Releases: andrew867/rgcs
Release list
RGCS v8.5.3
RGCS v8.5.3 — polish: autosave, undo, render control, library filters, neutral naming
Frequency Studio polish: dirty sessions autosave every 30 seconds and
on teardown with a Recover Autosaved Session action, undo/redo covers
layer/segment/title/wobble edits (Ctrl+Z / Ctrl+Y), and the full
shortcut set lands (Ctrl+N/O/S, Ctrl+Shift+S, Ctrl+W, Ctrl+E,
Ctrl+R, Ctrl+Q with a dirty prompt). Renders run off the UI thread
with duplicate-click protection and cancel; import failures state the
exact gate reason with Copy error / Open as text. The Session Library
gains search (title, category, tags, id, frequency), Factory/User/
Imported/Recent/Favorite filters, sorting, and favorites. Exports get
collision control (overwrite or auto-increment), reveal-folder, and
receipts carrying the app version, git commit, and input hash; the
Phryll v2 single-file export asks where to write and shows the full
path. The New Session page is redesigned around a Session Summary
card with the recipe JSON behind a toggle.
The curated session library uses neutral naming throughout — no
vendor or device branding in any filename, ID, or text field — and
older factory directories migrate safely to the workspace trash.
Factory updates write a prior-hash backup manifest before touching
anything.
The release pipeline now runs a 9-step scenario smoke inside the
frozen build on every platform (create workspace, open the curated
Schumann session, save as user session, render a 10 s preview, export
the session PDF, export a Phryll STL only, close and reopen) plus a
five-scenario first-run/upgrade matrix. All builds unsigned per
project policy.
Designs, estimates, and audio recipes remain model outputs and
reproducibility records. Measurements decide.
RGCS v8.5.2
RGCS v8.5.2 — Frequency Studio CRUD, curated curated corpus library, workspace lifecycle
Frequency Studio becomes a real editor: sessions have New / Open /
Open Recent / Save / Save As / Close / Duplicate / Delete-to-trash /
Import with dirty-state prompts, a stable session identity, and a
Session Library page over the workspace library. 61 curated
curated binaural sessions from an imported frequency corpus (Schumann, sleep, focus/memory,
relaxation/emotional, chakra/pineal, altered-state families) ship as
factory content with a per-file-hash manifest; factory sync is
idempotent — it adds missing files, updates only unmodified factory
copies, and never touches user-edited files. Source-language claimed
uses are recorded, not endorsed; nothing here is a medical claim.
Workspace lifecycle is fixed end to end: first runs and reinstalls
never crash when the workspace folder already exists, a failed
open/create leaves the current workspace usable, and Close / Switch /
Reveal / Repair commands tear down playback and jobs properly.
Exports are selectable by file type: Frequency Studio writes only the
checked types with the expected files listed first, and Phryll v2
exports single artifacts (STL-only among 12 kinds) without a bundle.
Phryll v1 is retired from the UI; v2 is the only Phryll path. Every
PDF sheet's section rules now sit under their text instead of striking
through it, and all demo sheets are regenerated.
The tag-triggered release pipeline now runs the full test suite and
leak firewall as a validate gate before building Windows (portable
zip + per-user installer), Linux, and macOS packages — each
smoke-checked, manifested, and SHA-256'd. All builds unsigned per
project policy.
Designs, estimates, and audio recipes remain model outputs and
reproducibility records. Measurements decide.
RGCS v8.5.1
RGCS v8.5.1 — wobbles, corrected lattice coils, session import, automated release builds
Frequency Key Studio gains wobble stage-table modulation: 38 packaged
presets (feathering, staged %, octave/odd/Fibonacci ladders,
golden-ratio and PHI ramps) applied per-layer to carrier or beat with
phase-integrated click-free stepping and a Nyquist guard. External
frequency_session JSON files import and render via
'rgcs-sonic render-file' (documented sample in examples/sonic/).
Phryll Generator v2 coils are corrected to the reference geometry:
copper-CW / silver-CCW multi-start families at +/-45 degrees crossing
in an X lattice centered on the Eye plane, carried through
SCAD/STL/3MF/DXF and the build sheet; crystal-bottom coupling model
(open bottom enforced, O-ring compliant-mount records), source-language
entries, hardware-first excitation ordering. OpenSCAD integration
covers standard Windows install locations and the nightly Manifold
backend, with stated render timeouts.
This release is built and published automatically by the tag-triggered
release workflow: Windows portable zip + per-user Inno Setup installer,
Linux tar.gz, and macOS tar.gz — each smoke-checked, manifested, and
SHA-256'd. All builds unsigned per project policy.
Every exported artifact keeps its claim boundary: designs, estimates,
and audio recipes are model outputs and reproducibility records.
Measurements decide.
RGCS v8.5.0 — Frequency Key Studio (Sonic Lab) + Phryll Generator v2
RGCS v8.5.0 — Frequency Key Studio (Sonic Lab) + Phryll Generator v2
Two new Design Studio modules in one release: a complete audio creation and recipe system (Frequency Key Studio, with the full v1.1 and v1.2 feature sets), and a crystal-first parametric CAD generator (Phryll Generator Designer v2).
Phryll Generator Designer v2
Enter your measured crystal (length, top/base diameters, max width, facets, Eye coordinate + uncertainty) and get a bespoke cone and coil sleeve — never a scaled stock mesh (enforced by test):
- Custom cone — inner = crystal envelope + clearance, outer = inner + wall, per-station fit check (demo: 26/39 mm crystal + 0.66 clearance + 1.8 wall → inner 40.32→27.32 mm, outer 43.92→30.92 mm, fit PASS).
- Eye-aligned ±45° crossed lattice coils — copper CW at +45° and silver CCW at −45° as multi-start families (83 starts per coil on the demo, rise one circumference per turn), crossing in an X centered exactly on the Eye (demo residual 0.0 mm; the crystal midpoint is 2.5 mm away and is never used). AWG 28 spacing per the source rule measured perpendicular to the strands: clear gap 0.66 mm, strand pitch 0.99 mm, coil-center standoff 2.375 mm (≈7.2 wire Ø).
- Full export bundle — deterministic SCAD module set (cone, sleeve, continuous twist-extruded helical wire slots, Eye marker, base adapter, cap, LED holder, jack holder, locker, annular pickup ring), STL + 3MF from a built-in mesh backend — including the grooved circular coil sleeve with the crossed wire slots carved in (104 976 triangles; OpenSCAD optional), SVG/DXF templates, compatibility + build PDFs, JSON receipts,
CHECKSUMS.sha256-verified bundle — demo attached. - Crystal-bottom coupling — open/lightly-coupled gap → flat pickup surface → annular pickup ring; the bottom is never solid plastic (mesh-tested), O-rings recorded as compliant mounts (compression bounded 5–30 %), excitation paths ordered hardware-first with intention/focus-only kept as source language.
- Reference discipline — 12 CC-SA assets registered with licenses and mesh-decode seed data; the conflicting M2 text and M2 mesh profiles stored separately, never reconciled; annular-ring craft locks stay out of cone sizing (source-scan test).
Frequency Key Studio (Sonic Lab)
New Session wizard — choose a type, carrier (100–1337 Hz family including the RGCS keys), and beat target (1.5–33 Hz, each with its status: standard band / source patent / source lead / candidate / project); add noise beds; preview the recipe JSON live with left/right frequencies (925 Hz + 7.83 Hz → 921.085 / 928.915 Hz).
v1.1 — live preview playback (QtMultimedia), STFT spectrogram preview (carrier lines visible over the noise floor), Timeline Editor (custom segment ramps replace the standard shape, invalid timelines refused with reasons), voice-cue / music-bed WAV import with start offsets, multi-carrier stacks, RMS loudness normalization (peak-capped at 0.95, recorded in the receipt), and batch render (UI + rgcs-sonic batch).
v1.2 — metadata-only web/video ingestion (frequency extraction with carrier/beat roles, claimed-use taxonomy with word-boundary matching), a URL-deduplicated corpus store with CSV export, duplicate clustering (frequency signature + title similarity), and deterministic recipe recommendations with stated reasons.
Engine — numpy + stdlib WAV only (no new dependencies): the 100/104 Hz patent-example binaural convention, click-free isochronic gating, seeded pink/brown/white/surf noise, constant-power pan, a mixer that never clips. 48 kHz float32 internally, 16-bit PCM export. Every render produces a schema-valid receipt (peak, RMS, input/output SHA-256), a session PDF sheet with receipt hash, deterministic recipe JSON, a YouTube metadata draft, and a checksummed bundle zip.
Seed recipes — 7 recipes ship (patent seed, Gateway-style Focus 10 seed, 6.3 Hz astral source lead, 7.83 Hz Schumann bridge, 925/1337 Hz frequency-key carriers, isochronic comparison) plus 11 beat targets.
Boundaries
Independent engine: no Hemi-Sync recordings, tape structures, narration, or branding are reproduced. No audio downloader exists — the corpus is titles/descriptions/URLs only, enforced by a test that greps the ingest module for network imports. Every export carries: Experimental audio recipe. Use comfortable volume. Results vary. Claimed uses are recorded from sources, not endorsed.
Verification
- Full suite: 8479 passed, 9 skipped, 1 deselected (D-V3-04), exit 0
- CI green on the 3-OS matrix (runs 31221779267 + 31224629851)
- UI smoke: 22 panels constructed OK
- ~145 new tests across both modules (sonic engine/timeline/recipes/exports/ingest/corpus/UI + phryll geometry/CAD/bundle/UI, including every pack skeleton assertion)
- Screenshot proofs and demo artifacts attached: 20 s Schumann-bridge WAV with session sheet + verified bundle, and the complete CRY-DEMO-120 phryll design bundle (Eye residual 0.0 mm)
- Final reports:
docs/reports/FREQUENCY_KEY_STUDIO_FINAL_REPORT.md·docs/reports/PHRYLL_GENERATOR_V2_FINAL_REPORT.md
Quickstart
python -m pip install -e ".[desktop]"
rgcs-workbench # Design Studio → Make binaural / frequency-key audio
rgcs-sonic render RGCS-SCH-0001 --duration 60 --out exports/Full docs: docs/user/FREQUENCY_KEY_STUDIO.md · report: docs/reports/FREQUENCY_KEY_STUDIO_FINAL_REPORT.md
Windows frozen build (attached)
rgcs-workbench-v8.5.0-windows-x64.zip — PyInstaller onedir build from commit d1d079c (v8.5.0 + the reference-review winding correction and coupling update); frozen --smoke-check passed (22 panels). SHA-256 attached alongside with the build manifest. Unzip and run rgcs-workbench gcs-workbench.exe. Unsigned, per project policy.
RGCS v8.4.0 — Design Studio v1
RGCS v8.4.0 — Design Studio v1
The desktop workbench (rgcs-workbench) becomes a coherent user-facing product: a guided, task-first Design Studio layered over the existing Advanced Scientific Workbench (every research panel preserved).
What's new
Design Studio home — task cards answering "what do you want to do?", open by default; Advanced Mode one click away.
Crystal Validator — measured dimensions + required uncertainty → schema/semantic validation, derived geometry (aspect ratios, volume estimate, density consistency, axial half-wave model estimate), labelled SVG diagram, and a certification-sheet PDF with receipt hash.
Phyrll Generator Designer — custom-fit holder geometry inherited from the selected validated specimen (cavity = specimen + 2×clearance; refused, not clamped, when parameters don't fit), deterministic SCAD export, STL through external OpenSCAD when installed, build-sheet PDF with print settings.
Coil / Pulse Designer — wire length and resistance estimates from declared formulas, six pulse modes with duty-cycle validation, and exact sideband arithmetic: 4096 ± 925 → 3171 / 5021 Hz · 4096 ± 1337 → 2759 / 5433 Hz · 4096 ± 963.026 → 3132.974 / 5059.026 Hz.
Annular Ring Designer — exact-closure cell geometry (default RGCS fixture: OD 288 / ID 188 mm, 37 cells, 33 active / 4 blanked), mask validation, probe plan, ring SVG/SCAD, phase-map and mask CSVs, engineering-sheet PDF.
Frequency Key Library — 17 keys as sourced / mathematical-relation / candidate records from a packaged data file; custom keys allowed only with an explicit custom status and warning.
Provenance everywhere — export receipts (input/output SHA-256, git commit, software version), bundle MANIFEST.json + CHECKSUMS.json with verification, and a dependency-free pure-Python PDF writer (text-extractable; no NaN can reach a sheet; unavailable estimates say "unavailable", never zero).
Docs & installers — INSTALL.md, task-first user docs under docs/user/, developer packaging guide, scripts/install_linux.sh (+ launcher + install receipt), tools/packaging/windows/build_windows.ps1 reusing the existing PyInstaller spec, release-manifest schema + generator.
This release also carries the physics-spine research pack (RC2 workbench lane) previously on the research branch.
Verification
- Full suite: 8362 passed, 9 skipped, 1 deselected (D-V3-04), exit 0
- UI smoke:
rgcs-workbench --smoke-check→ 20 panels constructed OK, background job succeeded - Linux install script executed end-to-end on a clean checkout (fresh venv, PySide6 6.11.1, smoke passed)
- Screenshot proofs of every workflow:
docs/assets/design-studio/screenshots/ - Final report:
docs/reports/DESIGN_STUDIO_FINAL_REPORT.md
Windows frozen build (attached)
rgcs-workbench-v8.4.0-windows-x64.zip — PyInstaller onedir build from tag commit 2c9834a via tools/packaging/windows/build_windows.ps1; frozen --smoke-check passed (20 panels + background job). Zip SHA-256:
f91efb90f442bc83f5e4ee5251cfed31534b7e813f37213e176f77a1901b5873 (also attached as .sha256, with the build's release_manifest.json). Unzip and run rgcs-workbench\rgcs-workbench.exe. Unsigned, per project policy.
Demo artifacts (attached)
Certification sheet, Phyrll SCAD + build sheet, coil/pulse build sheet, annular ring engineering sheet, and the checksummed export bundle — all generated by tools/design_studio_demo.py from the golden fixtures.
Claim boundary
Every exported sheet states it explicitly: designs and estimates are model outputs and reproducibility records. They do not by themselves validate an anomalous physical effect. Measurements decide.
Quickstart
python -m pip install -e ".[desktop]"
rgcs-workbenchRGCS Workbench Public RC2
RGCS Workbench Public RC2 Release Notes
Date: 2026-08-05
Tag: workbench-v1.0.0-rc2
Supersedes: workbench-v1.0.0-rc1
Verdict
RGCS_WORKBENCH_PUBLIC_RC2_READY
NO_PHYSICAL_CLAIM_ADVANCED
TERRA_RC4_PRESERVED
RELEASE_FILTER_CLEAN
ALL_TESTS_PASS
CI_GREEN_ON_MAIN
What changed since RC1
RC1 shipped with a locally green suite; the first public CI run then
failed on three platform and hygiene defects. RC2 is RC1 plus those
fixes, cut from a tree that is green locally AND on CI.
- A test comment and skip reason named the private repository
literally. The r10 privacy firewall flagged both lines on the
committed tree. The wording now says "the private archive" and the
firewall passes. The old literal remains only in git history,
where it is a declared, non-blocking residual by policy; history
is not rewritten. - The manifest tamper test mutated the checksum text with a
replacement that was a no-op on CI, because checkout line endings
change file hashes per platform. The tamper is now a deterministic
digit flip and the test detects it everywhere. - The diatomic Gamma-point check held an eigensolver zero to 1e-9,
tighter than BLAS noise across platforms. The eigensolver assert
now allows 1e-7, still eight decades below the optical branch;
the closed-form assert keeps 1e-12.
What this release is
RGCS Workbench is a public research workbench for reproducible
coordinate parsing, phase and resonance modeling, measurement
planning, and provenance tracking. Physical interpretations remain
hypothesis gated unless backed by bench receipts and independent
validation.
What this release is not
This release does not claim propulsion, lift, antigravity, gravity
control, source authentication, free energy, or validated craft
performance. The claim firewall scans every tracked markdown file
and the cage's own code and data as a standing release gate.
Terra status
RGCS Terra RC4 remains the frozen operational calibrated profile.
repo: andrew867/rgcs-terra
tag: v1.0.0-rc4
commit: 4fdee3e7fbdb416d8e4b32dcb422d0977e6f20af
verdict: GREEN_TERRA_ALIGNMENT_SOLVED_CALIBRATED_V1
B01A=CLOSED
B02A=CLOSED
B01B=VALIDATION_PENDING
B02B=PHYSICAL_VALIDATION_PENDING
B10=OPEN
Independent physical endpoint validation remains HOLDOUT_REQUIRED.
Workbench lanes
Unchanged from RC1: variable-length codec parse receipts and
correction ledger, crystal measurement objects with a bench-receipt
gate, Phyrll measurement lane with force fields refused, H-ME-SSP-001
protocol and derived arithmetic, craft-path hypothesis registry and
archive schema with public-safe seed records, claim firewall, and
the release manifest and checksums (release/workbench-rc2).
Receipts
Suite at the RC2 content commit: recorded in the seal commit
message, zero failures required. CI run on the same fixes: green
(run 31028439454 on commit 7040b96). A fresh clone runs the suite
without any private files present.
RGCS Workbench Public RC1
RGCS Workbench Public RC1 Release Notes
Date: 2026-08-05
Tag: workbench-v1.0.0-rc1
Verdict
RGCS_WORKBENCH_PUBLIC_RC1_READY
NO_PHYSICAL_CLAIM_ADVANCED
TERRA_RC4_PRESERVED
RELEASE_FILTER_CLEAN
ALL_TESTS_PASS
What this release is
RGCS Workbench is a public research workbench for reproducible
coordinate parsing, phase and resonance modeling, measurement
planning, and provenance tracking. Physical interpretations remain
hypothesis gated unless backed by bench receipts and independent
validation.
What this release is not
This release does not claim propulsion, lift, antigravity, gravity
control, source authentication, free energy, or validated craft
performance. The claim firewall runs as a standing release gate over
every tracked markdown file in the repository and over the cage's
own code and data.
Terra status
RGCS Terra RC4 remains the frozen operational calibrated profile.
repo: andrew867/rgcs-terra
tag: v1.0.0-rc4
commit: 4fdee3e7fbdb416d8e4b32dcb422d0977e6f20af
verdict: GREEN_TERRA_ALIGNMENT_SOLVED_CALIBRATED_V1
B01A=CLOSED
B02A=CLOSED
B01B=VALIDATION_PENDING
B02B=PHYSICAL_VALIDATION_PENDING
B10=OPEN
Independent physical endpoint validation remains HOLDOUT_REQUIRED.
New workbench lanes
variable-length coordinate codec with parse receipts and a
correction ledger (rgcs_coordinate + public_cage/codec_receipts)
crystal phase engine measurement objects with a bench-receipt gate
Phyrll generator measurement lane schema, force fields refused
H-ME-SSP-001 annular slow-wave measurement hypothesis, protocol and
derived arithmetic checked by test
craft-path hypothesis registry, append-only, public-safe seed
records imported
source and provenance archive schema, four-step community intake,
public-safe seed records imported
claim firewall and full-tree release scan
release manifest and checksums (release/workbench-rc1)
Status language
PUBLIC_WORKBENCH
OPERATIONAL_CALIBRATED_PROFILE
MEASUREMENT_HYPOTHESIS_NOT_VALIDATED
PHYSICAL_VALIDATION_PENDING
HOLDOUT_REQUIRED
NO_PHYSICAL_CLAIM_ADVANCED
Receipts
Test receipts for this release line: baseline 8022 passed before the
cage; 8124 passed after the cage; the record import and packaging
additions raise the count further and the final number is recorded
in the seal commit message. Zero failures at every step. A fresh
clone runs the suite without any private files present; the private
source corpus stays in the private repository.
The machine-readable module registry is
rgcs_workbench/public_cage/module_registry.json. The release
manifest and SHA256SUMS live in release/workbench-rc1/.
RGCS v8.0.0 — R15 Experimental Phase Infrastructure
RGCS v8.0.0 - R15 Experimental Phase Infrastructure
Release date: 2026-07-25
Predecessor: v7.0.0 (R13)
Final verdict: R15_GREEN_EXPERIMENTAL_INFRASTRUCTURE_READY_NO_PHYSICAL_CLAIMS_ADVANCED
R15 adds the r15/ package: 36 phases across 8 tranches (33 modules, 11 JSON
schemas, 36 terminal receipts) turning the R13 architecture into an
instrument-ready, calibration-bound, uncertainty-aware experimental platform.
Every hardware-facing lane ships REAL/REPLAY/SYNTHETIC/FAULT_INJECTION modes,
an error budget, protocols, tests, and docs; only physical acquisition is
blocked (no-purchase rule). The evidence ladder (E0-E7) caps any observation
missing a binding below a physical measurement; the strongest unreplicated
residual is UNEXPLAINED_INSTRUMENT_RESIDUAL, and there is no PHRYLL_DETECTED
state. See docs/v8/R15_FINDINGS.md and docs/v8/R15_NON_CLAIMS.md.
Additive: no prior work reset, no public history rewritten.
PHYSICAL_VALIDATION_NOT_CLAIMED.
expect: 6533 passed (1 archived-environment byte test deselected by policy D-V3-04)
RGCS v7.0.0 — R13 Discovery & Experiment Architecture
RGCS v7.0.0 — R13 Complete Discovery and Experiment Architecture
Release date: 2026-07-24
Predecessor: v6.3.0 (R12)
Final verdict:
R13_GREEN_COMPLETE_SOFTWARE_SIMULATION_AND_EXPERIMENT_ARCHITECTURE_NO_BENCH_CLAIM
What R13 adds
R13 executes the full 48-phase discovery-and-experiment pack as the new
r13/ package — 32 modules, 32 test files, and 48 phase receipts
(docs/v7/receipts/). Nothing here is a measurement; the strongest claim
class any module reaches from software alone is
REPOSITORY_COMPUTATIONAL_RESULT.
The architecture
- Governance core (
claimtypes): the claim ladder and the seven
forbidden promotions, enforced as typed refusals. - Common linear-response core (
response,bridgegraph,srcregistry):
Green functions, Kramers–Kronig consistency, S-matrix unitarity, and a
certificate-gated coupling-graph search that extends R12's bridge rule. - Quartz chain (
crystalframe,atomistic,homogenize,
piezobridge): direct/reciprocal frame, an analytic phonon model with an
enforced acoustic sum rule, homogenization to continuum elasticity, and a
piezo→BVD bridge emitted as a certificateAWAITING_FALSIFICATION. - Transform layer (
symplectic,quadfield,qpm,floquet,
avoided,chiral,boundaryenergy,heterodyne): the rotation-vs-squeeze
discrimination, quasi-phase-matching, Floquet tongues, avoided crossings,
chiral phonons, and a boundary energy ledger that closes with no new energy. - Apparatus & detector stack (
apparatus,qcmstack,sixangle,
imaging,daq,diskdrive): all preregistered designs — nothing is
built — with the six-angle ring's planar-uniformity-is-not-isotropy rule. - External validation (
euphonic,scattering): a force-constant
interface (real DFTBLOCKED_MISSING_INPUT) and synthetic INS/IXS
predictions. - Coordinate codec (
coordfinal,holdout,magroot,shellmap,
epochsolve,serialize): the finalized icosahedral codec that decodes
only to a 32-member alias set, with a blinded decoder holdout. - Prospective experiments & controls (
experiments,preregister): six
preregistered, not-run bench experiments and the sealing/blinding protocol.
The red team
tests/v6/test_r13_redteam.py attacks all seven promotions plus the bridge,
alias, energy, and isotropy governance — 16 attacks, all refused.
Non-claims (unchanged discipline)
No cross-domain transfer is measured; no new energy; no isotropic emission;
no decoded destination; no authenticated source; no carrier. Bench, DFT
force constants, neutron facility, and beam time are all
BLOCKED_MISSING_INPUT, stated explicitly in phases 25–30, 33, 34, 45.
Provenance
Additive: no prior work reset, no public history rewritten.
PHYSICAL_VALIDATION_NOT_CLAIMED.
expect: 5638 passed (1 archived-environment byte test deselected by
policy D-V3-04)
RGCS v6.3.0 — R12
RGCS v6.3.0 — Release Notes
Status: SOFTWARE_VERIFIED_PHYSICAL_UNTESTED
Baseline: v6.2.0
Licence: MIT, unchanged. No relicensing.
R12: a new r12 package — the icosahedral packet grammar, quaternary
refinement, eight shells, a South-referenced body frame, an IGRF-14
magnetic-root certificate, isotopic/astronomical epoch candidates, a
reciprocal-space crystal model, typed cross-domain coupling certificates,
and home-lab experiment plans. Nine modules, additive.
Final verdict:
R12_GREEN_ICOSAHEDRAL_AND_BRIDGE_ARCHITECTURE_IMPLEMENTED_NO_UNIQUE_DECODE
The grammar — valid, and not a decoder
A 30-bit word parses as F5 | Q22 | S3: 5 face bits, 22
quaternary-refinement bits (11 levels × 2), 3 shell bits. All five
registered vectors decode exactly as the pack specifies (face 4, shared
3301 path prefix).
Headline null (R10.6 lesson, six tranches on): the shared 13-bit binary
prefix is a fact about numeric range, not content — 14 bits are
expected from the values' span, band-clustering p = 1.0 (random
same-band draws share a longer prefix on average) → EXPLAINED_BY_RANGE,
with a power control that catches a planted 24-bit prefix. Geographic
decoding is refused until face numbering, body orientation, magnetic root,
handedness and shell projection are all frozen — and even then the result
is GRAMMAR_ONLY_NO_GEOGRAPHIC_CLAIM.
The cross-domain rule changed
R11.1 refused every transfer between physical domains. R12 supersedes
that (does not delete it):
NO_AUTOMATIC_EQUIVALENCE
TRANSFER_ALLOWED_WITH_EXPLICIT_COUPLING_CERTIFICATE
A transfer may now be licensed — by a certificate declaring all nine of:
domains; state variables and units; coupling operator; overlap; detuning
and damping; phase matching and symmetry; energy accounting; uncertainty
and null model; and a falsifying measurement. A complete certificate is
only ENGINEERING_CANDIDATE until measured, and nothing here can be
measured, so every certificate is AWAITING_FALSIFICATION. Certificates
are directional and do not compose. The default is still refusal.
The rest of the mission
bodylock— South-referenced frame without a forced hemisphere
inversion; the double-inversion audit fires only when an improper op
and a latitude flip cancel to net +1 while claiming to invert.shells8— 8 shells; four spacing laws give different radii for the
same index, so the radial projection isUNRESOLVED; atomic-shell
physics refused.igrf14root— IGRF-14 root, epoch mandatory, coefficient grid
BLOCKED_MISSING_DATA, identification always refused.epochcand— Ba-130 / Cs-137 / Cs-133 / astronomical clocks span
38.5 orders of magnitude; ~8.7×10¹⁸ aliases; no unique epoch.reciprocal— standard crystallography;aᵢ·bⱼ = 2π δᵢⱼexact;
reciprocal/Q-space is not a place.homelab— mode/phase/ringdown/isotropy experiments preregistered,
not run; anisotropy in anisotropic quartz is the null, not an anomaly.
Red team
Twelve attack families, each failing with a typed reason. The privacy guard
now recognises this pack's private_do_not_commit/.
Verification
git clone https://github.com/andrew867/rgcs && cd rgcs
pip install -e ".[dev]"
pytest -q --deselect tests/regression/test_generator_determinism.py::test_generator_deterministic
# expect: 5175 passedCI note. Free-tier GitHub Actions minutes are exhausted; the
verification of record is the full local suite on the exact commit.
Scope
The pack ships 38 agent prompts. This release implements the master
orchestrator's 10-item mission; the neutron/phonon theory, chiral-phonon
INS, beamtime, Floquet-QPM converter, response-function and related
sub-prompts are a recorded R13 seed backlog, deferred rather than
dropped.
What this release does not claim
No externally transmitted address, no identified craft, no creation epoch,
no artificial geomagnetic structure, no novel particle, no anomalous energy
transfer. Mathematics, models and an experiment plan — nothing measured.
See R12_FINDINGS.md for the full analysis.