QUESTION
Is the case study ready for public inspection?
The release gate asks whether a reader could inspect a sanitized, versioned artifact and understand exactly what a local fixture check supports.
Case study 001 · hold
Evidence exists.
Packaging is not ready.
The route-aware synthetic fixture remains HOLD, with PUBLIC—DEMO as a possible future tier—not a benchmark result. Its compact project repository remains withheld until provenance, baselines, representative traces, and clean reproduction instructions are complete.
Release decision
HOLD
Route-aware synthetic fixture case study.
No downloadable fixture, model result, benchmark rate, or deployment claim is published here. The possible future tier is a synthetic mechanism and scorer-regression DEMO only.
target: public demoQuestion → evidence → decision
INPUTS / SOURCES
Release-gate review.
The review considered the proposed synthetic fixture, its local validation summary, provenance requirements, and publication boundary. Raw working material and private locations are not public inputs.
METHOD / VERSION
Validity audit · 2026-08-23.
The audit separates code integrity from scientific validity and requires versioned provenance, baselines, representative traces, a clean reproduction path, and explicit non-claims.
OBSERVED
The DEMO packaging gate is incomplete.
The synthetic generator and scorer have internal integrity checks, but a compact versioned public packet, independent baselines, and clean-checkout reproduction are not yet complete.
LIMITATION
This is a governance decision, not a failed experiment.
The hold does not show that the fixture is valid or invalid. It establishes no model ranking, generalisation rate, controller reliability, field performance, or deployment safety.
NEXT DECISION
Extract a bounded DEMO repository.
Publish only after immutable provenance, sanitized versioned inputs and expected results, trivial baselines, representative traces, and a tested clean-checkout path are recorded.
Fixture counts and local pass statements previously shown here were withdrawn on 2026-08-17 because they could be mistaken for benchmark performance. The changelog records this correction. The current portfolio audit is in the public research registry.