SPECTR
Structural Primitive Extraction & Compositional Topology for Replay
Deterministic Structural Replay
Ingest any binary stream. Record its structure. Replay it exactly, while exposing the structural signal that entropy and compression algorithms cannot see.
↓ See the evidence
The Core Invariant
The Replay Guarantee
Every claim SPECTR makes derives from one foundational property. It is not approximate. It is not probabilistic. It holds for every input, every time.
Same input, same configuration: identical output every time. No stochastic elements, no approximation. The result is the same on any machine, any run.
The original stream is fully reconstructible from the ledger alone. Nothing is discarded, approximated, or summarised. Every bit that went in comes back out.
The ledger records more than the stream. It records the arrangement: which primitives appeared, in what order, and how they repeated. That structural record is what the replay invariant preserves.
How It Works
Three operations. One invariant.
Ingest
Any binary stream (files, network packets, genomic reads, sensor data) is decomposed into a primitive vocabulary. Every selection is recorded in an append-only ledger entry. The engine operates at bit resolution with no byte-alignment requirement.
# → (seed, status)
Discover
The discovery curve reveals structural signal invisible to entropy metrics: non-aligned boundaries, arrangement, and repetition patterns. Where Shannon entropy returns a single scalar marginalised over order, SPECTR produces a temporal structural record.
# → discovery_rate, ...
Replay
replay(ingest(X)) == X: the original stream is reconstructed exactly from the ledger alone. No separate raw byte storage. No approximation. The same structure that enables discovery also enables lossless reconstruction.
# → exact original bytes
For the Skeptics
What SPECTR Is Not
Every person who encounters SPECTR tries to fit it into a familiar category. They are always wrong. Here is why.
The ledger grows with novel data. SPECTR makes no claim about size reduction. Deduplication of recurring primitives is a structural observation, not an optimisation goal.
The original stream is fully reconstructible from the ledger. Nothing is one-way. The seed is a structural identifier, not a content fingerprint.
Purely deterministic. No training data, no weights, no approximation, no gradient descent. Given the same input and configuration, SPECTR returns identical output on every machine.
SPECTR makes no confidentiality guarantees. The ledger encodes structure, not secrets. It is a substrate layer; what you build on top handles access control.
What it is: A deterministic, append-only substrate that decomposes any binary stream into a primitive vocabulary, records the arrangement as a replayable timeline, and reconstructs the original stream exactly. replay(ingest(X)) == X, always.
Verified Results
Structural signal invisible to entropy.
Visible to SPECTR.
A hidden pattern that gzip is blind to
SPECTR AUC: 0.9847. Baseline: 0.5000.
Same letters, different order — and standard tools can't tell the difference
SPECTR AUC: 0.9645. Baseline: ~0.5000.
The standard measurement says they're identical. They're not.
SPECTR AUC: 0.9861. Baseline: 0.5000.
All results are from deterministic test runs. Every claim maps to a reproducible test configuration. Retracted claims are documented with root causes.
Built for Research
Every result is reproducible.
Every claim maps to a passing test.
SPECTR is built for researchers who need deterministic, auditable results. The engine is Rust. The bindings are Python. The claims are test-gated.
REST API
All engine capabilities available over HTTP. Ingest, replay, structural profiles, batch analysis, and semantic normalisation through one authenticated endpoint.
curl -X POST /v1/engine \
-H "X-Api-Key: $KEY" \
-d '{"data": "<base64>", "mode": "auto"}'Python Wheel
The Rust engine ships as a native Python wheel. Run experiments locally without any network round-trips. Full API parity with the hosted service.
import ufm
engine = ufm.InvariantIdentityEngine(
bit_depth=21, storage_path="ledger.bin"
)
seed, status = engine.process(data)
original = [bytes(seq) for seq in engine.replay(seed)]Open Claim Registry
Every quantitative result on this site maps to a claim ID (C-DISCOV-XXX, C-CORE-XXX) in a public test registry. The tests are deterministic, seeded, and re-runnable. Retracted claims are documented with root causes, not removed. Honest failures are part of the record.