Publication boundary: Implemented and regression-tested software behavior using SYNTHETIC_TEST_DATA. No biology, reactor, analytical method, yield, economics, or safety result is reported.
What synthetic data is allowed to prove
A synthetic fixture is useful when the question is whether software accepts a schema, preserves units, seals provenance, detects impossible values, replays deterministically, or fails safely.
Eden runs synthetic observations through the same qualification and model-consumer plumbing used by a future instrument record. That exercises interfaces and conservation logic without changing the source’s evidence class.
- Ingestion and schema validation work end to end
- Unit and conservation checks run against expected boundaries
- Provenance hashes, replay, rollback, and failure states remain traceable
- Real model functions can be exercised in SHADOW mode
- Known missing consumers are recorded as DOWNSTREAM_MODEL_NOT_IMPLEMENTED rather than fabricated
The line synthetic data cannot cross
Synthetic observations may create qualified shadow events and may move a sensitivity score, but measured_eden_parameters remains empty. A shadow score change cannot reopen a campaign.
Pipeline test doubles require the exact non-production gate EDEN_ALLOW_PIPELINE_TEST_DOUBLES=1. Production refuses them, and any gated test-double record remains tagged PIPELINE_TEST_DOUBLE_NOT_REAL_EDEN.
The next observation that would change the claim
Only qualified real-instrument observations from an authorized physical run can approach MEASURED_EDEN. They must pass the same calibration, contamination, closure, provenance, and promotion gates exercised by the synthetic path.
Until that happens, the defensible public claim is narrow: Eden has tested evidence plumbing and a fail-closed promotion boundary. It has not produced physical process evidence.
Artifact receipt
The supporting repository is protected. Eden publishes the reviewed artifact identifiers, dates, source locators, and SHA-256 receipts without presenting the underlying planning files as public measurements.
Open the machine-readable artifact and checksum manifest
Technical review date: .
Public evidence contract
Artifact-backed publication
An artifact link establishes traceability, not a broader production, safety, environmental, economic, or deployment claim. The claim state and exclusions below define the public boundary.
- Artifact type
- Software release
- Artifact status
- Published
- Claim state
- Synthetic test data / not a physical result
- Technical review date
What changed
The experiment pipeline now qualifies, seals, replays, recalculates, and shadow-rescores synthetic records while hard-refusing their promotion into measured Eden evidence.
What the artifact narrowly supports
Synthetic data validates software plumbing, conservation checks, provenance, replay, model wiring, and fail-closed behavior.
Eden interpretation
Synthetic evidence can reduce software uncertainty while remaining categorically separate from physical process evidence.
Next evidence gate
Land qualified observations from an authorized real-instrument run and pass QC, contamination, balance-closure, provenance, and promotion gates.
Excluded inferences
- Pipeline success does not validate the biology, reactor, or analytical method.
- A model-output change under synthetic inputs does not validate model accuracy, yield, or economics.
- No safety, food-suitability, or physical Eden result is claimed.
Unknowns and risks
- Real observation quality and repeatability remain unknown.
- Model accuracy against physical data and process performance remain unknown.
- Named downstream model gaps remain unresolved.
Evidence and references
- Synthetic promotion-boundary artifact and checksum manifest (Reviewed Eden artifact receipt)
- Current Eden evidence program (Public evidence taxonomy)