a cube of its own. living.
every domain has its own words for the same few things. a notion database says row, column and select option. axes say X, Y and Z. this design said piece, dimension and option; the cubes code says piece, kind and value; anlat says row, column and value. compared side by side they point at the same things, so the words are the pieces here and the domains are one of their dimensions.
we do not have the higher-level name for each of these things yet. until we do, each one is a variable (?1, ?2, ...) with a sentence or two on what it is. as more domains are compared, a name may appear for a variable; then the name replaces it. nothing here is final.
the concepts (variables until they are named)
| variable | what it is |
|---|---|
?0 |
the whole collection: everything that sits together and can be seen through its ways of being the same. |
?1 |
one thing in the collection. seen closer, it opens as a collection of its own. |
?2 |
a way things are the same: a question that can be asked of every thing in the collection. |
?3 |
one answer to that question: how things that are the same in that way still differ. carries a colour. |
?4 |
one thing's answer on one question. when it is missing, it is unknown, never empty by mistake. |
?5 |
a choice of one to three questions to look at the collection through. changing it regroups the things. |
working words, for now
in selim's projects today: row (?1), column (?2) and option (?3), the words of a notion database, because they are the easiest to explain with. code that already uses other words keeps them; this cube maps them.
when selim says X, Y and Z, he means row, column and option: X is the row, Y is the column, Z is the option.
the words, by domain
| domain | ?0 |
?1 |
?2 |
?3 |
?4 |
?5 |
|---|---|---|---|---|---|---|
| working words (selim, 2026-10-29) | database | row | column | option | cell | view |
| axes (selim) | space | X | Y | Z | not known yet | not known yet |
| notion | database | row (page) | column (property) | select option | cell | view |
| relational database | table | row (record) | column (attribute) | value (from a domain) | field | query |
| spreadsheet | sheet | row | column | value | cell | filter |
| cube design 0.1 and 0.2 | cube | piece | dimension | option | placement | look |
| cubes app (UI, 2026-10-29, after a rubik's cube) | cube | piece | face | colour | sticker | not known yet |
| cubes app code (still the old words) | cube | piece | kind | value | placement | not known yet |
| anlat | cube (küp) | row, piece (parça), story | column, lens | value | not known yet | not known yet |
| mosaic home cube | mosaic | app, then tab | none: fixed by code | none | none | side of the cube |
| mosaic blog | the blog | post | dimension | category | placement (with a weight) | not known yet |
| todarch cube (mosaic, 2026-10-30) | todarch | to-do (code: item) | face (a lens, states, rhythms) | circle, state, rhythm | placement (a circle); the item's own field (state, rhythm) | the cube turned to a face |
| agents' usage (mosaic, 2026-10-30) | usage | report (one session's usage) | face (parts, roles, surfaces, scheduled tasks, areas, projects, models, days) | a part, a role, ... | the report's own field | the usage turned to a face |
| todarch | todarch | item | lens | circle | placement | the lens shown |
| OLAP | cube | fact | dimension | member | cell | slice |
| rubik's cube (selim) | cube | piece (corner, edge, centre) | face (up, front, right, ...) | colour | sticker: a piece's colour on one face | how you hold it: the three faces you see |
| 3D space (geometry) | space | point | axis (X, Y or Z) | coordinate (x = 3) | a point's coordinate on one axis | projection onto a plane |
| periodic table | the table | element | group, period, block | group 1, period 3 | an element's group | the layout (long, short, spiral) |
| field biology | a flora | specimen | character (leaf shape) | character state (lobed) | a specimen's scored state (? when unknown) |
identification key |
| linguistics | a language | word form | grammatical category (case, number) | feature value (plural, locative) | gloss (ev-ler-de: PL, LOC) | paradigm |
| library, faceted (Ranganathan) | collection | document | facet | focus (isolate) | a document's focus in a facet | browsing by facet |
| SET (card game) | deck of 81 | card | attribute (colour, shape, number, shading) | value (red, oval, two, striped) | a card's colour | not known yet |
| machine learning | dataset | example | feature | category (level) | an example's value (NaN when missing) |
projection (PCA, an embedding plot) |
| music | score | note | parameter (pitch, duration, loudness) | a pitch (C4), a length (crotchet) | a note's pitch | not known yet |
"not known yet" cells are the point: they show which domain has not been compared on that concept yet.
what the comparison already shows
- understand once. whoever understands rows, columns and select options in notion already understands pieces, dimensions and options in a cube. the understanding transfers; only the words change. that is why the same three or four concepts keep appearing in every project.
- the differences show at once once the sameness is seen: a notion row has one value per select column, while a cubes piece may have several values on one kind; a mosaic tab has no dimensions of its own; the blog weighs a placement (strong, some, a trace).
- axes do not line up the way they first seem. in selim's mapping X is the row, Y the column and Z the option; in a drawn cube each of the three axes is usually a
?2. both readings are kept; the difference between them is a piece of its own (see the findings in the open questions).
ten domains, compared (selim, 2026-10-29)
selim gave three domains, each said as three words one inside the other: a rubik's cube (cube, piece, face), a notion database (row, column, option) and 3D space (X, Y, Z). seven more were picked to be as unlike them and each other as possible. their words are in the table above; here is what each one shows. "not known yet" is kept where a domain has no word.
- rubik's cube. same: a corner sits on three faces at once and one look shows at most three faces: the viewfinder rule, in plastic. a solved face is a question with one answer. differs: "face" is said at two distances, the cube's face (
?2) and a piece's face, the sticker (?4); a twist changes answers by moving pieces, not by editing them; a centre has one sticker and a corner three, so some answers are missing by build, not unknown. why: it is the object cube design is named after. - notion. same: the working words; whoever knows a database knows a cube. differs: the columns are decided before the rows; a select cell holds one option; a row opens as a page, not as a database of its own. why: it is where selim's data lives, and the easiest to explain with.
- 3D space. same: every point answers three questions at once. differs: geometry puts X, Y and Z side by side at one distance (three
?2), while selim's reading puts them one inside the other (row, column, option): three distances. every answer is a number in order, and no point is unknown. why: it is the picture of "three dimensions" everyone carries, and the cube drawing comes from it. - periodic table. same: Mendeleev grouped elements by how alike they behave, decades before anyone knew why (colour before vocabulary), and left empty cells for elements nobody had found, then predicted them (empty is the point). differs: the answer is a place in the table, one per question. why: the best-known cube found from sameness.
- field biology. same: characters are noticed by comparing specimens, and an unknown state is written
?: "not known yet" exactly. differs: a key asks one question at a time and narrows the pile: a zoom, not a turn. why: nobody designed the things being compared. - linguistics. same: turkish evlerde sits on number (plural) and case (locative) at once; a paradigm is a look of two questions. differs: the answer is built into the thing (a suffix you can read off), and some cells are empty because the language has no such form. why: selim learns languages, and grammar is cube thinking.
- library, faceted. same: Ranganathan (1930s) argued against one master hierarchy: a book sits on several facets at once, the first principle ninety years early. differs: the facets are meant to be fixed and universal (five of them), and on the shelf a book stands in one place, so one facet becomes the master again. why: a book must be findable from any side.
- SET. same: three cards make a set when, on every attribute, they are all the same or all different: sameness and difference are the whole game. differs: the deck is full, 3×3×3×3 with every combination once, so nothing is unknown and nothing opens; four questions show at once. why: it trains the eye for sameness.
- machine learning. same: data has many dimensions and one plot shows two or three; a missing value is a known gap (
NaN). differs: many features are found by the machine and have no name at all (an embedding has hundreds): colour before vocabulary at scale. why: everything done with AI sits on it. - music. same: a note sits on pitch, duration and loudness at once; a score shows two of them (pitch up, time across). differs: in music "piece" is the whole (
?0), in cube design one thing (?1); time orders everything. why: a cube for the ear.
what the ten show together. ?1, ?2 and ?3 have a word in every domain; ?4 is often a phrase ("a card's colour") rather than a word, and ?5 is the thinnest: SET and music have none, and biology's key narrows instead of regrouping. missing answers come in two sorts the words do not tell apart: not known yet (Mendeleev's gaps, ?, NaN) and impossible by build (a centre's other stickers, a form a language lacks). and every one of selim's three triples went one level deeper with each word, while geometry's X, Y, Z sit side by side. these are collected as findings in the open questions, not settled here.
the two journeys, by every name (2026-10-29)
principle 17's two journeys have been called many things, and every name is kept (principle 18). each one says something the others do not.
| name | one piece in | the whole cube | where it came from | what it brings out |
|---|---|---|---|---|
| journey one, journey many (chosen for cube design, for now) | journey one | journey many | selim | how many pieces you are looking at |
| journey A, journey B | journey A | journey B | anlat's docs/discovery.md |
that there are two, and nothing more |
| close up, from afar | close up | from afar | suggested in the cubes thread; the cubes code and decisions | distance (principle 6) |
| into a piece, across the cube | into a piece | across the cube | suggested in the cubes thread | direction: inward or sideways |
| zoom in, turn | zoom in | turn | suggested in the cubes thread; the moves in 04-moves.md | the gesture that starts it |
| one name, many names | singular (duygu) | plural (duygular) | principle 11, anlat | the word changes with the distance |
same: every pair names one side from a single piece and the other from the whole cube, and none lets close up lead out to afar. differs: some pairs name a count (one, many), some a distance (close, afar), some a direction (into, across), some a gesture (zoom, turn), and A and B name nothing. not known yet: whether a rubik's cube or notion has its own words for the two (notion's closest: a page opened, a database grouped by a column).
each project speaks its domain's words (selim, 2026-10-29)
a project uses the words its own domain already has, and only those: never words from two domains in one place (a screen, a reply, a document about the project). its row in the table above says what each word maps to, so the projects still understand each other. a domain that has not found its own words yet borrows the most basic ones, cube and piece, and replaces each as soon as its own word is found (anlat is there today). when a new domain starts, first map what its things are onto the variables, then name them; building is easier once the mapping is clear.
the cubes app is about cubes themselves, so its domain is a rubik's cube: the cube, its pieces, and a face (the cube turned to one way its pieces are alike). seen on one piece, the same face shows that piece's colour on it. its UI said "kind" and "value", the words of its own code, and a reply about it said "pieces, kinds and values": three domains in one sentence. that is the mistake this rule names.
the mapping also depends on distance. a row is one piece of a cube; a column, seen closer, is a cube of its own, and its options are its pieces; an option opens as the rows that carry it. which variable a thing is depends on where you are looking from, so a word names a thing at a distance, not for good.
the empty place, by domain
every piece of a cube has the same shape (principle 16), so an empty place is always there, and every domain needs a word for it: cubes says not known yet, anlat henüz bakılmadı (not looked into yet), field biology ?, machine learning NaN. a domain that has no word yet borrows "not known yet". missing by build (a rubik's centre has one sticker) is a different thing and is not shown as empty.
names for the agents sub-app (2026-10-30)
selim's own sub-app in mosaic for his claude sessions at work (sessions and usage). none chosen yet; every one is kept (principle 18).
- agents: the working name, in selim's own words ("ai agent coordination"); plain, says what is inside.
- loom: threads woven into one cloth. not picked: taken by java's virtual threads.
- helm: where one steers. not picked: kubernetes' package manager.
- hive: many workers at once. not picked: taken (apache hive).
- bridge: a ship's bridge, where every station is watched. not picked: selim did not feel it.
- earlier, for sessions alone (2026-10-28): tandem and relay, not felt right; selim wondered about "something that will dissolve and solve itself".
adding to it
when a new domain comes up (a tool, a field of knowledge, a way selim talks about something), add a row to the table: its words for each variable, and "not known yet" where it has none. when a name for a variable appears that fits every row, propose it as a decision; do not rename by yourself.