Independent evidence review · draft scope only

Saved review · Original source proposals/2026-10-09-continuation/research-review/section01/review.md · original SHA256 f2f7ec9bf9afc18b5dc4d0ba5edf150d68782beec7cc95c90646d82424f1e97c · immutable published snapshot proposals/research-artifact-archive/00e2a98e88c7/f2f7ec9bf9afc18b5dc4d0ba5edf150d68782beec7cc95c90646d82424f1e97c.md · SHA256 f2f7ec9bf9afc18b5dc4d0ba5edf150d68782beec7cc95c90646d82424f1e97c

Bounded research/review evidence. This artifact does not certify physical manufacturing closure or later changed routes.

Independent review of the first five BFS research reports

9 October 2026. Target v1, scenario coordinated-discovery-v1. Scope: the original five reports under research-bfs/section01, their structured package and eight staged operations R01–R08. This is a technical review of evidence drafts, not route execution, inventory-completeness approval, terminal review or physical acceptance.

The reports are useful, bounded research drafts. Their main cited claims are supported by the primary sources independently opened for this review. They retain all 63 declared immediate requirements and correctly preserve unknown target demands, module hierarchy, workshop power, physical software hosts and independent qualification. Three bounded amendments were required in the originals: simpler polymer conditioning/screening alternatives, explicit separation of glass stress/strength measurements, and the RGB mask-pattern set and installed tooling allocation. These have now been inspected and confirmed in amended research-bfs/section02. All five amended reports and eight corrective findings are reviewed-as-draft. No staged corrective finding is rejected. Original and amended decisions remain identifiable by hashes in final-audit.json and decisions.json.

Scope, provenance and drift

Order was selected at revision52; original scoped records were captured at53; original author final reread was69. Reviewer captured current scoped records at74 in live-audit.json. Every one of the five selected route objects exactly matches the revision53 object, every report SHA256 matches its original manifest, and selected obligation counts remain34 + 7 + 5 + 9 + 8 =63. Comparison against original revision69 correction context also found no changes to the nine overlapping selected downstream routes (correction-context-drift.json). These are exact-object and hash checks on a single read snapshot, not a transaction or a fresh global gate pass.

Current canonical and Tool indices were searched by function/name/aliases; plausible records and selected routes were inspected. There is no canonical glass stress or glass edge-impact/bend identity in the searched revision74 index. optical-metrology selects frontier-optical-metrology-v1, consisting of microscope/length/jigs/light/surface inspection, and optical-inspection-microscope selects its compound-optics/stage construction route. Neither declares stress coupling, glass optical constants or an independent stress reference. Joint-pull, protective-barrier and abrasive-wheel fixtures have different geometry/acceptance duties and no selected manufacturing routes in this snapshot. Their existence cannot qualify a glass test.

mask-making has a candidate construction route discovery-mask-making-v1 in the catalog but no route selected in this scenario. moisture-verification-instrument likewise has no selected route. The molding selected route is discovery-memory-molding-equipment-v1; its polymer-film-extrusion-stretching-line requirement was already present in original correction context and remains reachable. Candidate routes or shared definitions do not establish usable equipment stock.

Decisions on the original eight operations

approved-as-draft below approves the corrective finding and ownership scope for subsequent work. It does not approve a live mutation, suitability, complete inventory, selected alternative or physical capability. needs-correction identifies a bounded draft amendment; its valid underlying finding is retained.

Independent source-to-claim check

All13 original cited primary sources were independently opened. The following gives claim scope rather than copying source recipes. Numerical commercial values remain unadopted.

H1 supports completed modular parts, screw/cable/BTB assembly and model-specific mechanical/electrical/firmware compatibility. Its current repair instructions also mention a controlled antistatic environment for the core module. This supports inventory/hierarchy obligations, not original manufacture or flagship equivalence. H2 supports permitted USB Type-C negotiation signaling and connector certification requirements within USB-IF's program, not this sink circuit or a frozen charging target. H3 supports an authorized-lab workflow and checking additional operator-specific requirements; it does not establish that every carrier requires PTCRB. Fairphone, USB-IF, PTCRB.

E1 describes antenna splits co-molded into a particular aluminum tablet enclosure and CNC finishing. It is valid precedent for a conditional RF integration option, not a mandatory phone insert or adopted geometry/alloy. E2 discusses commercial milling stability/workholding and chip evacuation; no manual-machine or cutting-fluid choice follows from it. These support R02's caution and required operation capability. Apple enclosure, Sandvik milling.

C1 supports moisture-related visible defects and hidden polymer-chain degradation; its material/grade distinctions matter. Pages2–5 also describe drying cabinets for small batches, several dryer arrangements and a simple TVI test limited to unreinforced Makrolon/Apec. TVI screens sufficient/insufficient dryness; it is not a quantitative laboratory moisture measurement. The brochure distinguishes low-moisture laboratory methods from simple weighing and should not be treated as a current universal weighing-instrument performance limit. Add these conditional simpler options to R03; no material selection, setpoint or target threshold is adopted. C2 supports matching machine size to material, shot/mold/clamp demands and mold heating/cooling capability; it does not qualify the project's manual apparatus. Covestro drying, Covestro equipment.

G1 supports potassium/sodium ion exchange and distinct cover/substrate functions. It does not provide our composition, dwell or process recipe; its approximate bath temperature remains context. G2 supports controlled edge-damage introduction followed by four-point bending. Its study uses an air/belt-slide arrangement, defined contact insert, force sensor and velocity photogates. These are one supported implementation, not automatically mandatory features of every acceptable earlier fixture. A different implementation needs its own evidence, reference transfer and geometry/control qualification. Preserve separate dynamic-contact and bending subassemblies/duties and destructive specimens under R05. Corning FAQ, Corning edge study.

G3 documents optical-waveguide stress measurement, a physical PC/software host, CS/DOL ranges and different wavelength variants. It advertises cross-section stress distribution and double-ion-exchange features, so profile capability is not disproved. Its ranges depend on the glass's photoelastic constants and refractive-index distribution; quoted accuracy applies to standard glass. The catalog does not independently qualify full depth profile/interior tension in our unspecified glass or establish instrument construction/calibration. Make this precise in the glass identity request while retaining unmet demands. Orihara FSM-6000X.

D1 supports vacuum organic evaporation with aligned fine metal masks and explicitly describes a different mask for each RGB color. It also places subsequent complementary layers/encapsulation on the same processed substrate. R08 should retain this installed mask set and its physical operation duties; a reticle-writing tool is not the resulting perforated metal mask. No pixel pitch, complete recipe or process yield is established. Samsung FMM.

D2 supports one rigid OLED glass-cover/frit/laser sequence; that commercial sequence also has seal preconditioning and UV hardening of an additional mother-glass seal. If this specific implementation is selected, preserve those duties or justify a distinct one-off cell-level alternative. D3 supports alternating inorganic/organic barrier layers in a flexible-panel example and particle/pinhole mitigation. Neither supports our rigid/TFE combination's thermal/chemical compatibility or lifetime, and neither establishes TFE cannot be used on rigid glass. R07 correctly keeps alternatives separate. Samsung encapsulation, Samsung TFE.

Hierarchy, inventory and alternatives

The reports distinguish installed modules from tools that manufacture them. Repeated substrate/FPC/material IDs are not automatically extra incorporated stock. R06 should produce an explicit state/quantity ledger: bare prepared glass, backplane-bearing glass, OLED-bearing intermediate and encapsulated panel; identify actual driver/FPC installation stage. If later schema revisions split states, create only meaningful independently understandable outputs, retain material/utility/tool obligations through the actual consumer, and keep the selected graph acyclic. A same-name edge deletion is insufficient evidence of allocation.

Operation inventory remains incomplete. Known open duties include assembly contacts/protected storage/waste; machining cutting/chip collection and qualified stock; cover feed conditioning/mold vents/ejection/thermal capacity; glass cut/edge/rinse/bath-contact tools, references, coupon stock and waste; display installed FMMs/carriers, process monitors, clean transfers, contacts, independent photometry and barrier-life tests. Whether these are supplied by a reusable platform or a distinct consumer is a quantity/capacity/availability decision. Construction/commissioning inventory never automatically replenishes a later production job's fuel, water, salt, feed, cutting compound or cleaning stock.

Modular screw/clip versus bonded final assembly, machined versus near-net metal versus polymer enclosure, molding versus sheet machining/compression forming, chemical versus thermal strengthening, and lid sealing versus TFE are coherent alternate methods. They remain hypotheses except for the narrow source precedents above. None is a proved shorter from-nature route or a capability-equivalent substitute. The newly cited small-batch drying/screening options provide a genuine simpler-route investigation while leaving grade-specific and low-moisture requirements open.

Identity and ownership recommendations

glass-stress-profile-instrument: approve as a distinct *draft capability request* compared with current microscope/length routes, subject to owner review of extension versus separate identity. Include coupling optics/liquid, wavelength/detection, glass constants, surface/depth/profile demand, independent references and earlier physical host/loading/power. Manufacturer feature claims never certify our device. No new ID has been inserted or qualified.

glass-edge-impact-and-bend-fixture: approve as a distinct *draft capability request* compared with wheel/joint/barrier fixtures, with R05 corrections. Prefer reusable force/length/motion/containment components where compatible; identify dynamic-contact and four-point-bend stages explicitly. A combined parent is acceptable if its independently understandable subassemblies remain exposed. No assumption that the instrument under test independently certifies its own load/velocity.

Coordinator owns target/interface freeze and allocation decisions. Mechanical owns enclosure/covers/tooling and physical glass/bath duties; facilities owns display intermediate/encapsulation/operation allocation and chemical environment; electrical/metrology owns specialized glass and panel measurement/reference/host duties. Findings were sent directly to the original research agent for staged same-five amendments only. Discovery owners must reread current canonical records before changes and use the shared atomic gate; reviewer makes no live registry or generated-page edits.

Remaining limits and handoff

Five original reports reviewed against their saved hashes; all13 source claims checked within bounded scope; five corrective findings approved-as-draft and three need amendment. There are zero executable route, complete-inventory, quantitative suitability, physical-output, endpoint or closure approvals. The target acceptance test denominator remains unfrozen. No sixth item was researched. Next work is bounded amendment verification and separately owned physical correction discovery; numerical target and compatibility gaps stay visible.

Final amended-draft check

Final reviewer snapshot84: all five selected producing route objects still exactly match revision53, original and amended report hashes match their respective manifests, and amended reports retain every one of63 immediate requirement IDs. No canonical glass stress or glass edge-impact/bend identity was found in the current index. final-audit.json preserves hashes and selections; later concurrent changes require owner rebase before implementation.

Amended section02 reports, changes, evidence and correction entries confirm R03's conditional cabinet/grade-limited qualitative screening, R05's manufacturer-feature versus actual-glass qualification distinction and separate damage/bending stages, and R08's RGB FMM set/installed carrier allocation. Covers retains the existing polymer-film line requirement; display identifies the unselected mask-making candidate. Original R03/R05/R08 remain needs-correction in provenance; amended R01–R08 are reviewed-as-draft. All recipe/stock/reference/target gaps remain open. G2's commercial drive/sensor arrangement is a supported implementation example, not a universal prerequisite for alternatives.

Original research agent section02 discovery feedback was inspected: its three labels correctly distinguish declared-inventory fact, method-supported obligation and unresolved decision/suitability. Reviewer discovery-feedback.md independently supplies route examples, owners and lightweight prompt checks. Physical correction discovery stays with assigned owners. This bounded review is complete; no further same-five amendment is required for these findings.