The Weaving Void

Visible meaning. Hidden operation. One shared surface.

The Weaving Void is an executable laboratory for artifacts that support more than one reading: prose and program, surface and generator, visible meaning and hidden operation.

The repository began as an archive of esoteric programming languages. Those languages remain as working studies, but the central object is now the weave: the deterministic relation that lets a human-readable surface carry an executable thread.

Start here

• The crossing — a reading room is a self-contained browser entrance: open the downloaded HTML file, switch between the two public readers, and step through sentences, memory and output. Whitespace reflow and a one-letter revision make the rule's limits visible. The page replays six traces from the actual Python runtime; it is not a second interpreter or an arbitrary-prose editor. No server or network is needed.
• The Weave defines the operational spine and the boundary of the project.
• Subtext is the first canonical specimen: ordinary prose deterministically compiles to a Marginalia program and leaves the trace VOID.
• Constructing Subtext takes a first bounded step in the inverse direction: select authored sentences for a target thread. A second surface tells a garden story while carrying the same VOID program.
• Authoring Subtext plans sentence shapes, diagnoses a draft, and proposes repairs using authored candidates.
• Multiple readers gives one crossroads story two public readings, yielding GO or NO through different routing.
• Joint construction selects one surface against several target programs at once, with a reproducible selection witness.
• Marginalia is the first canonical loom, a language of notes, doubt, folding, and verification.

surface --reader/key--> thread --loom/runtime--> trace

The inverse direction is the harder construction problem: given an intended human meaning and a desired program or trace, construct a surface that carries both.

Reproduce the first weave

Rebuild or check the reading room and its embedded witnesses from the existing crossroads specimen:

python3 -m weaves.subtext.explorer
python3 -m weaves.subtext.explorer --check
python3 -m unittest weaves.subtext.test_explorer

The downloadable witness contains the exact surface, profile, extracted program, execution snapshots and SHA-256 provenance. The existing authored surface and profiles are unchanged. Variant hashes distinguish a modified surface from its source; hashes do not authenticate authorship.

python3 weaves/subtext/reader.py weaves/subtext/carrier.md --explain
python3 Marginalia/interpreter.py weaves/subtext/extracted.mrg
python3 -m unittest weaves.subtext.test_reader

The carrier, public reader profile, extracted thread, expected trace, and mutation tests are all committed together. The specimen uses visible sentence features only; it is an experiment in dual reading, not secrecy or cryptography.

Construct another surface

python3 -m weaves.subtext.weaver weaves/subtext/construction/target.mrg --bank weaves/subtext/construction/sentences.json
python3 -m unittest discover -s weaves/subtext -t .

The weaver chooses the first unused sentence that encodes each target instruction, then verifies the complete extraction. The ordered bank contains human-authored alternatives. Changing its order can change the surface without changing the thread. This is finite sentence selection; it does not generate meaning or solve construction for arbitrary programs or traces.

To write a carrier yourself, the authoring workflow gives constraints before you need a sentence bank, then diagnoses your draft:

python3 -m weaves.subtext.authoring plan weaves/subtext/construction/target.mrg
python3 -m weaves.subtext.authoring check weaves/subtext/construction/target.mrg weaves/subtext/construction/draft.md

The supplied draft intentionally differs by one final word; check exits 1 and explains the changed operand. See the authoring guide for repair, structured reports, and the distinction between minimal sentence changes and preserving human meaning.

The archive

The existing languages are retained as studies of different carriers, memory models, and runtimes:

Language	Study	Implementation status
Distaff	Auditory drafts and a bridge from notation to execution	Working
Vortex	Early cyclic tape-machine and syntax study	Working
MandelMemory	Spatial and dynamical memory addressing	Working
SFract	Thread construction through L-system growth	Working
Entanglement	Linked-state variables and destructive observation	Working
Gastronomy	Culinary surface vocabulary over a tape machine	Working
Clockwork	Register-machine computation and synchronization	Working
Bloom	Constraint-oriented continuous simulation	Concept sketch
Marginalia	Annotated memory, compression, and verification	Working / canonical loom

New languages should be added only when they test a property that the current archive cannot: a new kind of surface, reader, loom, trace, or transformation.

Project discipline

A canonical weave must include:

1. a meaningful human-facing surface;
2. a public deterministic reader;
3. the extracted thread;
4. an executable loom;
5. the expected trace;
6. tests for invariance and structural failure.

Zero-width characters, opaque metadata, appended payloads, and secret binary blobs are outside the canonical scope because the visible and executable readings do not depend on the same material.

Running the archive

Interpreters target Python 3.10+:

python3 <LanguageFolder>/interpreter.py <SourceFile>

Bloom additionally depends on NumPy:

python3 -m pip install -r Bloom/requirements.txt

Run the full archive smoke test with:

python3 scripts/smoke_test.py

Related work

• systems-and-intelligence holds the broader Trace-to-Generator and construction-versus-deduction questions. This repository is a narrow executable laboratory for them.
• the-weaving-sound explores generative organization in sound. Distaff is a deliberate bridge, while the projects keep separate runtimes and scopes.

License

MIT License — see LICENSE.