Native source anchorapply_nqfc_c3_commit_certificate_durablePersist then advance
PRODUCES / NEXT BOUNDARY
End of this selected sliceNo further effect or completion is inferred.
Source selection & scope
Selected local C3 dependencies only. Open the dedicated BraidCell planar view for the persistence lane and exact witness study; live validator events are not connected.
src/nqfc_c3_consensus.rs
Illustrated refusal · Missing local lock
Certificate verification does not create the applying node's missing durable commit lock.
Illustrated refusal · Different exact witness
CommitQC must refine the supplied proposal and exact LockQC witness.
Relationships identify selected source artifacts and prerequisites. Reading order and motion are presentation controls, not execution traces.
End of this selected sliceNo further effect or completion is inferred.
Source selection & scope
Queue admission, routing and dispatch are distinct from job completion and durable recovery. Historical scalar/batch and worker measurements remain separate.
src/glap_udp1/worker_plane.rs
Backpressure · Queue full
A full bounded queue returns backpressure with the item retained by its contract; it is not completed application work.
Illustrated refusal · Allocation not admitted
Reject an invalid or unadmitted allocation profile before materializing its worker buffers.
Relationships identify selected source artifacts and prerequisites. Reading order and motion are presentation controls, not execution traces.
USE CASE
Study how a caller routes work without silently dropping a full queue.
EVALUATION SCOPE
Selected admission, sharding and queue source path. Historical benchmarks are separate; this diagram executes no worker job.
Read the complete mechanism
Input batchOperations + profile
The caller supplies the bounded batch and selected worker profile.
PlanAllocation budget
Compute the plan before materializing the admitted worker buffers.
MaterializeAdmitted buffers
Materialize the worker plane only after its allocation plan is admitted.
RouteStripe ownership
Resolve a flow to its admitted stripe and worker mapping.
QueueItem or backpressure
Enqueue bounded work, or return the full-queue outcome without pretending it executed.
Dispatch / snapshotReceipt + queue state
Dequeue selected work and inspect the retained queue receipt and state; dispatch is distinct from application completion.
Refusal to inspect
A full bounded queue returns backpressure; it must not be drawn as completed work.
A full bounded queue returns backpressure; it must not be drawn as completed work.
Compute / 04
GL-C / canonical IR
Trace a typed operation from source to execution.
WHAT TO INSPECT
Compilation and execution share the checked type model and canonical IR boundary.
PLANAR STUDY / SOURCE DEPENDENCIESSelect an artifact to inspect its contract
Selected artifactRequired dependencyAlternative / separate boundarySOURCE STUDY · NOT LIVE TELEMETRY
Reading pace · animation timing is independent of execution latency.
SOURCE CONTRACT
REQUIRES
Input to this selected source sliceIts caller owns admission and authority.
Native source anchorcompileNamed operation
PRODUCES / NEXT BOUNDARY
Type checkCanonical source · Artifact dependency
REQUIRES
GL-C sourceCanonical source · Artifact dependency
Native source anchorcompile_paired_observedValidated model
End of this selected sliceNo further effect or completion is inferred.
Source selection & scope
Compilation and execution are separate native callers and costs. Source artifacts arriving from outside their caller still require decode/verification.
The selected admitted execution limits and supported operation remain required; this browser runs no arbitrary source.
Relationships identify selected source artifacts and prerequisites. Reading order and motion are presentation controls, not execution traces.
USE CASE
Define a bounded computation once and inspect its typed inputs, result and refusal.
EVALUATION SCOPE
Language/IR source study. This browser diagram does not compile or execute arbitrary visitor code.
Read the complete mechanism
GL-C sourceNamed operation
Provide explicit arguments, result type and operation definition.
Type checkValidated model
Resolve types and reject incompatible inputs before preparing execution.
Checked IRCanonical program
Compile the selected operation into the shared canonical representation.
PrepareAdmitted limits
Prepare the checked operation under its selected execution limits.
ExecuteTyped arguments
Run the selected IR operation with the admitted input values.
ResultValue or refusal
Return the typed value or explicit execution error; compilation is measured separately.
Refusal to inspect
An invalid type, unsupported operation or exhausted admitted budget must not become a successful result.
An invalid type, unsupported operation or exhausted admitted budget must not become a successful result.
Authority / 05
Policy & Need
Keep missing evidence visible in a decision.
WHAT TO INSPECT
ObservedOutcome preserves Need; a pure decision is separate from authority and protected effects.
PLANAR STUDY / SOURCE DEPENDENCIESSelect an artifact to inspect its contract
Selected artifactRequired dependencyAlternative / separate boundarySOURCE STUDY · NOT LIVE TELEMETRY
Reading pace · animation timing is independent of execution latency.
SOURCE CONTRACT
REQUIRES
Input to this selected source sliceIts caller owns admission and authority.
Native source anchorbindSubject + action
PRODUCES / NEXT BOUNDARY
EvaluateRequest inputs · Required input
REQUIRES
Input to this selected source sliceIts caller owns admission and authority.
Native source anchorprepareTyped expressions
PRODUCES / NEXT BOUNDARY
EvaluatePrepared policy · Required input
REQUIRES
Input to this selected source sliceIts caller owns admission and authority.
Native source anchorinvoke_observedRequired facts
PRODUCES / NEXT BOUNDARY
EvaluateObserved facts · Required input
REQUIRES
RequestRequest inputs · Required input
PolicyPrepared policy · Required input
ObserveObserved facts · Required input
Native source anchorObservedOutcomeDecision + Need
PRODUCES / NEXT BOUNDARY
Effect boundarySeparate effect admission · Separate boundary
Decision traceDecision support · Artifact dependency
REQUIRES
EvaluateSeparate effect admission · Separate boundary
Separate authorityCaller-owned effect and authority admissionSeparate admission
PRODUCES / NEXT BOUNDARY
End of this selected sliceNo further effect or completion is inferred.
REQUIRES
EvaluateDecision support · Artifact dependency
Native source anchordecision_supportDecision support
PRODUCES / NEXT BOUNDARY
End of this selected sliceNo further effect or completion is inferred.
Source selection & scope
Decision support is distinct from an effect receipt. Observed facts, provider freshness and actual authority/effect admission require their own qualified caller.
Unresolved required observations remain visible. A Need or decision is not proof that an external effect completed.
Separate qualification · Effect not admitted here
This selected pure caller has no protected effect or principal authentication. Separate authority and completion checks remain the effect caller's responsibility.
Relationships identify selected source artifacts and prerequisites. Reading order and motion are presentation controls, not execution traces.
USE CASE
Investigate an access decision with incomplete observations and inspect the separate effect boundary.
EVALUATION SCOPE
Selected pure decision and observation source study. Authority, customer effects and provider freshness require their own caller qualification.
Read the complete mechanism
RequestSubject + action
Select the resource, authority and operation under evaluation.
PolicyTyped expressions
Prepare the selected decision expression and its admitted inputs.
ObserveRequired facts
Resolve the named observations under the selected provider boundary.
EvaluateDecision + Need
Keep unresolved required information explicit in the decision result.
Effect boundarySeparate admission
The selected pure caller performs no protected effect or principal authentication; effect admission belongs to a separately qualified caller.
Decision traceDecision support
Retain the selected decision context without treating its trace as a signed receipt or proof of effect completion.
Refusal to inspect
Missing required observations cannot be portrayed as a completed authorized effect.
Missing required observations cannot be portrayed as a completed authorized effect.
Authority / 06
GS sessions
Follow a session through refresh and revocation.
WHAT TO INSPECT
Server-owned state binds opaque handles, device proof, freshness and monotonically checked counters.
PLANAR STUDY / SOURCE DEPENDENCIESSelect an artifact to inspect its contract
Selected artifactRequired dependencyAlternative / separate boundarySOURCE STUDY · NOT LIVE TELEMETRY
Reading pace · animation timing is independent of execution latency.
SOURCE CONTRACT
REQUIRES
Input to this selected source sliceIts caller owns admission and authority.
End of this selected sliceNo further effect or completion is inferred.
REQUIRES
Check requestRevocation · Alternative path
Native source anchorrevokeAuthority state
PRODUCES / NEXT BOUNDARY
End of this selected sliceNo further effect or completion is inferred.
REQUIRES
IssuePinned journal · Alternative path
Native source anchorimport_native_journal_pinnedPinned journal head
PRODUCES / NEXT BOUNDARY
End of this selected sliceNo further effect or completion is inferred.
Source selection & scope
Issue, request, refresh, revoke and restoration are lifecycle alternatives. The native journal is authenticated plaintext; callers protect its bytes and key.
src/gs_pq_v2_session_control.rs
src/gs_pq_v2_session_control/native_journal.rs
Illustrated refusal · Refresh replay
A consumed refresh continuation must not become another admitted continuation.
Illustrated refusal · Revoked authority
Revocation is an authority-state transition, not a normal step every request executes.
Illustrated refusal · Stale journal head
An authenticated old journal is not fresh merely because its tag verifies; the pinned caller checks the trusted head.
Relationships identify selected source artifacts and prerequisites. Reading order and motion are presentation controls, not execution traces.
USE CASE
Study why an old refresh continuation must not become a second valid continuation.
EVALUATION SCOPE
Session-control and native-journal source study. No customer session, device or production cluster is opened here.
Read the complete mechanism
IssueOpaque handles
The selected authority issues the session and refresh handles.
Bind devicePossession proof
Bind the selected device and channel context to their server-owned state.
Check requestCounter + freshness
Check the request counter and freshness against authoritative session state.
RefreshOne continuation
Consume the admitted refresh continuation and retain its next generation.
RevokeAuthority state
Record revocation under the selected authoritative mutation boundary.
RestorePinned journal head
Restore the native journal against its trusted head; authentication alone does not admit a stale snapshot.
Refusal to inspect
A reused refresh continuation, stale authenticated snapshot or revoked authority must not regain admission.
A reused refresh continuation, stale authenticated snapshot or revoked authority must not regain admission.
Data / 07
GL-DDBS / MVCC
Read a snapshot while writes are checked.
WHAT TO INSPECT
The selected durable path binds transaction and storage receipts while preserving conflict and rollback outcomes.
PLANAR STUDY / SOURCE DEPENDENCIESSelect an artifact to inspect its contract
Selected artifactRequired dependencyAlternative / separate boundarySOURCE STUDY · NOT LIVE TELEMETRY
Reading pace · animation timing is independent of execution latency.
SOURCE CONTRACT
REQUIRES
Input to this selected source sliceIts caller owns admission and authority.
Native source anchormulti_operation_requestSnapshot version
End of this selected sliceNo further effect or completion is inferred.
Source selection & scope
This selected source is a local qualification runner over transaction and durable-store callers. The Lab neither invokes it nor connects a customer store.
End of this selected sliceNo further effect or completion is inferred.
Source selection & scope
The selected source is a local research report. Noise simulation and reference bootstrapping are explicitly distinct from production cryptographic admission.
Parameter validation must admit the selected ring/moduli profile before the corresponding local arithmetic.
Research boundary · Production FHE boundary
Local arithmetic, typed artifact round-trips, a noise model and reference bootstrap do not close production FHE security.
Relationships identify selected source artifacts and prerequisites. Reading order and motion are presentation controls, not execution traces.
USE CASE
Understand where parameters, approximation and a noise budget affect an encrypted-compute evaluation.
EVALUATION SCOPE
Local research arithmetic and reference-bootstrap study. Production FHE security and external cryptographic review remain open.
Read the complete mechanism
ParametersRing + moduli
Validate the selected native RLWE/CKKS parameter profile.
Ring arithmeticPolynomial product
Inspect the selected local negacyclic arithmetic over the admitted ring.
RNS / NTTLocal round-trips
Follow the report's residue and transform round-trip checks.
Typed artifactsCiphertext + bootstrap key
Inspect local typed ciphertext/key artifact construction and canonical round-trips, without a production encryption claim.
Noise studyOperation budget
Track the selected noise study and reference-bootstrap boundary.
Reference bootstrapLocal reference path
Read the selected reference-bootstrap path separately from production FHE security admission.
Refusal to inspect
An unsupported parameter profile is rejected by parameter validation before the selected arithmetic is admitted. Production security remains a separate qualification.
An unsupported parameter profile is rejected by parameter validation before the selected arithmetic is admitted. Production security remains a separate qualification.
Research / 12
SPDZ2k shares
See how an authenticated product is checked.
WHAT TO INSPECT
The local algebraic core combines shares, Beaver-style multiplication and MAC verification over Z / 2^k Z.
PLANAR STUDY / SOURCE DEPENDENCIESSelect an artifact to inspect its contract
Selected artifactRequired dependencyAlternative / separate boundarySOURCE STUDY · NOT LIVE TELEMETRY
Reading pace · animation timing is independent of execution latency.
SOURCE CONTRACT
REQUIRES
Input to this selected source sliceIts caller owns admission and authority.
End of this selected sliceNo further effect or completion is inferred.
Source selection & scope
This slice prepares the native request and binary response contracts. The actual engine caller and replay/audit operations are separate boundaries, not a single execution chain.
src/universal_intent_external_api.rs
Illustrated refusal · Malformed envelope
Reject an unsupported binary operation or malformed admitted fields before native guard request construction.
Separate qualification · External effect boundary
Guard, replay and audit-export contracts do not themselves authorize this website to execute arbitrary external effects.
Relationships identify selected source artifacts and prerequisites. Reading order and motion are presentation controls, not execution traces.
USE CASE
Study admission of an agent, API or wallet intent before a separate effect executor is considered.
EVALUATION SCOPE
Selected binary contract and request-construction study. The actual engine caller is separate; this website grants no arbitrary effect authority.
Read the complete mechanism
IntentUIG1 envelope
Encode the selected intent under the versioned external binary contract.
Policy packFamily + profile
Select the named policy pack and admitted action family.
DecodeBounded fields
Decode the admitted binary request and bind its selected context.
Guard requestExisting Core caller
Construct the native guard request for the existing engine caller; request construction is distinct from invoking that caller.
ReceiptUIB1 response
Return the fixed receipt and canonical binary payload for the selected outcome.
Replay / auditRetained evidence
Inspect the separate replay and audit-export boundaries without implying effect execution.
Refusal to inspect
A malformed binary envelope, unsupported profile or unbound request cannot be promoted to an authorized effect.
A malformed binary envelope, unsupported profile or unbound request cannot be promoted to an authorized effect.
Compute / 17
GL-COD / object admission
Bind an object before composing a computation.
WHAT TO INSPECT
Descriptor, schema, payload and evidence contribute separately to the canonical object and admission receipt.
PLANAR STUDY / SOURCE DEPENDENCIESSelect an artifact to inspect its contract
Selected artifactRequired dependencyAlternative / separate boundarySOURCE STUDY · NOT LIVE TELEMETRY
Reading pace · animation timing is independent of execution latency.
SOURCE CONTRACT
REQUIRES
Input to this selected source sliceIts caller owns admission and authority.
End of this selected sliceNo further effect or completion is inferred.
Source selection & scope
Preserve the broader compute-over-data capability as a separate caller to qualify. This diagram accurately exposes the selected object-admission asset.
Separate qualification · Computation remains separate
This caller returns an object-admission receipt. It does not execute a query or prove a durable computation result.
Relationships identify selected source artifacts and prerequisites. Reading order and motion are presentation controls, not execution traces.
USE CASE
Inspect what an object-admission receipt actually binds before a compute-over-data integration.
EVALUATION SCOPE
This selected source admits objects and encodes reports. Query execution, data journals, durable compute results and GLAP integration remain separate callers to qualify.
Read the complete mechanism
DescriptorObject identity
Validate the selected object descriptor and its bounded identifiers.
SchemaUnique fields
Validate schema identities and field uniqueness before hashing the canonical schema.
PayloadBounded bytes
Validate and hash the non-empty bounded payload.
EvidenceAvailability bindings
Validate required evidence and its digest bindings; retain all blockers.
Canonical objectFour bound hashes
Combine descriptor, schema, payload and evidence hashes into the canonical object identity.
Admission receiptDecision + blockers
Return admitted status, blocker list and deterministic receipt hash. This is not a compute result.
Refusal to inspect
Invalid object identity, schema, payload or required evidence remains a blocker; a receipt hash does not prove a computation executed.
Invalid object identity, schema, payload or required evidence remains a blocker; a receipt hash does not prove a computation executed.
Money / 18
Offline wallet / GS
Keep an offline replay tuple bound to authority.
WHAT TO INSPECT
The selected composition checks wallet replay context and GS authority without claiming that offline acceptance is final settlement.
PLANAR STUDY / SOURCE DEPENDENCIESSelect an artifact to inspect its contract
Selected artifactRequired dependencyAlternative / separate boundarySOURCE STUDY · NOT LIVE TELEMETRY
Reading pace · animation timing is independent of execution latency.
SOURCE CONTRACT
REQUIRES
Input to this selected source sliceIts caller owns admission and authority.
Width 4 · 4096 requested bytes · 4096 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 4 · 8192 requested bytes · 4096 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 4 · 12288 requested bytes · 4096 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 4 · 16384 requested bytes · 4096 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 4 · 24576 requested bytes · 4096 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 4 · 32768 requested bytes · 4096 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 4 · 49152 requested bytes · 4096 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 4 · 65536 requested bytes · 4096 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 8 · 4096 requested bytes · 4096 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 8 · 8192 requested bytes · 8192 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 8 · 12288 requested bytes · 8192 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 8 · 16384 requested bytes · 8192 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 8 · 24576 requested bytes · 8192 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 8 · 32768 requested bytes · 8192 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 8 · 49152 requested bytes · 8192 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 8 · 65536 requested bytes · 8192 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 12 · 4096 requested bytes · 4096 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 12 · 8192 requested bytes · 8192 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 12 · 12288 requested bytes · 12288 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 12 · 16384 requested bytes · 12288 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 12 · 24576 requested bytes · 12288 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 12 · 32768 requested bytes · 12288 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 12 · 49152 requested bytes · 12288 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 12 · 65536 requested bytes · 12288 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 16 · 4096 requested bytes · 4096 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 16 · 8192 requested bytes · 8192 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 16 · 12288 requested bytes · 12288 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 16 · 16384 requested bytes · 16384 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 16 · 24576 requested bytes · 16384 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 16 · 32768 requested bytes · 16384 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 16 · 49152 requested bytes · 16384 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 16 · 65536 requested bytes · 16384 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 24 · 4096 requested bytes · 4096 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 24 · 8192 requested bytes · 8192 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 24 · 12288 requested bytes · 12288 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 24 · 16384 requested bytes · 16384 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 24 · 24576 requested bytes · 24576 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 24 · 32768 requested bytes · 24576 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 24 · 49152 requested bytes · 24576 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 24 · 65536 requested bytes · 24576 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 32 · 4096 requested bytes · 4096 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 32 · 8192 requested bytes · 8192 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 32 · 12288 requested bytes · 12288 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 32 · 16384 requested bytes · 16384 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 32 · 24576 requested bytes · 24576 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 32 · 32768 requested bytes · 32768 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 32 · 49152 requested bytes · 32768 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 32 · 65536 requested bytes · 32768 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 48 · 4096 requested bytes · 4096 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 48 · 8192 requested bytes · 8192 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 48 · 12288 requested bytes · 12288 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 48 · 16384 requested bytes · 16384 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 48 · 24576 requested bytes · 24576 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 48 · 32768 requested bytes · 32768 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 48 · 49152 requested bytes · 49152 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 48 · 65536 requested bytes · 49152 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 64 · 4096 requested bytes · 4096 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 64 · 8192 requested bytes · 8192 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 64 · 12288 requested bytes · 12288 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 64 · 16384 requested bytes · 16384 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 64 · 24576 requested bytes · 24576 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 64 · 32768 requested bytes · 32768 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 64 · 49152 requested bytes · 49152 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
Width 64 · 65536 requested bytes · 65536 effective body bytes · 20 release loops. Times are recorded local means per batch, not percentiles or network latency.
1,037.544–1,069.013 ns/op in the retained local path.
Explicit backpressure · 0 observed dropped items · allocation-plan maximum 242,944 bytes. Recovery and end-to-end latency excluded.
WORKERPLANE / CONCURRENT MATRIX4 profiles
187.541–1,183.939 ns/op across the retained configurations.
Thread creation/join included · disjoint stripes · 0 observed dropped items. These ranges do not establish a universal parallel speedup.
Measurement identity, scope and next experiment
Retained release micro-benchmark: 20 loops; scalar/batch parity checked before reading time ratios. Worker matrices use their own workload and scheduling profile. The source report is identified by its exact digest in the qualification record; these are historical observations, not a qualification of later source changes.
The following experiment requires the same artifacts, workload, semantic outcomes and environment: skewed queues, contention, interrupted work, recovery and an independent result oracle. RSS, allocation, compilation and execution must retain separate measurement scopes.
Evidence has a named scope.
Source studies explain selected callers. The native execution tab runs three real GLAP cases. Recorded measurements describe their retained artifacts. Finite navigation certificates concern their exact model.
The presentation model checks six explanation positions, pause/refusal retention and the distinction between a source study, a recorded measurement and an observed native result. It has 16 proved checks and one deliberately refuted negative control across 184 finite valuations. Native Rust transition parity is checked separately. These certificates do not prove the browser, native mechanisms, cryptography, performance or distributed guarantees.
Verification uses the existing GL-Core FormalVerify service on this exact module, with the compiler, verifier, codec, Rust toolchain, OS and hardware in the trusted base. Model source SHA-256: 3de2649a90180b56292ba7b53d41d54a1f68d9834ed06fffed353382b386a3e5
The retained finite certificates concern their original control model. They do not certify the later dependency graphs, boundary scenarios or browser adapter.
SELECTED NATIVE CALLERS
Inside the execution.
Inspect a component and its actual caller dependencies. GLAP measures six server call windows; BraidCell / C3 shows selected source relationships.
ONE WORKSPACE / THREE NATIVE EXPERIMENTS
Allow. Repeat. Refuse.
Trigger a request. Inspect what the runtime returned. Keep the exact evidence.
Run an experiment to obtain its own native result.
Three isolated sessions. Only the replay experiment resubmits the exact same packet within its run. These requests connect no customer data writer or settlement service.
SELECTED EXPERIMENT
No captured result yet.
Select a completed card to restore its measured windows and downloads without sending another request.
Choose and run an individual native request
SIX LAB OBSERVATION BOUNDARIES · NESTED RUNTIME CALLSWaiting for a request
Select a component or use the component picker below.
COMPONENT / CALLER CONTRACT
Caller contracts and source identity
Session & sourceGlapRootBinding
The Lab acquires its private slot, creates random session material, obtains the current source binding and opens the local native journal. Measured together at the Lab caller boundary.
Packet encoderRoot-bound request v2
The real action, verb, session, nonce and source binding are encoded into the native request packet. The displayed byte count comes from this buffer.
Native runtimeGlapNativeBackend<16,16>
The existing handle_root_bound_v2 caller receives the actual packet and PublicReadOnly profile. Admission, authentication, catalog lookup and dispatch are inside this measured call; their individual durations are not observed.
Same-packet replaySame runtime + same bytes
Only the replay scenario calls the same native runtime a second time with the exact same packet. Other scenarios display Not called, never a fabricated zero-duration invocation.
Response verifierGlapNativePacketOutcome
The Lab checks the expected native class/error and zero effects, then encodes, authenticates, decodes and compares the actual response fields. It reports the digest and exact response bytes.
Owned cleanupCompleted private slot
After dropping the runtime, the Lab validates and removes only its completed synthetic journal files and slot. Failed namespaces remain preserved.
Admission & catalogBounds · replay · auth · root
The selected runtime caller invokes admit_packet_with_body_validator, resolves the action and checks source/catalog binding. This is source architecture inside one runtime call, with no separate live stage events.
Local journalPrivate run state
The selected run owns a local journal and replay bookkeeping. This is a bounded isolated Lab namespace, not a customer database or a remote validator node.
Executor boundaryGL-T / GL-DDBS: None
This Lab passes None for observed trust, trust admission and GL-DDBS executor arguments. Customer data writers, operator Health and consensus-network telemetry are not connected by these requests.
Selected source SHA-256: 12d5f0dde2c4ea2454fba7c9d30a70b6857986cb888ea9825926ac82db101bff. The Lab calls the retained native binary; this graph is not an inventory of every deployment component.
DURABLE LOCAL CONSENSUS CALLERS · SOURCE PROJECTIONLive C3 events not connected
Select a component or use the component picker below.
COMPONENT / CALLER CONTRACT
Caller contracts and source identity
ProposalNqfcC3Proposal
issue_nqfc_c3_proposal binds the chain, epoch, height, round, registry, checkpoint and value. The registry and signed checkpoint dependencies are separate inputs.
Lock voteNqfcC3PhaseVote
sign_nqfc_c3_lock_vote_durable verifies proposal/state/registry, derives the Lock context and delegates to issue_phase_vote_durable. A signed slot is persisted before returning the vote.
LockQCPhase certificate + witness
assemble_nqfc_c3_lock_certificate aggregates verified votes. The exact certificate-body digest is distinct from the certified context digest.
Commit voteDurable matching lock
sign_nqfc_c3_commit_vote_durable verifies the supplied LockQC, carries its exact witness, records the local lock and persists the signing transition before returning.
Registry & cryptoWeights · signatures
Registry binding, phase vote verification and checkpoint signing/verification are actual dependencies of these callers. Drawn nodes do not denote a real validator population.
Persist transitionNode state publication
issue_phase_vote_durable and local apply both call persist_state_transition. Local memory is replaced after the durable operation returns successfully.
Local applyLock check → persist → memory
apply_nqfc_c3_commit_certificate_durable verifies proposal and both certificates, requires the matching durable local lock, prepares the next height, persists it, then updates memory and clears the completed lock.
CommitQCExact LockQC witness binding
assemble_nqfc_c3_commit_certificate verifies the supplied LockQC and builds the Commit context with that exact justification witness. Local application remains a separate operation.
Timeout pathSeparate round transition
Timeout voting, certificate assembly and application have their own checked path. A successful diagram traversal does not establish distributed safety, liveness or recovery.
Selected source SHA-256: bec98a1073f5dc1344bc9b72aeaacea3191da8a02f120217ea54219e98f0a5c2. No validator network, operator state or live C3 event stream is connected to this view.
ACTUAL NATIVE OUTCOME
Ready for a native request.
Choose a scenario. Its result and measured call windows appear here.
Awaiting a real request
SERVER CALL WINDOWNot observed
NATIVE PACKETNot observed
NATIVE RESPONSENot observed
RUNTIME INVOCATIONSNot observed
Measurements arrive after completion, not as a streamed inner-event feed.
MEASURED WINDOW / SIX BOUNDARIESLengths follow captured durations
Individual server observation · excludes browser/network delivery · select a window to inspect its caller.
Session & sourceNot observed
Packet encodingNot observed
First runtime callNot observed
Repeated runtime callNot observed
Response verificationNot observed
Owned cleanupNot observed
Session and observation identity
Session
Not observed
Server time
Not observed
Native response SHA-256
Not observed
Observation record SHA-256
Not observed
GLLOBS01/v1 records six measured Lab wrapper boundaries and binds the native response/session. Its exact bytes are downloadable from the native result. These are individual observations; benchmark comparisons stay in the study tab.
Study the internal source path separately
This source explanation is separate from the measured Lab call boundaries above.
Each POST creates an isolated session. The replay case accepts the first request and then resubmits that exact packet within the same run. Comparison panes do not share a session.
The worker authenticates and decodes its native response before displaying it. Open the execution details to inspect its fields and download its exact bytes. Session keys remain private; the download alone is not independent server attestation.
No customer database, payment, signing/broadcast or external settlement is connected. Busy and unavailable outcomes are explicit errors, not successful executions.
Context, results & evidence
CURRENT WORKSPACE
Component details
Run an admitted GLAP scenario to capture a native result.