# ISS fidelity review — work in progress

**The reconstruction has not passed the full rubric. This playtest contains improvements under review; publishing a demo does not certify full fidelity.**

This partial review records 91 individual criteria. Each ledger row retains the development candidate it actually reviews; latest reviewed source is `revision3-9f7287342518`. 370 unreviewed, 69 unverified, 3 pass, 19 failed.

Scores are manual visual judgments against the rubric anchors. Unknown evidence is left unverified, never converted into a zero or a passing score. Geometry and movement tests are reported separately.

## Recorded visual failures

- **S08.04 · 6/10:** OCO-3, ECOSTRESS and H8 use distinct original visualization meshes. The reconstructed MAXI, HISUI, GEDI and H9 are identifiable by broad shape but their covers and external details remain visibly simplified against the saved photographs.
- **S08.05 · 6/10:** Duplicate legacy grapples were removed and a single authored fixture fitted per reconstruction. Generic circular interface collars lack the photographed latch and connector detail; geometric contact does not establish that match.
- **S08.06 · 5/10:** Exposed facility utility runs have not been traced to named connectors from photographs. The current source has sparse routing, and the new instruments do not yet have verified service connections.
- **S08.07 · 5/10:** The nadir capture exposes a largely featureless platform bottom. Several reconstructed payload bottoms are also plain closed rectangles. Underside equipment and panel detail require reference evidence and reconstruction.
- **S08.08 · 6/10:** Dark apertures, painted interfaces and blankets are distinguishable, but H9/HISUI silver covers currently read as nearly uniform white material. Wrinkle/seam placement is not matched to flight photographs.
- **S11.03 · 5/10:** Coarse cover shape and absent straps/brackets; dated inventory and fifth-tank placement remain unresolved.
- **S11.04 · 6/10:** EVA rim, adjoining handrails and cover attachments do not yet match the close photograph's detail.
- **S11.05 · 5/10:** Relief improved, but lettering, dated insulation/stowage and fittings still fall short of the close photograph.
- **S11.07 · 1/10:** The requested reference-quality visible Quest crew-lock interior is missing.
- **S11.08 · 6/10:** Distinct aluminum response improves the prior matte result; unmeasured finish, incomplete material segmentation and remaining source simplification prevent reference-quality acceptance.
- **S22.03 · 6/10:** Both reconstructed IDAs have an overly simple ring and flange profile. Dimensional matching and local attachment depth remain unsupported.
- **S22.05 · 5/10:** IDA2 and IDA3 lack resolved docking guides, petal/cover and actuator detail. The ring reconstruction does not meet close reference quality.
- **S22.06 · 6/10:** PMA black blanket faces have sparse closure seams and coarse segmentation; source collars are visibly faceted.
- **S34.03 · 7/10:** Root-kit reconstruction remains visibly simpler than the flight installation hardware; fitting to a source canister is not one-to-one kit reconstruction.
- **S34.04 · 7/10:** Deployment fittings and mandrel covering lack the reference's small structural and material detail.
- **S34.07 · 5/10:** Complete supported power-cable routing and destination connector identities are still missing.
- **S38.04 · 6/10:** The original meshes improve frame and equipment identity, but close ELC-2 views show large coarse boxes with blurred atlas detail. Source provenance does not meet the close-up criterion.
- **S38.08 · 6/10:** Frame, blanket and radiator-like surfaces differ, but texture resolution is visibly insufficient on close ELC-2 equipment and the carried-unit materials are not independently calibrated.
- **S39.08 · 5/10:** ESP1's blankets have especially blurred vertical streaks. ESP2/3 also use coarse repeated cloth atlases; separate base materials alone do not establish realistic close surfaces.

## Latest focused review

- B.01: both OpenAI marks pass the recorded visual review at 9/10. Separate suit and station evidence includes four static directions each, 120 motion frames each, and physical backing/fastener measurements. All 240 rectified glyph images were inspected; no missing strokes or flicker was observed. This is sampled review evidence, not a mathematical proof of every continuous viewpoint. [Open the branding review](OpenAI-branding-review.html).
- S34: all six iROSAs now have separate front and back surfaces, two blanket strips, cell-level coverglass, outer booms, split tip mandrels and fitted roots. Their full geometries do not intersect the source station. Twelve measured foot clearances are 0.199–0.201 mm. The source rim-slot problem is corrected; full root-kit identity and rear electrical routing remain unresolved. No complete solar-array package passes.

## Bounded measurements

- Kibo: twelve payloads, 66 distinct payload pairs, no pairwise triangle intersections. Maximum payload-to-host closest distance is 0.08774 mm. This does not prove every individual support or every host penetration is correct.
- BEAM: outer diameter 3.23 m and length with grapple 4.01 m, calibrated to NASA’s published envelope. Its exact surface profile and collar stack remain unverified.
- Candidate37 station probe run: 146 reachable body/camera approaches across forty packages passed; 94 directions were obstructed or missed the geometry and remain untested. Candidate38 Quest geometry changes require a new run.
- Animated suit: 72 sampled poses remain within the collision capsule, with a minimum sampled margin of about 3.36 cm. Continuous joint-space and tether coverage are still separate checks.

## Still required

All 38 earlier child contact sheets (380 images) and six new iROSA sheets (60 angles) have been inspected. Continue matched photograph comparisons, refinement of coarse original fittings and textures, Earth/cloud/orbit sequences, the remaining criterion assessments, current collision coverage, and production performance testing. Historical findings remain tied to their own candidates. No full-fidelity or deployment claim follows from the partial results above.

[Full acceptance rubric](ISS-fidelity-rubric.html) · [Complete evidence ledger](ISS-fidelity-evidence-ledger.csv)