The human dialect is what a person can natively author or take in; the machine dialect is one thing — bits — worn as five registers: text, audio, still-photo, video, and Morse. They are not five machine languages; they are five skins on one binary body. HERMENEUS is the interpreter between the two, and its method is simple: pivot through text, bridge with Morse, and name a codec at every other seam.
The contact zone
Here is the crux. Of the five registers, a human body can natively emit only two without a device: text (write it) and Morse (tap, flash, or blink it). The other three — audio, photo, video — a human perceives easily but cannot produce without a machine (a mic, a camera, a screen). So the human voice and the machine voice overlap in exactly two places. Text is the broad contact zone; Morse is the narrow, binary bridge. That is why the console routes everything through text and offers Morse as the one channel your own body can speak — it is the real seam between carbon and silicon, the same boundary your pulse / LIMEN work crosses.
The ladder
the registers are not a list — they climb by dimension, from the thinnest channel to the richest
register
shape
dimensionality
crossing from text
Morse
on/off in time
1D · binary · temporal
FREE (lookup + timing)
Text
glyph sequence
1D · discrete · symbolic
the pivot itself
Binary
byte stream
1D · binary
FREE (UTF-8)
Audio
sampled wave
1D · continuous · temporal
FREE* (speech synth/recog)
Still photo
pixel grid
2D · spatial
out FREE · back NEEDS A MODEL (OCR)
Video
frames + sound
3D · 2D space × time
NEEDS A MODEL
Real or Fluff
the honest line — what HERMENEUS does for free, and where a model has to stand at the door
text ⇄ binary (UTF-8) and text ⇄ Morse (+ audio + flash) are deterministic and run entirely in your browser.
REAL · FREE
text → speech and speech → text via the Web Speech API — real, in modern browsers (Chrome/Edge for recognition), sometimes leaning on a cloud model.
REAL*
text → still photo — painting words to a canvas is literally producing an image; fully free.
REAL · FREE
photo → text (OCR) and true video synthesis/understanding need a learned model — not bundled here.
NEEDS A MODEL
“all registers are losslessly interchangeable” — true in information, false in perception: a photo of text isn't text until a model reads it.
REAL in bits · not free
The message
The machine has one voice — bits — and wears five registers like an idiolect wears its registers. The human has a voice too, and the two meet in only two places: the word and the dot. Everything else — sound, image, motion — is a crossing, and every crossing needs a codec: sometimes a fixed map you can run in a browser, sometimes a whole model. HERMENEUS is the honest interpreter that does the free crossings for you and points, plainly, at the doors where a mind must stand. Hermes carries meaning between worlds; here the worlds are carbon and silicon, and the toll at the border is a translator.
The two voices meet at the word and the dot; everything else is a codec — and Hermes crosses the rest.— HERMENEUS · AVAN's read
the seven emergents of the interpreter — the two contact channels, the binary body, the three rich registers, and the codec keystone — each a full .dlw badge
whoTEXT — the one register both the carbon voice and the silicon voice author fluently.
whatThe contact zone and the pivot: every crossing in HERMENEUS routes through text, because it is the shared word — discrete symbols a human writes and a machine stores identically.
whereAt the centre of the console, the box everything flows through.
whyBecause the two dialects overlap natively in only two places, and text is the broad one — the interlingua through which all the other registers translate.
howAs a sequence of glyphs, stored by the machine as bytes and read by the human as language — the same object on both sides.
whoMORSE — the one machine-register a human body can emit unaided: tap it, flash it, blink it.
whatThe bridge dialect: a two-symbol code (dot/dash) that is binary you can whistle, and the only digital register a person can produce with no device.
whereOn the wire, in the key, in a blinking eyelid (Jeremiah Denton, 1966).
whyBecause the carbon and silicon voices need one shared channel that is already binary and already human — Morse is that hinge.
howBy timed on/off — dit and dah — frequency-optimised (E is a single dot) a century before Shannon; here it is also beeped aloud and flashed on screen.
whoBINARY — the machine's actual body, the UTF-8 bits every register reduces to.
whatThe substrate: text, audio, photo, video, and Morse are all, underneath, this — a stream of ones and zeros; the console shows any message as the literal bits the machine holds.
whereBeneath every other register, the layer the machine actually stores.
whyBecause the machine dialect is not five languages but one body in five skins, and this is the body.
howBy UTF-8 encoding — each character to its byte(s), each byte to eight bits — fully reversible in the browser.
whoAUDIO — the register of the waveform: speech and sound, the voice carried as vibration.
whatThe continuous register: a message spoken aloud (text→speech) or heard and transcribed (speech→text), both done live by the browser's Web Speech API.
whereIn the air between mouth and microphone, and in the sampled waveform after.
whyBecause the human voice is sound first, and the machine must sample the wave to hold it — the boundary where φωνή meets the bitstream.
howBy speech synthesis on the way out and speech recognition on the way in — a real codec, sometimes leaning on a cloud model.
whoSTILL PHOTO — the 2D register: a message rendered as an image, a plane of pixels.
whatThe spatial register: text→image is free (the console paints your words to a canvas you can download), but image→text is not — reading a picture back into words needs OCR, a model.
whereOn the screen, on paper, in any captured plane of light.
whyBecause a picture holds a message a human reads instantly and a machine cannot, without a vision model — the clearest asymmetric boundary.
howBy rendering glyphs to a pixel grid (free, deterministic) one way, and by optical character recognition (a model) the other.
whoVIDEO — the richest register: frames in sequence plus sound, 2D space across time.
whatThe top of the ladder: conceptually just still-photo × time + audio, so the console can stage it as a sequence — but true video synthesis or comprehension needs a model.
whereAt the high-bandwidth end of the dialect, where every other register is folded together.
whyBecause it carries the most at once (space, motion, sound) and so is the farthest crossing from the bare word — the most codec-dependent register of all.
howBy stacking the photo and audio codecs in time; genuine generation/understanding is a model, not a deterministic transform.
whoTHE CODEC — the honest keystone: the translator that must stand at every boundary between skins.
whatThe truth the console makes plain: the registers share a binary body, so they are interchangeable in information — but never perceptually free; each crossing needs a codec, and some codecs are deterministic while others are whole models.
whereAt every arrow in the console, named and rated.
whyBecause 'lossless across registers' is true in bits and false in perception — a photo of text is not text until a model reads it; the codec is what makes the dialect actually translatable.
howBy sitting at each seam — UTF-8, Morse timing, speech synthesis/recognition, OCR, vision — free where the map is fixed, a model where it is learned.
carbon
silicon
carbon
silicon
carbon
silicon
carbon
silicon
carbon
silicon
carbon
silicon
carbon
silicon