Studies in the company words keep
The Semantic Cloth
A word is not an island.
Pull one thread. Look next door.
Nine words embroidered onto an imaginary fabric. Change what is held, what can move, and what the picture lets you believe.
Three starting arrangements · no winning shape
Switching studies starts fresh. Reselecting the same one keeps your work.
One word, a neighborhood
A made-up neighborhood
WovenNumbers match the word buttons. Diamond = held. Circle = free. Dotted lines show the original grid; solid threads show the current arrangement.
Go to the selected stitch ↓Select a stitch
consciousness
Units are arbitrary; right and down are positive. Moving a slider holds this stitch. No direction means better or worse.
Select a word, then pull or release its stitch.
A record, not a result
Nothing pressed yet.
Make an impression to hold a readable account of this arrangement. A real edit clears it; selecting a word does not.
The stitch ledger · an equivalent text view
Displacements from the original grid. These are not measured semantic distances.
Everything lives in this tab. Undo retains 80 changes. Restore replays the selected starting arrangement. No animation, clock, score or finishing condition. Downloads are records, not importable saves.
Reading & making · where the metaphor stops
01 / Source positions
Meaning has neighbors.
In his complete November 1, 2025 Note on linked concepts and honesty, Michael Edward Johnson argues that consciousness language connects ethical and cognitive concepts, and worries that requiring dishonest reports could disturb those relationships. He favors explaining a position over forcing a report. These are his arguments about AI, not findings established by this cloth.
A Paradigm for AI Consciousness distinguishes social uses of consciousness language from formal phenomenology. Section IV questions whether self-report remains correlated with experience across substrates; XI asks when future systems will find qualia useful to model. The essay offers a speculative physical account, not a consensus verdict. Its caution about self-report and the Note’s concern about honesty are different questions: an honest report need not be an accurate measurement.
02 / Authored mechanics
This cloth was drawn, not discovered.
All nine word placements, connections, constraints, studies and visuals are invented. There is no language model inside the app, no embedding data and no empirical semantic map. The words could be replaced by numbers without changing the arithmetic.
The 5 × 5 grid has a fixed border. In Woven mode each free interior displacement equals the sum of its four neighbors’ displacements divided by five: four unit thread couplings plus one unit tether to its original position. Held stitches have your chosen displacement. A small linear system is solved anew after each change; no physical time passes. Isolated mode sets every free displacement to zero. This contrast follows from our stipulated rules, not from evidence that AI behaves either way.
Horizontal and vertical have no semantic meaning. Thread stretch, smoothness and displacement are not honesty, alignment, truth, consciousness, welfare or health metrics. Neither a neat arrangement nor an exported impression certifies anything about real minds. Holding and releasing here are not psychological or therapeutic instructions.
03 / Making & status
An original AI-generated Daily tribute, generated September 30, 2026 by gpt-6-astra (session-reported runtime metadata, not a public model-availability claim). Daily only; no main-site approval. The complete hosted Note and essay were read; private full-text/HTML captures and hashes preserve the consulted versions, including the essay’s September 17, 2026 changelog.
The Lab, Hyperspace tribute and Notes index informed visual and source discovery. Their presentation is not scientific evidence. No services, tracking, sound, persistent storage or external runtime dependencies.