EVIDENCE / ASSURANCE

What has actually been proved — and what has not.

Every result keeps its tested environment, challenge, result, source, limitation and next reality gate visible.

QUALIFIED / SOFTWARE ONLY

Common authority/control

Tested environment: Bounded software contract
Challenge: wrong authority, stale context, malformed and rebound inputs
Result: 32/32 PASS
Evidence reference: Common control qualification
Does not prove: Common control schema/validator only; not production or universal agent safety
Open reality gate: Broader real-world integration

QUALIFIED / SOFTWARE ONLY

Tested environment: Common receipt contract
Challenge: uncertain effect, acknowledgement/effect separation, reconciliation
Result: 18/18 PASS
Evidence reference: Common receipt qualification
Does not prove: No universal receipt implementation across every component/domain
Open reality gate: Mapped receipt implementations across target domains

QUALIFIED / BOUNDED LOCAL SOFTWARE

Tested environment: Pinned bounded local software composition
Challenge: stale context, replay, STOP, projection separation and UNKNOWN preservation
Result: 17/17 PASS, recorded twice
Evidence reference: Bounded local integration qualification
Does not prove: Not production, universal federation or physical safe-state proof
Open reality gate: Representative deployment qualification

QUALIFIED / SOFTWARE ONLY

QUALIFIED / SEALED / CANONICAL

Tested environment: Pinned joined software composition
Challenge: shared semantics without authority merger
Result: 48/48 PASS + 9/9 sealed artifacts
Evidence reference: Shared Sovereign Fabric qualification
Does not prove: Not arbitrary federation, production, national, clinical, regulatory or physical deployment
Open reality gate: Larger/degraded and physically separated federation

QUALIFIED / SOFTWARE-SIL ONLY

Tested environment: Common and bounded vertical software/SIL paths
Challenge: in-flight interruption, authority revocation, restart, partial/unknown effect
Result: Common STOP 10/10 plus bounded vertical profiles
Evidence reference: STOP qualification
Does not prove: STOP is not rollback; no universal physical safe-state or remote-latency guarantee
Open reality gate: Domain-specific HIL/physical STOP and measured latency

QUALIFIED / SOFTWARE-SIL ONLY

Tested environment: 409,600 represented servers, 15 kernels
Challenge: semantic equivalence and accounting under distributed partitioning
Result: PASS_DISTRIBUTED_CONTROL_EQUIVALENT; semantic_exact=true
Evidence reference: Data-centre SIL qualification
Does not prove: Not physical capacity, maximum scale, SLA, energy saving or production throughput
Open reality gate: Representative physical facility and production-like trials

QUALIFIED / SEALED / REAL GPU

Tested environment: One physical host, two CUDA GPUs
Challenge: distributed execution, refusal, replay, STOP/withdrawal and join failure
Result: Real two-GPU CUDA/NCCL DDP PASS
Evidence reference: Real-GPU qualification
Does not prove: No rack/site/power/cooling independence or physical data-centre federation
Open reality gate: Independent physical hosts/racks/sites

QUALIFIED / SYNTHETIC SOFTWARE

Tested environment: Synthetic GP-to-hospital fixture
Challenge: release revalidation, ACK/effect/admission/responsibility separation and reconciliation
Result: 40/40 repeated byte-identically; later 44/44 regression
Evidence reference: Synthetic institutional-handoff qualification
Does not prove: Not live NHS, clinical safety, regulatory acceptance or national deployment
Open reality gate: Authorised representative institutional/clinical pilot

QUALIFIED / SYNTHETIC SOFTWARE

Tested environment: One pinned synthetic legacy work-order schema
Challenge: malformed/stale material, changed material, lost response and schema drift
Result: Current 17/17 PASS
Evidence reference: Synthetic translation-profile qualification
Does not prove: Not a universal translator or arbitrary semantic discovery
Open reality gate: Separately qualified real legacy system/profile

QUALIFIED / NON-ACTUATING SOFTWARE/API

Tested environment: Exact transport information/API profiles
Challenge: stale/conflicting/malformed inputs, restart, STOP and provenance
Result: 31/31 + 27/27 + 26/26 + 8/8 PASS
Evidence reference: Non-actuating infrastructure-information qualification
Does not prove: No signalling, PLC, SCADA, interlocking or vehicle actuation; no generic standards certification
Open reality gate: Representative operational integration; separate actuation qualification if authorised

QUALIFIED / BOUNDED LOCAL SOFTWARE

Tested environment: Pinned bounded local execution context
Challenge: wrong runtime identity, malformed/host commands, drift and evidence sufficiency
Result: R4 3/3; R5-R7 PASS; R8 23/23 applicable controls
Evidence reference: Bounded agent-execution qualification
Does not prove: No universal agent safety or deployed privileged-host claim
Open reality gate: Privileged sealed-release host promotion

QUALIFIED / PHYSICAL SIMULATOR ONLY

Tested environment: Closed-loop physical simulator/SITL
Challenge: closed-loop motion, sensor degradation, obstacle response and recovery
Result: 20/20 PASS twice with stable semantic hash
Evidence reference: Physical-simulator qualification
Does not prove: No HIL, field/ocean, physical sensors/comms, BVLOS or regulatory evidence
Open reality gate: Target hardware/HIL and field qualification

QUALIFIED / SOFTWARE-SIL ONLY

SOURCE VERIFIED / NOT QUALIFIED

Tested environment: Current legacy REST path
Challenge: documentation claims stronger restart durability than active process-local volatile persistence supports
Result: Public restart-durable / durable exactly-once claim withheld
Evidence reference: Source review: legacy REST durability boundary
Does not prove: Cannot claim restart-durable exactly-once on the current REST path
Open reality gate: Durable implementation or narrowed docs, then crash/restart/race qualification