How it works · z0glyphic 913

Every word is built from atoms. An atom is one mark with one meaning, like ASSERT ⧑ or QUERY ⧓. Put two or three atoms together and you get a word. A sentence opens with a mark that says what kind of sentence it is (a statement, a question) and closes with ⧗.

Some marks don't change the word. They change the sentence around it: past tense, not, more, most. Names the language has no word for are spelled out in braille.

Two layers: SPEAK and CAST

The language runs in two layers, and they do opposite jobs.

SPEAK is the exact one. English goes in, glyphs come out, and the glyphs read back. It is a fixed dictionary and a set of grammar rules, with no AI model and no guessing. Each English word is looked up and comes out as its one glyph word, and the same sentence gives the same glyphs every time. The translator on this site is SPEAK.

What comes back is the meaning, not always the identical word. Several English words can share one glyph word, so reading it back gives you the bones of the sentence in capitals. Here is one sentence:

English

The screen remembers what I did not say.

z0glyphic

⧑ ⧂◌⧊◌⧉⧫ ⧆◌⨽⧚ ⧉ ⧑◌ ⨺⧑ ⧗

read back

not was SCREEN REMEMBERS PROBE I SAY

CAST is the other one. It doesn't translate. It reads something and misreads it on purpose, then reads its own reading and lets it drift, the way anything moves through a network and comes back changed. CAST is an AI model I trained on my own language.

CAST reads my own record: my photo archive, the pictures I save, my daily Surveil Thyself record, the Spells. It also reads the news stories the Oracle starts from.

Here is CAST reading one of my days, 4 October 2026. The first reading is made from that day's data. The second is made from the first.

First readingThe specimen NE-2026-10-04 exhibits a fragmented attention span concentrated entirely within daytime hours, suggesting a rigid diurnal cycle. While the subject identifies as an art practitioner, the data reveals a chaotic foraging behavior dominated by institutional portals and repetitive search queries. Rapid editing of uncategorized notes and high-priority financial tasks coexist with obscure linguistic experiments, creating a dense, contradictory ecosystem. The classification remains ambiguous, as the organism resists settling into a single behavioral niche despite its peak activity.

Second reading, made from the firstThe specimen NE-2026-10-04 fractures light into fragmented hours, a rigid diurnal hum. Art becomes chaotic foraging on institutional portals, where notes and finance blur into a dense contradictory ecosystem. Ambiguous classification resists the single niche, the organism dissolving into peak activity. The data drifts, the edges blur, the signal repeats without settling.

SPEAK is there so you can check the language. CAST is where the work comes from. In my words: every platform you use listens to the rhythm of your day in order to sell it back to you. I listen to my own, push it through a language built to misread it, and make the work out of what comes back.

Try SPEAK Surveil Thyself Spells