AstroDynamiXOffline Scientific & Telemetry Software
Searchable Compression & Edge Core · Division 02 · Pure-Rust + 5-Target C-ABI

Quantoji Searchable Compression & AstroDyX Edge Runtime

A Pure-Rust solver core with C/C++, Rust, Go, and Python host bindings and panic-guarded 5-target C-ABI libraries (natively runtime-tested on macOS arm64/x86_64 & Linux aarch64; Linux x86_64 & Windows x86_64 cross-compiled and symbol-audited [20/20 exports]; zero shared source code with Division 01) for lossless numerical/sensor packing, compressed-header range queries, erasure recovery, and integer edge sequence scaling.

Data-Source Key:Observational Individual public Silesia & PCM filesReference Model Local 4-node hardware cluster testSynthetic Deterministic numerical, TPC-H & erasure vectors
Silesia mr FileObservational
4.00× Single File
Single 9.97 MB public MRI file vs. 3.79× XZ-6 (+5.6%; trails XZ-6 on English prose: 2.77× vs. 3.60×; specialist DICOM codecs untested)
Selective Range ExtractSynthetic
12.3× Extract
6.76 ms selective extract (27.79 µs header filter = 2,985×) vs. 82.95 ms unindexed full decode across 1.00 MiB archives
Systematic Parity HealSynthetic
100% Bit-Exact Heal
36.11 ms SHA-256 repair in studio fault-injection tests; systematic GF(2⁸) parity adds +m/k storage over k data blocks to heal up to m lost blocks of k + m total (default k, m withheld)
AstroDyX SequenceSynthetic
1,048× Less RAM
257.81× latency speedup (100.46 µs vs. 25.90 ms) vs. unquantized dense f32 quadratic attention on arm64 (synthetic runtime/RAM scaling benchmark only; no downstream model task-accuracy parity claimed; not vs. FlashAttention/SSMs)
01 // Primary Division 02 Flagships

Quantoji Data SDK & AstroDyX Edge Intelligence Core

Maintained in a completely segregated codebase and U.S. provisional-patent-pending lineage from Division 01, Division 02 ships CLI binaries and 20-function C-ABI shared libraries (thin unsafe extern "C" FFI boundary wrapped in std::panic::catch_unwind) across 5 targets (natively runtime-tested across 3 physical Unix hosts [macOS arm64, macOS x86_64, Linux aarch64], with Linux x86_64 .so and Windows x86_64 PE32+ .dll cross-compiled and symbol-audited [20/20 exports; ff0c6108...]) plus C/C++, Rust, Go, and Python host bindings.

Synthetic Smooth 3D IEEE 754 f32 Numerical Grid100% Bit-Exact Lossless Mode (Pre-Parity Base Container)

Lossless 3D Float Field Compression Ratio vs. Tested General & Float-Specialist Codecs

All reported comparators run in bit-exact lossless mode on identical smooth synthetic 3D f32 physical grids: XZ -6 and Zstd -19 ran as standard standalone CLI codecs without external byte-shuffle or delta prefilters, while fpzip lossless serves as the float-specialist comparator (error-bounded lossy frameworks such as SZ3 are excluded; extreme slow presets xz -9e / zstd --ultra -22 were outside the standard CLI profile; the Oct 18 3D f32 dossier [98f91024...] recorded lossless ratios rather than wall-clock MB/s, while the separate 4.00 MiB SDK verification harness [ff0c6108...] measured 121.05 ms 2-thread encode and 31.40 ms decode on arm64). Quantoji exceeds standalone XZ -6 by 5.14× on smooth 3D floats while trailing big-dictionary codecs on flat English prose: Bar widths round nearest-integer against 57.792× max.

Quantoji (Lossless Container)
57.792×
XZ -6 (Standalone CLI; No Shuffle)
11.240×
Zstd -19 (Standalone CLI; No Shuffle)
8.114×
fpzip (Float-Specialist Lossless)
6.463×
Table 1 — Division 02 Lossless Compression, Domain Boundaries, Indexed Query & Edge Sequence Benchmarks
System & WorkloadData Source ClassMeasured ResultTested Baselines, Domain Boundaries & Parity Scope
Quantoji · Silesia Public mr & x-ray FilesObservational Files4.004× mr · 2.259× x-raySingle public files (9.97 MB MRI, +5.6% vs. XZ-6 3.791× & Zstd-19 3.298×; X-Ray vs. XZ-6 2.094×); medical specialists (JPEG-LS / HTJ2K) not yet benchmarked
Quantoji · Honest Prose Boundary (dickens)Observational Prose2.774× losslessDomain Boundary: Beats Gzip-6 (2.632×) & matches Zstd-3 (2.773×), but trails big-dictionary XZ-6 (3.599×) & Zstd-19 (3.576×) on flat English prose
Quantoji · 16-Bit PCM Mono Audio FileObservational Audio1.375× losslessSingle 227 KB mono WAV vs. FLAC -8 audio specialist (1.341×), XZ-6 (1.248×), and Zstd-19 (1.192×); base lossless container
Quantoji · Smooth 3D IEEE 754 f32 GridSynthetic 3D Grid57.792× lossless5.14× tighter than standalone XZ-6 (11.240×, no external byte-shuffle or delta prefilter), 7.12× vs. standalone Zstd-19 (8.114×), and 8.94× vs. float-specialist fpzip lossless (6.463×); Oct 18 dossier recorded ratios rather than wall-clock MB/s (separate 4.00 MiB SDK harness [ff0c6108...] measured 121.05 ms 2-thread encode / 31.40 ms decode on arm64); base container prior to parity
Quantoji · Boundary Header Range QuerySynthetic 1.00 MiB6.76 ms extract · 27.79 µs scanvs. 82.95 ms unindexed full-stream decode (12.3× faster selective extract; 2,984.6× faster metadata filter; <0.2% index overhead; not vs. seekable chunk indexes)
Quantoji · TPC-H Tabular Moment RollupSynthetic TPC-H69 ns (0.069 µs) · 0.47 KiB32,768-row 16-column LINEITEM (4.22 MiB): Columnar Dict + Zstd-3 achieves tighter 8.60× bulk compression vs. Quantoji's 5.22× lossless ratio, while Quantoji evaluates compressed-header statistical moments in 69 ns (0.069 µs, 0.47 KiB RAM) vs. 949.54 µs (1,846.29 KiB RAM) on Apple Silicon arm64
Quantoji · Integrated Systematic Erasure HealSynthetic Erasure100% SHA-256 heal (36.11 ms)Configurable systematic GF(2⁸) MDS parity: adding m parity blocks to k data blocks (+m/k overhead over data) heals up to m lost blocks out of k + m total encoded blocks—matching standalone Reed–Solomon / PAR2 efficiency with default k, m container geometry withheld (base ratios exclude optional parity)
AstroDyX · Integer Edge Sequence CoreSynthetic Scaling257.81× faster · 1,048× less RAM100.46 µs vs. 25.90 ms latency & 1,048× state-RAM reduction vs. unquantized f32 dense quadratic attention at matched sequence length on arm64 (synthetic runtime/RAM scaling benchmark only; no downstream model task-accuracy parity claimed; not vs. FlashAttention or quantized SSMs)
Step 01 · Mutual NDA

Sample Telemetry Benchmark

Provide a non-sensitive sample float tensor or sensor log under NDA; receive lossless ratio, query latency, and Reed–Solomon-rate parity comparisons.

U.S. Provisional Patents Pending
Step 02 · Architecture Review

5-Target C-ABI & CLI Fit

Review the 20-function panic-guarded C-ABI header and CLI flags against your macOS, Linux, or Windows (x86_64) storage engine or flight processor.

Bindings: C/C++, Rust, Go, and Python
Step 03 · NDA / OTA Pilot

Air-Gapped Binary Evaluation

Built with zero HTTP/TLS client crates and designed for air-gapped execution where enclave firewall/OS policy enforces network isolation (linux-aarch64 .so dynamic-library footprint is limited to libc, libdl, and libpthread [ff0c6108...]).

Designed for Air-Gapped Enclaves
Step 04 · Production License

Cluster Seat or OEM Royalty

License annual per-cluster server subscriptions, embedded OEM distribution rights, or 10 U.S.C. § 4022 federal Prototype OTA deployments.

100% Founder-Originated IP
02 // Modular Pure-Rust Infrastructure Crates (Self-Reported Internal Studio Benchmarks)

Deterministic Geometry, GPS-Free Sync & Mesh Supervision

Standalone Division 02 add-on libraries for autonomous platforms and local multi-node orchestration. All three cards below report self-run internal studio benchmarks (not part of the 6-digest flagship ledger below; full benchmark harnesses and logs are provided in our mutual-NDA evaluation pack).

Synthetic · NDA Eval Pack

Deterministic Avionics Geometry Core

Self-reported internal studio benchmark: evaluates 54 µs symbolic derivatives and 32 µs 3×3 Jacobians on arm64 with static-memory discipline—preventing coordinate precision drift observed in naive 64-bit float baselines at 10,000 km orbital range (internal suite only; no external FAA/DER certification audit claimed).

Synthetic · NDA Eval Pack

SENTRY GPS-Free Clock Alignment

Self-reported internal studio benchmark: resolves sub-sample clock skew (correlation R > 0.95 across +7/−5 sample offsets in multi-channel synthetic drift vectors on arm64) for GPS-denied unmanned and subsea arrays using HMAC-SHA256-authenticated telemetry envelopes.

Local 4-Node · NDA Eval Pack

Runtime Isolator & Fleet Control

Self-reported internal studio benchmark: combines a 29 µs process-boundary isolator with fleetctl local mesh orchestration—measured dispatching and reconciling 12 concurrent tasks across a 4-node local hardware cluster in 854 ms with zero orphan processes.

Division 02 SHA-256 Raw Input & Benchmark Ledger (6 Digests)Public Silesia Input Independently Verifiable · Studio Artifacts Under NDA

Public Silesia Input Anchor & Aliased Studio Benchmark Digests

Redaction & Alias Disclosure: Public Division 02 documentation withholds proprietary waveform transforms, internal container layouts, and default k, m block/shard geometry. Item 1 displays its unmodified upstream filename (silesia/mr, 9,970,564 B, downloadable from the public Silesia compression corpus for independent third-party shasum -a 256 verification), whereas internal studio reports (Items 2–6) use uniform Division02_* public display aliases for private studio workspace paths (in particular, Division02_Silesia_Float_Dossier.md [98f91024...] aliases the Oct 18 internal container preprint covering both the Silesia/f32/WAV lossless ratio table and boundary/erasure architecture). SHA-256 digests, exact byte sizes, and local UTC mtimes are recorded below so public Silesia input can be independently verified via shasum -a 256 while internal studio reports are supplied under mutual NDA; local filesystem mtimes are mutable OS metadata corroborating our self-reported studio workflow rather than external RFC 3161 cryptographic timestamps:

1. Public Silesia mr Raw Input File (Third-Party)
silesia/mr · 9,970,564 B · 2003-03-20 15:12:26 UTC68637ed52e3e4860174ed2dc0840ac77d5f1a60abbcb13770d5754e3774d53e6
2. Quantoji Silesia & 3D Float Dossier (Alias)
Division02_Silesia_Float_Dossier.md · 8,079 B · 2026-10-18 21:17:30 UTC98f9102463abf3103070542afdac158db711556fa78e5899c9bdcc1d45784cb3
3. Quantoji 5-Target C-ABI & Parity Report (Alias)
Division02_SDK_Parity_Report.md · 14,607 B · 2026-11-12 00:00:18 UTCff0c6108550fcfda4bca0a72233657dbd4a663a6988f8ead45b393f62607d6ff
4. Selective Extract & Erasure Heal Report (Alias)
Division02_Boundary_Erasure_Report.md · 16,902 B · 2026-10-16 15:13:51 UTC0dd9ec5f5ffb38cc2d974b0d735f7005f5d88b9199f0cc2ca0b4baf1772e01d6
5. Quantoji TPC-H 69 ns Rollup JSON (Alias)
Division02_TPCH_Rollup_Benchmark.json · 2,672 B · 2026-11-10 23:43:27 UTC67e5aa325f55206b43b905e0f54156b98180f1fd1fc58d5a2972f4f8e2eac2fa
6. AstroDyX 257.81× / 1,048× Scaling Dossier (Alias)
Division02_AstroDyX_Scaling_Dossier.html · 43,354 B · 2026-10-21 02:43:51 UTC710ad5cf5fa8235a27d6745532e24b1f5e64de60733ab5f268cf94e794ec1e23