Adludo documentation · Development pilot

Adludo evidence register

Development pilot documentation. Reviewed 5 November 2026. Specified means a requirement exists; implemented means inspected source supports it; tested means a recorded check supports its scope. No independent assessment is established. This revision ran 29 focused local receiver tests and eight delivery-function cases. Earlier SDK and phone evidence retains its original date. Reader-facing evidence links below resolve within this documentation. Detailed source IDs remain in the restricted inventory and structured register.

E01 Active compressed upload recovery

Status: Tested

Build or version: Build 43 / 0.33.0-compressed-recovery

Test date: 4 November 2026 Europe/Madrid (source updated 3 November 23:29 UTC)

Environment: Samsung SM-N975F, Android 12; Adludo pilot

Method: Three fresh recordings; background, secure screen lock and restart after acknowledged batches

Result: 41 / 52 / 74 pending at interruption; zero after each recovery; one matching server result each; 212 original and 43 compressed-copy files unchanged

Limits: Not interruption within a single HTTP request, prolonged suspension, other devices or hardware power loss. Server checks recorded arithmetic.

Verification: Internal project evidence; no independent assessment established

Evidence: Method, result and limitations.

E02 Pending queue lifecycle recovery

Status: Tested

Build or version: Build 43

Test date: 3 November 2026 checkpoint

Environment: Samsung SM-N975F, Android 12

Method: Sequential background, lock and restart with pending compressed-copy upload

Result: Qualification records PASS; subsequent recovery completed

Limits: Separate scenario from active uploads; does not establish all Android lifecycle behavior.

Verification: Internal project evidence; no independent assessment established

Evidence: Method, result and limitations.

E03 SDK release checks

Status: Tested

Build or version: Runtime eb5d488cd2cf64f7f90a5e5971352129deb4fff1; metadata aaa14c6a75d5978ee5d6809c783da9c3b814198e

Test date: 4 November 2026 Europe/Madrid; suite ended 3 November 23:43 UTC

Environment: Local clean source export, .NET 8; Unity 6000.6.4f1 reference compilation

Method: Full SDK suite and isolated HTTP checks; release packaging checks

Result: 175 tests passed; HTTP authorization, recomputation, hold, retry, conflict, malformed body and restart passed; release manifest reports compile/build/APK/provenance checks passed

Limits: A test count is not system certification. Compilation does not establish runtime compatibility for all Unity versions. Original runtime export requires metadata patch for corrected fixtures and fingerprint.

Verification: Internal project evidence; no independent assessment established

Evidence: Method, result and limitations.

E04 Native visible and qualified time

Status: Implemented; focused test identified in recorded release suite

Build or version: Inspected SDK source; build 43 release context

Test date: No individual test date mapped

Environment: Native local recorder and exposure tracker

Method: Source inspection plus release suite reference

Result: Visible intervals require known, ready, active endpoints with positive visibility and valid sampling gap. Qualification applies policy thresholds and continuity separately; focused test “interrupted fragments never combine into an impression” asserts 1,000,000 qualified microseconds and no impression candidate after two interrupted half-second fragments.

Limits: Sampling cannot prove visibility between observations. No pixel authenticity or attention proof. Per-policy thresholds must accompany reports.

Verification: Internal project evidence; no independent assessment established

Evidence: Method, result and limitations.

E05 Supported capture and upload methods

Status: Implemented; focused local verification passed

Build or version: SDK 0.4.0-preview.2; common and pilot indexed upload contracts

Test date: 5 November 2026 Europe/Madrid

Environment: Local macOS, Node v24.22.1, temporary isolated fixtures; receiver c721c37eb83b011678b174edfb11221c32c9ac19

Method: Pinned source review and named focused checks; 29-test suite passed, with separate delivery-control probe

Result: Upload contract lists owned-opaque-overlay-v1 and owned-depth-quad-conservative-v1; shape methods v1–v4 remain local only; pilot adapter selects adludo-native-upload-pilot-v1 with separately bounded indexed finalization; see compatibility reference. Sender and receiver limits differ: pilot SDK archive cap 64 MiB, receiver 128 MiB; effective bounds use the stricter side.

Limits: Not arbitrary 3D/GPU visibility or general engine support. QuickStart and README contain older scope statements; do not treat their milestone lists as latest release status. Actual SDK receiver fixtures are from preview.1 commit 30532e3, not newly generated build 43 fixtures; full maximum-size qualification not rerun.

Verification: Internal project evidence; no independent assessment established

Evidence: Method, result and limitations.

E06 Opt-in upload, receipts and retry

Status: Implemented; focused local verification passed

Build or version: Common and pilot indexed contracts, envelope schema 1

Test date: 5 November 2026 Europe/Madrid

Environment: Local macOS, Node v24.22.1, temporary isolated fixtures; receiver c721c37eb83b011678b174edfb11221c32c9ac19

Method: Pinned source review and named focused checks; 29-test suite passed, with separate delivery-control probe

Result: Original individual queue messages and receipt hashes remain stable through optional bundling and gzip. Compressed-copy recovery checks an existing prepared queue; it neither reconstructs a missing queue nor changes pending bytes. Fresh authorization is required after restart; an explicitly supplied unexpired grant is also supported.

Limits: Focused checks establish local gzip/bundle identity, malformed-input rejection and scoped recovery after withdrawal. No live endpoint test or new phone run. Compressed-copy phone recovery remains E01.

Verification: Internal project evidence; no independent assessment established

Evidence: Method, result and limitations.

E07 Ad-only collection and local default

Storage and deletion by copy. Source review covers build 43 SDK recording and recovery, receiver finalization and the pilot backup worker. A fixed retention schedule and a completed erasure across all copies remain unverified.

Status: Specified project requirement

Build or version: Current SDK project rules

Test date: Not a privacy audit

Environment: SDK, integration, schema and reports

Method: Standing rule and integration documentation review

Result: Only ad-measurement context; no general gameplay telemetry; no ad-free tracking or installation-ID creation; normal capture local, authorized QA upload separate

Limits: Requirement is not proof of comprehensive compliance. Retention, deletion, consent/age handling and production operational policy need separate review.

Verification: Internal project evidence; no independent assessment established

Evidence: Method, result and limitations.

E08 Browser duration meanings

Status: Implemented; focused local verification passed

Build or version: visibilityMetricsVersion 2

Test date: 5 November 2026 Europe/Madrid

Environment: Local macOS, Node v24.22.1, temporary isolated fixtures; receiver c721c37eb83b011678b174edfb11221c32c9ac19

Method: Pinned source review and named focused checks; 29-test suite passed, with separate delivery-control probe

Result: observedVisibleMs is known foreground non-idle positive visibility across adjacent endpoints with gaps at most 250 ms; qualifiedVisibleMs equals legacy visibleMs; durationMs is longest qualified streak Focused version 2 durations test passes; inspected current browser policy uses 50% visibility, 2% coverage, one second and a 250 ms gap.

Limits: Native microseconds and browser milliseconds are distinct. Older observed duration is unknown, not zero. Historical README thresholds and gaps must not override the versioned metric specification.

Verification: Internal project evidence; no independent assessment established

Evidence: Method, result and limitations.

E09 Creative review and placement controls

Status: Implemented; focused local verification passed

Build or version: Marketplace prototype, historical validation snapshots

Test date: 5 November 2026 Europe/Madrid

Environment: Local macOS, Node v24.22.1, temporary isolated fixtures; receiver c721c37eb83b011678b174edfb11221c32c9ac19

Method: Pinned source review and named focused checks; 29-test suite passed, with separate delivery-control probe

Result: Seven approval-race checks, withdrawal API/history/reactivation rejection, review reopening and recovery-after-withdrawal passed. Separate delivery-function probe passes baseline and seven blocked cases: pause, placement/game disable, category, missing either approval, changed placement revision.

Limits: Local fixtures only; not current hosted-state verification or real owner approval. Category/disable supplemental checks exercise the shared eligibility function, not an end-to-end live campaign.

Verification: Internal project evidence; no independent assessment established

Evidence: Method, result and limitations.

E10 Measurement and financial outcomes remain separate

Status: Specified

Build or version: Native QA and marketplace prototype

Test date: Not a new test

Environment: SDK and marketplace

Method: Project rule and prototype boundary review

Result: SDK records measurements; marketplace money is simulated; recovery qualification grants no billing eligibility

Limits: Do not present qualification, a stored receipt, replay verification or simulated ledger as a payable outcome.

Verification: Internal project evidence; no independent assessment established

Evidence: Method, result and limitations.

E11 Historical audit findings

Status: Historical remediation completion record located; scoped controls rechecked

Build or version: 28 October audit; 29 October plan

Test date: 28–29 October 2026

Environment: Preserved audit snapshots

Method: Completion record maps 16 findings to commits and regression files; focused current checks for approval concurrency, upload compatibility and visibility semantics

Result: Final verification records all 16 remediation findings with results, 446 marketplace checks and 144 SDK checks at the historical integration revisions. Internal audit map preserves each finding and its regression reference.

Limits: Historical completion is not a new audit of all current code. Remaining operational/device limits persist; no independent certification or current live deployment assertion.

Verification: Internal project evidence; no independent assessment established

Evidence: Method, result and limitations.