The Seven-Layer Blueprint.
One stack, seven layers, six modules. This page gives you the layer definitions, the module-to-layer map, and the reasoning behind the one intermediate representation that spans all seven — so you can see exactly which part of EigenOS sits above your vendor stack and which part is still on the roadmap.
L1 Hardware → L2 Execution → L3 Orchestration → L4 Programming → L5 SDK / API → L6 Security → L7 Discovery. EigenOS operates at the boundaries between them; the boundaries are the product.
Seven layers, one responsibility each.
Each layer has a distinct job and a distinct set of product modules resting on it. Read a row top to bottom and you get the same information a system architect needs: what the layer owns, and which module touches it.
L1 HardwareQubits, cryogenics, photonics, and the physical control plane that sits under every execution.
BSaaS (indirect) · T5.4 (audit) · QORCH (abstract) · QUOTA (budget) · Discovery (explore) · EQL (describe)
L2 ExecutionCircuit compilation and runtime — where a program becomes shots, cost, and error.
BSaaS (run) · T5.4 (validate) · QORCH (orchestrate) · QUOTA (per-shot ledger) · Discovery (mine) · EQL (multi-vendor abstraction)
L3 OrchestrationMulti-vendor control: dispatch, supervision, and the error budget that binds vendors into one system.
BSaaS (dispatch) · T5.4 (track) · QORCH (supervise) · QUOTA (error budget) · Discovery (correlate) · EQL (system spec)
L4 ProgrammingThe EQL compiler and authoring surface — a program written once, at source.
BSaaS (test) · T5.4 (audit reference) · QORCH (job spec) · QUOTA (constraints) · Discovery (explore) · EQL (core language)
L5 SDK / APIThe BSaaS dashboard and public API — the product surface where every module meets the user.
BSaaS (primary) · T5.4 (audit API) · QORCH (job API) · QUOTA (ledger API) · Discovery (query API) · EQL (compiler API)
L6 SecurityPQC boot chain and authenticated command path — the trust boundary under every call.
All six modules (auth · signed commands · PQC-authenticated references · provenance)
L7 DiscoveryMining, comparison, and scoped claim audit — where results become evidence.
BSaaS (mine) · T5.4 (primary) · QORCH (correlate) · QUOTA (meta-analyse) · Discovery (primary) · EQL (cite)
Three tiers, six modules, seven layers.
The tiers are a maturity statement, not a marketing one. Tier 1 is shipped and priced. Tier 2 is in validation. Tier 3 is designed and documented but not live — nothing in the roadmap tier runs today. Within each module, the cell shows what that module does at each layer.
Tier 1 — Live
Shipped and running today. Public API, public docs, public pricing.
BSaaSQuantum Backend-as-a-ServiceOne REST endpoint that runs the same circuit on 8 vendor backends and normalises the measurements into a single schema.
Tier 2 — Beta
In validation. End-to-end behaviour is public; the interfaces are still moving.
T5.4Claims AuditIndependent validation of a scoped quantum-performance claim: 9 truth rules, N0–N4 evidence tiers, citable verdicts.
QORCHPhoton TransferInter-vendor state-transfer modelling: fidelity, latency, entanglement rate, link-loss budget, and cost per logical operation.
Tier 3 — Roadmap
Designed, not shipped. Architecture is fixed and documented; nothing here is live.
QUOTAQuota & LedgerPer-shot accounting, an error-budget compiler, and a per-shot ledger — track every shot, enforce every budget, on any vendor.
DiscoveryDiscoveryExplores the trade-off space between configuration, timing, and measured outcomes. Requires real experimental data, not simulators.
EQLEigenOS Quantum LanguageThe cross-layer IR. Vendor-portable circuit plus unit definitions — the same source spans L1 description through L7 citation.
EQL is the IR that crosses all seven layers.
A compiler IR normally lives at one boundary. EQL does not. The same typed source describes hardware, carries QEC policy, drives classical control, expresses application intent, prices the job, and proves where the result came from — which means one artefact can be handed from L1 to L7 without being re-authored at each hand-off. Whoever owns the IR owns the orchestration layer above it. That is the strategic claim, and it is the claim P11 MMQCR is filed to defend.
The provenance half matters as much as the portability half. A canonical hash plus an Ed25519 envelope turns a circuit into something citable: the same source compiled in two separate sessions produces the same hash, so a T5.4 verdict, a BSaaS measurement, and a paper figure can all cite one identifier that a third party can verify.
COMPILEREQL v0.1, written in TypeScript. Zero runtime dependencies in the core library; the implementation is Apache 2.0.
PARSERLexer plus recursive-descent parser. 20 AST node types, 8 precedence levels, line and column tracked on every token.
UNITS60+ unit types. Every numeric literal carries a unit symbol, so rx(angle=2.0 Hz) is a compile-time error rather than a hardware fault. Dimensional type checking is the W3-4 deliverable.
GATES25-gate catalogue: Clifford, non-Clifford, parametric 1Q and 2Q, and higher-order (ccx, cswap). Vendor-native extensions hang off the same registry.
PLATFORMS10 hardware platforms registered — the 8 BSaaS vendor backends (Origin Quantum wuKong-180, IBM Quantum, IonQ, Google Quantum AI, Rigetti, Quantinuum, AWS Braket, Xanadu) plus QuEra and Pasqal. All 8 BSaaS vendor backends are simulated.
PROVENANCECanonical hash plus Ed25519 envelope. Source serialises deterministically and hashes with SHA-256, so the same circuit in two separate sessions yields the same citable audit ID (eig-t54-…). Anyone holding the source can verify the result downstream.
BUILD STATEW1-2 shipped: 0 type errors, clean build, 73/73 tests passing. W3-4 next: dimensional type checking. The public EQL product surface is still ROADMAP.
Why orchestration, not competition.
EigenOS is a complementary integration system. It sits above the compilers, the QEC libraries, and the vendor SDKs, and makes them interoperable, observable, and auditable. It owns no qubits, it does not fork any vendor toolchain, and it does not need to win for the field to lose — the field losing is a reason for it to exist.
NO HARDWAREEigenOS does not own qubits, cryogenics, or control electronics. L1 belongs to the vendors. We integrate against their APIs; we do not compete for the silicon.
NO FORKWe do not replace Qiskit, Cirq, Stim, Q#, or the vendor SDKs. EQL compiles to OpenQASM 3.0 and to native vendor targets, so an existing stack keeps working and gains one integration layer.
ONE CONTRACTOne API, one measurement schema, one canonical hash across every backend. Switching vendors is a parameter change, not a rewrite of the workload.
AUDIT THE FIELDT5.4 audits vendor claims the same way it audits ours. Independent verification is the product, and being the one who publishes the verdict is not a competitive advantage over anyone else.
HONEST LIMITSAll 8 BSaaS vendor backends are simulated. We do not claim real-hardware QPU access, we do not claim fault-tolerant quantum computing, and we do not claim a partnership that does not exist.
The honest version of the position: EigenOS is a vendor-neutral orchestration layer for distributed quantum computation. What it delivers is the L3 control plane and the L7 evidence plane — not fault tolerance itself, and not a claim on any vendor's hardware roadmap.
TIGHT v3 — the parent claims the stack.
The portfolio is deliberately small. P11 MMQCR is the parent, and almost everything else in the blueprint is either a continuation of it or a filing that had to stay separate because it sits in a genuinely different layer. Every claim in the portfolio maps to a row in the seven-layer stack above.
PORTFOLIOTIGHT v3. 3 families · 13 claims · AUD $300 total IP spend in Q4 2026. Compressed from the v2.3 plan of 6 families and 14 claims.
PARENTP11 MMQCR — Multi-Modal Quantum Compilation Runtime — is the parent family, covering L2 and L3. Filed at IP Australia on 13 Oct 2026 at the AUD $100 online minimum.
P11 ADDENDAP11-QEC addendum, 2 Nov 2026 — syndrome routing, declarative correction, PQC-authenticated commands. P11 L2 CIP, 13 Nov 2026 — per-shot accounting, error budget compiler, multi-vendor L2 abstraction. Both are continuations of P11, not new patents.
FOLDEDP17 photonic loss and erasure, P18 decoder tail latency, and P19 modular interconnect fold into P11 as a Q1 2027 CIP. Dependent claims are narrower, so they are harder to invalidate independently.
SEPARATEP16 Cryo-CMOS (L1, 16 Nov 2026) and T5.4-QEC-Normalize (L7, deferred to Q1 2027) stay separate filings. Different layers mean different statutory classes, different prior art, and different examiner pools.
LAYER MAPL1 P16 Cryo-CMOS · L2 P11 L2 CIP · L3 P11 base claims, the QEC addendum, and the Q1 2027 CIP · L6 PQC-authenticated command path · L7 T5.4-QEC-Normalize. Deferred filings are Patent Pending.
Run the stack, or talk to the author.
BSaaS is live and takes a circuit today. If you would rather talk through layer boundaries, an EQL manifest schema, or a co-development proposal, that is a founder conversation and it starts the same way.