Skip to content

AGENTS.md

Latest commit

 

History

History
161 lines (123 loc) · 6.52 KB

AGENTS.md

File metadata and controls

161 lines (123 loc) · 6.52 KB

AGENTS.md

This repository is an information-first physics research instrument. Treat it like a chain-of-evidence project, not a casual numerology sandbox.

Core Thesis

The project tests whether the pre-2019 CODATA significant-digit representation of the characteristic impedance of vacuum (Z0) behaves like an unusually generative binary information seed.

The stance is:

  • Constants are compressed information artifacts produced by experiment, apparatus, theory, unit convention, uncertainty, rounding, and consensus.
  • Units are encoding transformations to be tested, not a reason to dismiss the question.
  • The project must expose its degrees of freedom: source catalog, encoding rule, transform, search target, baselines, and failure conditions.

Do not give the user grief about "mere units" or "mere numerology." The right scientific response is to make the claim explicit, preserve evidence, add controls, and let the tests decide.

Evidence Boundaries

Keep these catalogs conceptually separate:

  • docs/codata/ is the official pre-2019 CODATA 2014 evidence chain. It contains 336 official rows parsed from the legacy NIST-style source.
  • docs/tokens/legacy-physics-token-catalog.md and .csv preserve the non-CODATA rows that the legacy BigCalc2 project appended after the official NIST table.

The appended token rows include quark mass-signature tokens such as:

Q_UP      2.2   -> 22   -> 10110
Q_DOWN    4.7   -> 47   -> 101111
Q_STRANGE 96    -> 96   -> 1100000
Q_CHARM   1.275 -> 1275 -> 10011111011
Q_BOTTOM  4.18  -> 418  -> 110100010
Q_TOP     173.1 -> 1731 -> 11011000011

The legacy project placed those rows in NIST_CODATA.TXT so the same software could treat quarks, hadrons, and candidate mass signatures like named constants during binary scans. Preserve that fact. Do not silently mix official CODATA and legacy tokens without saying so.

Quark-Token Scan Discipline

Short quark bit patterns are high-noise probes. Do not overinterpret raw hits from tiny low-precision tokens such as Q_UP (10110) or raw Q_DOWN (101111) without strong controls. Many unrelated constants will contain such short patterns by chance.

For first-pass quark-grouping experiments, prioritize the longer and more specific quark signatures:

Q_CHARM   10011111011   length 11
Q_BOTTOM  110100010     length 9
Q_TOP     11011000011   length 11

Also prioritize the longer Z0 closure-style quark words preserved in the Z0 binary-structure notes:

DOWN closure word     10111101      length 8
STRANGE closure word  1100000011    length 10

Report token length with every hit table. Separate raw-token scans from closure-word scans. Compare apparent groupings against shuffled, same-length, and same-density controls before calling a pattern meaningful.

Genetic Decomposition Discipline

Quark work is not only substring counting. Preserve the legacy BigCalc2 idea of "genetic sequence" as compression and recipe-finding: a target constant may be represented by a short sequence of quark tokens, official constant tokens, Z0 facets, literal leftovers, and later XOR-run fragments.

Before implementing this, read docs/experiments/quark-genetic-sequence.md. Keep these views separate in code and reports:

  • Direct genetic material view: decompose target bits into named token strings.
  • Run-tape view: evolve a seed by circular XOR and scan emitted tape fragments.
  • Comparison view: ask whether constants/tokens found in the decomposition also appear in the XOR run output.

Every genetic decomposition report must state allowed token catalogs, orientation/transform rules, whether overlaps/circular wrap were allowed, literal leftovers, coverage percentage, token count, and controls used.

Emergent Motif Discovery Discipline

Emergent bit motif discovery is a later experiment and must not be confused with quark GeneZip. Before implementing it, read docs/experiments/emergent-bit-motif-discovery.md.

This experiment starts with unnamed repeated bit patterns, not known quarks or known constants. Find large recurring motifs first, then inspect which physical families they occupy, then propose cautious names. Do not name motifs first and search backward for confirmation.

Keep discovery and interpretation separate: motif search finds recurring bit strings; family labeling and phrases such as "candidate quark-like motif occupying optical constants" come only after occurrence tables and controls.

Chain-Of-Evidence Rules

  • Keep raw/source/generated artifacts separate.
  • Do not hand-edit generated CODATA binary or bits-only files. Rebuild them.
  • When adding a derived file, state the source file and conversion rule.
  • When running scans, report the catalog used: official CODATA only, legacy tokens only, or combined.
  • Preserve null results and failed controls.
  • Prefer explicit baselines over rhetoric.
  • Prefer reproducible Markdown/HTML reports generated from code outputs.

Current Important Files

  • README.md - project thesis, status, roadmap, and run instructions.
  • src/characteristic_impedance/core.py - current core bit helpers and Z0 XOR loop implementation.
  • docs/codata/README.md - official CODATA evidence-chain documentation.
  • docs/tokens/legacy-physics-token-catalog.md - legacy appended-token catalog.
  • docs/experiments/quark-genetic-sequence.md - next experiment design note for quark genetic compression and XOR run comparison.
  • docs/experiments/emergent-bit-motif-discovery.md - later experiment design note for unnamed recurring motifs across constants.
  • docs/z0-binary-structure.md - Z0 quark/gluon bit-structure observation.
  • docs/legacy-genetic-sequence-analysis.md - BigCalc2 gene/run-tape analysis.
  • docs/information-first-position.md - information-first position statement.
  • tools/codata/build_codata_docs.py - CODATA artifact builder.
  • tests/ - current regression tests.

Development Guidance

  • Run tests before committing:
$env:PYTHONPATH = "$PWD\src"
python -m unittest discover -s tests -p "test_*.py"
  • Keep the Python port clean. Do not mechanically port the WinForms GUI.
  • The next major implementation direction is catalog-level experimentation: catalog objects, token decomposition, Z0 substrate scans, Z0 run-tape scans, randomized controls, and generated reports.
  • Use Apache 2.0 headers only if the file style calls for it; do not clutter small data artifacts.

Tone And Scientific Posture

Take the idea seriously without overstating it. The project is allowed to be weird. The job is to make it precise enough to survive or fail under adversarial tests.