OPENHOOPS / 2K14Final team archive · November 9, 2026Reference libraryChecksums

Dated source reference. Current release status: handoff state. Full original relative paths are in the developer archive.

Presentation framework acceptance

November 9, 15:07 UTC checkpoint. RG passes all 149 Windows groups over 817 frozen inputs, ten freshly built native profiles and all 666 API tests. It reruns every runtime group; 37 fixtures rebuild and 88 reuse exact verified compiler outputs. Its identical DLL passes original EG and college EH for 30.098/45.046 seconds after verified clock movement. Both observe the original shared-quad CRT allocation, capture the full 48,672-byte index span with zero unavailable DrawUP inputs, and pass mixed-target/direct-parameter joins. Selected CRT-origin tracking remains healthy with zero faults; no native free of that process-lifetime table is observed. Both restore 99 files and release custody. Their complete 4,074-file/184,807,589-byte and 3,354-file/191,069,406-byte archives and rollback members are hash-verified/fullfsynced externally. See analysis/presentation-framework75/takeover/rg-eg-eh-closeout.json. This accepts bounded copied input and reference evidence. Complete callback/resource/source/free/context closure, semantic owners/cues, full visual/lifecycle acceptance and final author handoff remain open. Packet mask zero; full framework unfinished. No game or custody is active.

November 9, 14:10 UTC working checkpoint. RB passes 143 Windows groups over 802 exact inputs, ten fresh native profiles and all 666 API regressions; 53 fixtures rebuild and 68 reuse exact binaries, with every runtime group rerun. Original EF runs 107.308 live seconds and passes mixed-target and direct-parameter capture. The mandatory DrawUP gate still fails: its 48,672-byte index span has no unique tracked heap/image token. A bounded mapping query grants no lifetime authority. All 99 files restore and custody releases at 13:57:38 UTC; its complete 3,352-file/220,890,720-byte archive and rollback are verified/fullfsynced, and temporary payloads retire254. No unchanged retry or college run follows. Static tracing identifies the shared quad CPU table, its original MSVCR100 realloc call and separate GPU-buffer creation. Isolated255 passes 2,916 original-x86 factory cases and local sanitizer checks; CRT generation/withdrawal validation is in progress and remains unpromoted. QS/DY/DZ remains the last paired scoped native baseline. Full ownership, effects/resource/source/free/context closure, cues, visual/lifecycle acceptance and final author handoff remain unfinished; packet mask zero. No game or GPU custody is active.

November 9, 13:41 UTC working checkpoint. QZ passes all 143 Windows groups over 802 exact inputs and 666 API regressions; ten native profiles and 96 fixtures build fresh, with 25 exact fixture reuses and every runtime group rerun. Original ED runs 101.877 live seconds and passes the mixed general-target and direct-parameter match gates. One match cohort accounts for all 28 targets, including 17 independent surfaces, zero unavailable targets and 11 explicit nulls; the surface bank retains/releases 54 references and closes empty, healthy and fault-free. ED fails the mandatory DrawUP gate because its indexed request remains unavailable; no unchanged retry or college run is started. All 99 files restore and custody releases at 13:29:55 UTC. Its 3,119-file/217,659,077-byte archive and rollback are verified/fullfsynced; temporary payloads retire 249. An isolated timing correction passes 9,647 input, 12,972 batch and 60,446 production-body Windows checks: exact validation remains the default, explicitly declared consumer indices validate after strict original-token recopy. Bounded preparation diagnostics distinguish shape/code/storage/index refusals; the native cause remains unconfirmed. This eight-file correction is promoted and RA startup checks are running before a fresh RB full gate. The separate buffer-factory prototype passes 366,355 x86 checks/82,944 scalar cases without resource-ownership acceptance. QS/DY/DZ remains the last paired scoped native baseline. Full effects, source/free/context closure, semantic owners/cues, visual/lifecycle acceptance and author handoff remain unfinished; packet mask zero. No game or GPU custody is active.

November 9, 13:14 UTC working checkpoint. QY passes 27 startup groups over 215 inputs. QZ is running 143 Windows groups over 802 frozen inputs with the general target profile enabled; no QZ game run has started. The promoted typed 2D/cube/direct-surface path passes 43,335 controlled production-body checks through publication, capture, surface retention, copied predictions, cancellation and retirement. Forty-eight receipt checks require exact metadata/physical/effect copy-retire joins and a mixed match-bound cohort with zero unavailable targets. The older default-only physical gate is superseded only when that stronger general gate and the same healthy closed surface bank pass. Eight oversized target metadata cases refuse before reads in the 28,121-check sanitizer run. QX remains the latest passed full CPU gate (137 groups, 666 API regressions); QS/DY/DZ remains the paired scoped native baseline. Whole effects/resource/source/free/context closure, semantic owners/cues, full visual/lifecycle acceptance and author handoff remain open. Packet mask zero; full framework unfinished. No game or custody is active.

November 9, 12:57 UTC working checkpoint. QX passes all 137 Windows groups over 783 frozen inputs, ten fresh native profiles and 666 API regressions. The QW fixed-window fixture failure is corrected by reserving its complete two-phase private windows before portable heap work; an intentional occupied-window probe exits before any original instruction. Forty-four receipt tests pass. Isolated texture-target inputs pass 15,111 Windows/sanitizer checks; typed 2D/cube/direct surface acquisition passes 19,315 controlled-COM checks; the general target batch is still under verification. None of this isolated target work is game-installed. QU EA/EC native refusals remain preserved; QS/DY/DZ remains the last paired scoped baseline. Complete targets, effects/resource/source/free/context closure, semantic owners/cues, full visual/lifecycle acceptance and author handoff remain open. Packet mask zero; full framework unfinished. No game or custody is active.

November 9, 12:33 UTC working checkpoint. QW is building 137 Windows groups over 782 frozen inputs after QV passes all 27 startup groups and 44 receipt-parser tests pass. The promoted DrawUP path has 9,216 original VM calls, 9,170 input checks, 11,384 batch checks, 34,457 mixed-plan checks, 7,076 candidate-window checks and 57,800 production-body checks. It copies original heap/image spans, validates vertex/index extents and rechecks the complete cohort. Direct parameter preparation now uses the existing candidate contract; consumption retains strict original-token and byte checks. QU passed 133 CPU groups, but original EA and unchanged retry EC failed required surface coverage after 102.777/110.767 live seconds; EA alone passed the new parameter match gate. Both restore 99 files and release, with full archives and raw retirement 235/236. No third unchanged retry or college run was started. QS/DY/DZ remains the last paired scoped baseline. General texture-backed targets, complete effects/resource/source/free/context closure, semantic owners/cues, full visual/lifecycle acceptance and author handoff remain open. Packet mask zero; full framework unfinished. No game or custody is active.

November 9, 11:50 UTC checkpoint. QS passes 131 Windows groups, 762 frozen inputs and 666 API tests. All ten native profiles and 33 fixtures compile fresh; 76 fixtures reuse verified exact source graphs, commands and artifact hashes, while every runtime group reruns. The same DLL passes original DY and college DZ for 30.142 and 30.127 live seconds. Device requests and consumer wrapper serials have positive match-bound copy/retire pairs with zero unavailable request inputs. Serial observations preserve the original allocation/source tokens and whole-wrapper rechecks; the two producer write prefixes are qualified separately and do not establish full producer lifetime. Both runs restore 99 files and release custody. Full 3,135-file/183,807,005-byte and 4,971-file/187,365,017-byte archives and rollback verify with fullfsync; temporary payloads retire 231/232. DS, DU and college DX failures remain preserved. Effect conversion stops at direct parameter opcode 0x0a; isolated original-VM, mixed-plan and consumer-batch work is not in QS. Full effects, resource/source/free/context closure, semantic owners/cues, visual/lifecycle coverage and author handoff remain open. Packet mask zero; full framework unfinished. No game or custody is active. See qs-dy-dz-closeout.json.

November 9, 09:54 UTC checkpoint. QK passes126 Windows groups/747 frozen inputs/666 API tests (source 8d8e25b0046ede242062451f69e5ec84d258b1304e0b2973e20d8285d0d4af2f, DLL dcd9fafd09d5351c1f956f6a0d536a1e9d3747c8dd211f5850a64ac470d4a95b). All ten native profiles and27 fixtures compile fresh;77 fixture binaries reuse only exact local include graphs, commands and artifact hashes, and every runtime group reruns. Identical original DQ/college DR pass54.244/70.983 live seconds: each has three bound/copied/retired environment sets; eleven/ten clear requests have zero unavailable inputs; four/six independent surface references retain/release with healthy empty closed banks. Both restore99 files and release. Complete2,985-file/197,531,719-byte and3,785-file/204,330,795-byte archives/rollback verify and fullfsync; temporary copies retire211/212. QG's Windows macro failure, DM's tag-byte refusal and DO's sole changed match-surface sample remain preserved. The correction proves all256 tag values/both bit-seven variants in3,264 original Clear cases; only observation bounds grow175→235/190→250 seconds inside the unchanged six-minute lease, with30 required live seconds and all gates retained. Conversion now stops at opcode0x23. Isolated twelve-method request/input/batch work passes3,456 original calls and1,213,376 Windows/sanitizer batch checks; it is not in QK. Complete driver effects, source/free/context closure, semantic owners/cues, full visual/lifecycle acceptance and author handoff remain open. Packet mask zero; full framework unfinished. No game or custody is active.

November 9, 08:16 UTC checkpoint. PZ passes121 Windows groups/666 API tests over728 exact inputs (source 679c4ffcf667d30ecde120306d0ff5cf70db197ad3fa9b4e12848002ea1a1066, DLL 89c79f8da24a69becdc564993c6b5070e9ff59ebf8c2a9259f9da45dfeeb20cc). Ten DLL profiles and96 fixture targets are reused only after exact input/command/hash verification; three corrected fixed-address fixtures rebuild and every runtime group reruns. PW's reserved-window refusal is retained. Identical PZ passes original DK/college DL after30.153/30.144 live seconds: nine/thirteen shadow sets bind and retire, including three/six match sets; sixteen/twenty-two independent surface references balance with healthy empty closed banks. Both restore99 files and release. Complete3,025-file/183,267,456-byte and5,065-file/196,903,663-byte archives/rollback verify and fullfsync; temporary payloads retire204/205. Effect conversion progresses to unsupported opcode0x13. Separate isolated work passes240 original no-read paths/96 required reads,84 initial-plan checks and1,152 mixed original target/root callbacks using flag-only target descriptors. Its production integration is CPU-testing separately and is not in PZ. Complete renderer/effects/source/free/context closure, semantic owners/cues, visual/lifecycle acceptance and final author delivery remain open. Packet mask zero; full framework unfinished. No owned game or custody is active.

November 9, 07:09 UTC checkpoint. PP2 passes119 Windows groups/722 exact inputs/ten native profiles and666 API tests. The receipt explicitly combines118 unchanged passed groups with a fresh666-test API rerun after storage recovery; the initial storage-preflight failure is preserved. Identical DLL 129cc2b3941df715d66653c9c0f020b2bc44a4d3f95a6ebedacab1b32da8a8d0 passes original DI/college DJ physical default-surface ownership after33.363/30.039 live seconds: six/seven sets, three/five match sets, twelve/fourteen retained and released references, zero bank faults and empty closed banks. Both restore99 files and release custody. Full3,116-file/180,135,778-byte and3,256-file/181,406,548-byte archives and rollback verify/fullfsync; temporary copies retire196/197. This proves references for the observed color/depth subset. Publication-exact state copies remain false; complete target effects and drawing authority remain unavailable. Separate PO passes6,686 Windows and sanitizer shadow/cache checks plus686 context regressions; production integration is next. Packet mask zero; the full framework remains unfinished. No game or custody is active.

November 9, 06:24 UTC checkpoint. PH passes117 Windows groups/ten qualified native profiles/666 API tests over711 exact inputs (source c61557f0203e25b4e864418d7cde6462a12caf3de2e91f552c7715b19474839a, DLL 7504c5d2f4eaa53ecf4cbf3a777ad90630ab6792b41b0ce992af82b5db68bad5). The identical DLL passes original DG/college DH mandatory context/default-target metadata gates after30.090/30.174 live seconds. Three/four cohorts, each with one color and one depth target, copy/retire; each profile has one positive match cohort and zero unavailable/unbound targets or joint faults/live frames. Both restore99/release. Complete3,827-file/186,865,061-byte and4,420-file/191,597,428-byte archives/rollback verify/fullfsync; duplicates retire190/191. This is metadata, not physical target lifetime or effect/drawing authority. Working PK151 batch checks, PI3 103 production/reset checks, five physical-receipt tests and PM27 startup groups pass. New PN119-group build over722 frozen inputs wires typed-getter references into renderer-thread close/drain/reset, with early reentry refusal. Its new native game proof is pending. Complete renderer/effects/source/free/context closure, semantic owners/cues, visual/lifecycle acceptance and final author delivery remain open. Packet mask zero; full framework unfinished. No game or custody is active.

November 9, 05:38 UTC checkpoint. The frozen OO original DE/college DF baseline remains preserved; working C++ now differs and is not yet game-qualified. Separate OR4 passes690 controlled-driver surface-ownership checks; OS2 passes85 actual D3D9 HAL checks with six typed getter acquisitions, twelve balanced retained/released references and one Reset/resize over two generations. Its10-file run is externally verified/fullfsynced,99 game files unchanged and custody released. OY passes858 Windows metadata checks, OZ578 batch checks and PA20,855 consumer-core checks. PB passes17,630 production checks for same-window target capture/retirement and pointer/flag redirection refusal. Candidate publication never grants quiescence; strict original-span consumer checks remain. Current startup verification and the expanded117-group combined gate are pending. Native target-input proof, physical surface integration, complete renderer/effects/source/free/context closure, semantic owners/cues, visual/lifecycle acceptance and final author delivery remain open. Packet mask zero; the full framework is unfinished. No game or custody is active.

November 9, 05:15 UTC checkpoint. OO passes 113 Windows groups/ten native DLL profiles/666 API tests over 698 exact inputs (source 23f74efb1d28af5124b7d6bae252c263b51ebb6fe590bf28f5444791f164b6c8, DLL e0883db9e6b7d4900373e7b2dbb32adba54ea7a468c1fb35edd6bd34ee355b34). The identical DLL passes original DE/college DF consumer-input, physical-closure and typed initial image-context gates after30.054/30.039 live seconds. All3/5 copied context sets bind/retire, including1/3 match sets, with equal18/18 branch words and zero faults/live frames. Both restore99 files and release. Complete3,160-file/177,187,050-byte and3,001-file/185,181,571-byte archives and rollback verify/fullfsync; duplicates retire184/185. Publication-exact state copies remainFALSE separately. Effect conversion still refuses opcode0x2f; this is input/context proof, not full effects or drawing. Separate OP2 executes5,775 original target/surface/release cases with65,533 Windows checks, including observed source kinds1/6/10. OQ default-target metadata passes796 Windows/local sanitizer checks plus686 context regressions; it is unintegrated and grants no COM lifetime or authority. Complete renderer/target ownership, remaining effects/source/free/context, semantic owners/cues, visual/lifecycle acceptance and final author delivery stay open. Packet mask zero; full framework unfinished:3 scoped accepted,17 partial,11 unavailable. No game or custody is active.

November 9, 04:42 UTC checkpoint. OJ passes 113 Windows groups/ten profiles/666 API tests over 698 exact inputs and original DC passes its consumer-input/physical-closure gate after 30.154 live seconds, restoring 99 files and releasing custody at 04:36:28 UTC. Four joint sets copy/retire with zero faults/live frames. Typed image facts bind two pre-match sets with equal state words18/18; both match sets lack publication tokens, so the new match-context gate is FALSE. The 22 pin refusals are metadata-window refusals, not accepted context samples. The whole 2,969-file/180,630,281-byte archive and rollback verify/fullfsync; duplicates retire182. College DD was not started. New working metadata_candidate selects only already-qualified current process-image storage tokens before a new publication observation; it mints no token and leaves strict consumer checks/refusal intact. OK passes 686 Windows/local sanitizer checks including active-publisher selection and no retry after failed consumer capture. OL/OM/ON and the next full OO gate are pending. Effect conversion now reaches unsupported opcode0x2f. The separate target predictor passes79 local golden checks only; original-instruction comparison and native integration are pending. Packet mask zero; full framework unfinished. No game or custody is active.

November 9, 04:24 UTC checkpoint. OC passes all 113 Windows groups, ten native DLL profiles and 666 API tests over 696 inputs. The separate image path passes OD 564 checks, OE 20,855 consumer-core checks and OF 16,703 production checks. OI passes all 27 startup groups against the exact source including pin-refusal diagnostics; OH preserved a source-preflight refusal before any Windows stage or game. Fresh OJ is building all 113 groups/ten profiles over 698 inputs (source 0dbcdcee0fa04d7b05d0c10d5a30ea504cf890a19ef5844d9f9102ba60f67f30). It carries exact typed main-image context tokens with the queued cohort; native image-context copy/retire proof is pending original DC/college DD. CPU archive181 verifies 185 outputs/180,101,095 bytes and three Windows restore samples; exact duplicates retired, with 4,030,234,624 bytes free at 04:19:14 UTC. The paired NX DA/DB baseline remains unchanged. Complete effects, remaining source/free/context closure, semantic owners/cues, visual/lifecycle acceptance and final author delivery remain open. Packet mask zero; full framework unfinished. No game or custody is active.

November 9, 04:09 UTC checkpoint. NY passes 21,085 Windows checks for guarded DDB8A0 copied-command conversion. NZ passes 330 typed context-fact checks and 20,854 consumer-core checks; OA passes 15,962 production-body checks including whole-cohort cleanup after bad code or changing context fields. OB passes all 27 startup groups. OC is freshly building 113 Windows groups and ten native DLL profiles over 696 frozen inputs; native proof of the new context facts is pending. The paired NX DA/DB consumer-input baseline remains unchanged. Closed CPU archive180 verifies 474 files/408,989,397 bytes and three Windows restore samples; exact duplicates retired, leaving 4,211,249,152 bytes free at 03:55:35 UTC. Context facts are a five-field initial subset with no complete renderer-state or callback authority. Packet mask zero; full framework unfinished. No game or custody is active.

November 9, 03:51 UTC checkpoint. NX passes all 113 fresh Windows runtime groups and 666 API tests over 694 inputs, rebuilding every fixture while reusing ten exact-input/command/hash-verified native DLL profiles from compiler-passing NW. NW preserves its new timing-fixture protected-window refusal; NX uses the existing bounded retry wrapper and never reclaims mapped memory. The identical NX DLL passes consumer-origin original DA and college DB after 30.069/30.124 live seconds. Each copies/retires four joint material/declaration/root/constant sets with zero faults/live frames; DB includes match sets with 209/234 materials and 45/50 independently retained declaration owners. All 99 files restore and both custody leases release. Complete 3,378-file/179,167,442-byte and 4,305-file/189,021,919-byte archives and rollback verify/fullfsync externally. Publication-exact state copies remain false and independently reported; these runs qualify the separate version-2 consumer payload contract. Every observed effect conversion stops at DDB8A0, which remains forwarded. Initial renderer state, guarded context binding, remaining effects/source/free/context closure, semantic owners/cues, native visual/lifecycle acceptance and final author delivery stay open. NX DA/DB is the paired consumer-input baseline; MU CO/CP remains the historical publication-exact baseline. Packet mask zero; full framework unfinished: 3 scoped accepted, 17 partial, 11 unavailable. Closed CPU-output preservation is underway to restore test storage margin. No game or custody is active.

November 9, 00:07 UTC checkpoint. MG passes all 101 Windows groups over 666 frozen inputs, nine x86 DLL profiles, the AMD64 composer and 666 API tests. Its identical DLL passes original CK and college CL for 30.140/30.073 seconds after copied-clock advance, including both legacy nested-material roles and complete material/effect-input capture and retirement. All three physical-reference banks finish healthy and closed with zero owners or faults. Each run restores all 99 files and releases custody. Their complete 3,008-file/185,244,627-byte and 3,341-file/187,156,648-byte archives and rollback verify/fullfsync externally. The separate complete-batch nesting path still observes role zero only. New MH passes 7,344 Windows/local sanitizer checks for physical snapshot handoff/cancellation; MI passes 7,056 production-body checks; MJ passes all 27 resource/startup groups. MK is building 105 fresh groups and ten DLL profiles to connect this handoff in its own opt-in profile. Its native evidence remains pending. Full framework is unfinished: packet mask zero; 3 scoped accepted, 17 partial, 11 unavailable. Remaining effects/source/free/context closure, semantic owners and cues, native visual/lifecycle acceptance and final author delivery stay open. No game or custody is active.

November 8, 23:39 UTC checkpoint. MA passes 100 Windows groups over 664 inputs, nine x86 DLL profiles, the AMD64 composer and 666 API tests. Its new declaration path retains and fully releases independent references in original CI and college CJ: both close healthy with zero live/pending/quarantined owners or faults. CJ passes the full forwarding gate after 30.022 seconds of copied-clock advance; CI records 47.525 seconds but fails both-role material nesting because its retained static lists are empty. CI/CJ restore all 99 files and release custody. Their complete 2,891-file/192,074,078-byte and 5,342-file/187,419,879-byte archives and rollback verify/fullfsync externally. MC2 passes 9,706 Windows/local sanitizer checks for publication-held declaration owners, MD passes 160 production-body checks and MF3 passes 10,992 material-publication checks, including refusal of an active-writer consumer without COM acquisition. ME2 passes all 25 resource/startup groups over 133 exact windows. MG is freshly building all 101 groups with these changes; no runtime pass is inherited. Whole-packet effects/source/free/context, remaining semantic owners/cues, native visual/lifecycle acceptance and final author delivery stay open. Full framework unfinished: packet mask zero; 3 scoped accepted, 17 partial, 11 unavailable. No game or custody is active.

November 8, 23:16 UTC checkpoint. LZ passes all 100 fresh Windows groups over 664 frozen inputs, nine x86 DLL profiles, the AMD64 composer and 666 API tests. The opt-in declaration profile adds four exact windows to the existing 129; fresh static review finds no overwritten-interior targets and unchanged ten literal bindings. LW3 verifies original entry/return machine state, LX4 passes 155 production-body/closure checks, and LY passes all 25 startup/resource groups with the held debugger barrier. Controlled HAL runs LU/LV2 separately verify declaration references and Reset; both close, preserve all 99 game files and release custody. The final receipt extension is included in fresh MA, now building. New-profile native game acceptance remains pending; LK/CG/CH is the last paired game baseline. See the declaration lifetime and ordered callback contracts. Full framework remains unfinished: packet mask zero; 3 scoped accepted, 17 partial, 11 unavailable. No game or custody is active.

November 8, 22:18 UTC checkpoint. LK passes 89 fresh Windows groups over 637 frozen inputs, eight x86 DLL profiles, the AMD64 composer and 666 API tests. The same DLL passes original CG and college CH for 30.087/30.043 seconds after copied-clock advance, with complete material/effect-input copying and retirement. Both restore all 99 files and release custody; their full 3,060-file/180,291,786-byte and 3,045-file/182,753,703-byte archives and rollback verify/fullfsync externally. Legacy nested/code witnesses pass both profiles; the new complete-batch collector observes role zero only and remains partial. LM separately passes 103,513 Windows checks over 1,344 original root-state callbacks, including three device branches and failed HRESULTs, plus local sanitizers. Newer mixed callback sequencing is under isolated CPU validation. These predictions grant no native callback authority. Whole-packet source/free/context, remaining owners and cues, visual/recovery acceptance and final author delivery remain open. Full framework remains unfinished: packet mask zero; 3 scoped accepted, 17 partial, 11 unavailable. No game or custody is active.

November 8, 21:27 UTC checkpoint. The fresh LH combined gate is running: 88 named Windows groups over 635 frozen inputs, including the original constant setup/remap predictor, strict source/map binding and ordered dependency batch, plus the new heap mutator tests. LD4 passes 504 original-heap baseline/observed scenarios and 111,638 checks; LE passes 362 source-span protocol checks and LF passes 574 normalized event-join checks. LG3 passes all 23 production resource groups, including the private version-3 startup transaction with 129 exact texture-profile windows. Static review verifies all original bytes, disjoint windows, 203 raw near-branch candidates with no overwritten-interior target, and ten entry-only literal bindings. Live proof for this new candidate is pending. KW2/CC/CD remain the last frozen paired native baseline. Current working files include newer allocator changes. See source-memory mutator evidence and the takeover receipts. Full framework remains unfinished: packet mask zero; 3 scoped accepted, 17 partial and 11 unavailable. No game or custody is active.

November 8, 20:39 UTC checkpoint. KW2 passes 83 fresh Windows groups over 619 exact inputs, eight x86 DLL profiles, the AMD64 composer and 666 API tests. Its unchanged DLL passes original CC and college CD complete batch and retained effect-input copying/retirement for 30.119/30.039 seconds after copied-clock advance, including 10/13 nonempty match-bound input sets and batches up to 230/234 materials. Both restore all 99 files and release custody. Their complete 3,705-file/185,076,093-byte and 3,594-file/191,464,765-byte archives, metadata and rollback verify/fullfsync externally. The pure ordered effect models remain separately compared with captured instructions and controlled drivers; native whole-packet effect/source/free/context admission remains open. New constant setup/remap predictions pass 8,459 local sanitizer checks over 6,912 actions; KY2 is testing the original Windows helpers with an authenticated relocated device selector after KY safely refused an occupied image window. Read-only owner sampling records cached roots only, with visual/semantic ownership explicitly unverified. No game is active. Remaining semantic owners, sequences, native visual/recovery acceptance and final author delivery remain open. Full framework unfinished; packet mask zero; 3 scoped accepted, 17 partial, 11 unavailable.

November 8, 18:54 UTC checkpoint. JY passes 78 fresh Windows groups, eight x86 DLL profiles, the AMD64 composer and 666 API tests over 603 exact inputs. Its original BX run prepares three candidate batches but aborts on the legacy heap-only descriptor receipt format; complete native batch evaluation is not accepted. All 99 files restore and custody releases at 18:46:02 UTC. The 2,886-file/154,755,959-byte archive and rollback verify/fullfsync externally; the exception prevented final probe-log/snapshot publication, so retained progress tails are partial diagnostics only. Version 2 descriptor receipts now bind explicit heap/image domain, token, module and range identities. KA passes 10,561 production-body checks, and ten Windows parser tests verify the actual image formatter round trip while retaining rejection of old zero-heap entries. KB passes 10,418 production batch checks with precise capture refusal diagnostics. JZ separately passes 178,872 Windows checks for 768 interleaved cache actions without refreshing native state. KC is rebuilding all 79 groups; no game is active. Complete packet effects, source/free/context, remaining owners and sequences, native visual/recovery acceptance and complete author delivery remain open. Full framework unfinished; packet mask zero; 3 scoped accepted, 17 partial, 11 unavailable.

November 8, 18:37 UTC checkpoint. JT passes 77 fresh Windows groups over 601 inputs, eight x86 DLL profiles, the AMD64 composer and 666 API tests. Original BW still refuses the complete batch after 63.358 seconds of live clock, while nested identity/code checks pass. It restores all 99 files and releases custody at 18:25:20 UTC; its complete 4,913-file/206,483,923-byte archive and rollback verify/fullfsync externally. The detailed trace identifies image-resident material metadata. A separate typed image-storage binding now passes 12,217 Windows and local sanitizer checks for full graph/owner recapture, mapping and byte changes, forged tokens, overlap and stale constructor refusal. Fresh resource regression JX is running; no new game is active. Whole-packet effects, source/free/context, remaining owners and sequences, native visual/recovery acceptance and complete author delivery remain open. Full framework unfinished; packet mask zero; 3 scoped accepted, 17 partial, 11 unavailable.

November 8, 18:00 UTC checkpoint. JO passes all 77 fresh Windows groups, eight x86 DLL profiles, the AMD64 composer and 666 API tests over 601 exact inputs. Original BU failed the complete-batch gate after 65.496 seconds of live clock; nested material/code checks passed, all 99 files restored and custody released at 17:48:59 UTC. Its complete 4,587-file/215,066,212-byte archive and rollback verify/fullfsync externally. The command census now uses one bounded 512 KiB metadata window, with staged dependency locks and unchanged strict consumer revalidation; JM/JN pass current Windows/sanitizer and actual production-body checks. JO original BV is currently running under the restoration supervisor. Ordered cache flush/write models and explicit-root parameter batches have separate native-compared CPU evidence; whole-packet effects remain unqualified. The new portable bundle CLI passes six tests on Mac and Windows and preserves declared resources and attribution. An editable version 5 neutral example is locally packaged and structurally verified; it grants no native capabilities. Source/free/context, remaining owners, sequence lifecycle, native visual/recovery acceptance and complete author delivery remain open. Packet mask zero; full framework unfinished; counts remain 3 scoped accepted, 17 partial, 11 unavailable.

November 8, 17:31 UTC checkpoint. IY passes 74 fresh Windows groups, eight x86 profiles, the AMD64 composer and 666 API tests over 591 inputs. Original BT still fails the complete-batch gate after 56.084 seconds of live clock, while its nested-code gates pass; all 99 files restore and custody releases at 17:24:31 UTC. Its complete 3,690-file/190,184,903-byte archive and rollback verify/fullfsync externally. Offline hash-checked control traces find all 430/602/685 material candidates on the explicit path, with no raw-only candidates in these samples. JB passes 12,151 Windows/sanitizer checks for staged dependency preparation and strict consumer re-enumeration, including added/removed/changed commands and mutation refusal. JC passes 10,418 actual production-body checks. This does not admit callbacks: the unchanged consumer fence, exact allocation generations, copied bytes and physical owners remain required. JD is rebuilding all 74 groups; no game is active. Ordered effects, scoped source/free/context, semantic owners, sequences, native visual/lifecycle coverage and author delivery remain open. Full framework unfinished; packet mask zero; 3 scoped accepted, 17 partial, 11 unavailable.

November 8, 16:53 UTC checkpoint. IU passes 72 fresh Windows groups, eight fresh x86 DLL profiles, the AMD64 composer and 666 API tests over 584 exact inputs. The same DLL passes original BQ and college BR for 30.050/30.057 seconds after copied-clock advance. Their dynamic/static entry-retirement counts are 16/4 and 14/4; all have authenticated complete native material code and physical parent identities. Original/college retain and release 1,040/994 shader and 711/712 texture references, ending with zero owners or faults. Both restore all 99 files and release custody. Their complete 3,204-file/182,757,343-byte and 3,080-file/189,159,684-byte archives, metadata and rollback verify/fullfsync externally. This closes paired nesting/code observation, not callback or drawing admission. The IV2 production complete-batch observer passes 10,417 Windows checks; its separate first fixture failed main-function embedding before runtime, and that failed freeze remains preserved. Current source now wires bounded complete dependency batches through publication, consumption, retirement, cancellation and Reset cleanup. Six new batch-receipt parser groups pass. IW is rebuilding 73 fresh groups over 588 exact inputs; complete native batch evidence and ordered material effects remain pending. Source/free/context, remaining semantic owners, sequence lifecycle, active-game Reset/resize and complete author delivery remain open. Packet mask remains zero; 3 scoped accepted, 17 partial, 11 unavailable. The full framework is unfinished and no game is active.

November 8, 16:35 UTC checkpoint. IO original BP fails the unchanged two-role nesting/code gate: ten dynamic entry/retire pairs have authenticated full material code, but static execution remains unobserved. It advances the copied game clock for 60.618 seconds, restores all 99 files and releases custody at 16:29:19 UTC. Its complete 4,348-file/210,902,269-byte archive and rollback verify/fullfsync externally. The wider read-only census succeeds within 0.100/0.176/0.201-second pauses: 348/776/1,166 headers contain 1/2/1 rare 76-byte static lists, typically deep in the packet. The captured 76-byte static and 408-byte dynamic lists are now literal production-decoder regressions; 11,749 local sanitizer checks pass. IT passes 10,562 Windows production checks for a tail-first static search using 32-byte command caching, a 32 KiB window and 64 KiB read bound; ordinary sampling retains its 2 KiB window/8 KiB read bound. IS passes 11,371 Windows and local sanitizer consumer-entry guards; ordered whole-packet cache effects remain explicitly unqualified. IU is rebuilding 72 named groups over 584 exact inputs with all four new components freshly compiled. No new callback/drawing authority is enabled. Native static/pair coverage, material effect admission, complete scoped source/free/context, semantic/sequence/lifecycle and author delivery remain open; the full framework is unfinished. No game is active.

November 8, 16:26 UTC checkpoint. IO passes all 68 fresh Windows groups, eight fresh x86 DLL profiles, the AMD64 composer and 666 API regressions over 571 exact inputs. Its code read view authenticates all three existing interpreter/drawing hook targets; IL/IM pass 13,320/241 Windows and sanitizer checks, including 111 altered native-code samples. IP/IQ/IR separately pass material write/input/control isolation, complete bounded candidate census and transactional publication/capture batches (10,446/902/11,197 Windows checks plus local sanitizers); those new components are not yet integrated into a native candidate and grant no callback authority. The 16-test read-only capture now surveys up to 2,048 headers before prioritizing 128 detailed samples, retaining its 2 MiB/0.65-second limits. The IO native harness is frozen, but the installer stopped at the idle guard before mutation because arena_bramlage acquired the Dell for its independent Tahoe test. No new custody, game install or game run occurred. Full-framework counts remain 3 scoped accepted, 17 partial and 11 unavailable; packet mask remains zero. Paired native nesting/code coverage, material effects, source/free/context, semantic/sequence and lifecycle/author delivery remain open.

November 8, 16:10 UTC checkpoint. IF passes 66 fresh Windows groups, eight fresh native profiles and 666 API tests. Its original BN run fails the new two-role nesting gate: 24 dynamic entry/retire pairs match, but no static entry is observed; all 37 complete sampled materials have zero static-list bytes. This is incomplete native coverage, not acceptance. All 99 files restore and custody releases; its 2,146-file/183,828,689-byte archive and rollback members verify/fullfsync externally. IG/IH pass authenticated own-patch projection and complete pinned material-code verification (3,990/233 checks); IJ passes the production observer and rotating, static-seeking bounded sampler (10,498 checks). IK is rebuilding all 68 named Windows groups over 571 exact inputs. Paired native nesting/code coverage, material-effect admission, source/free/context, semantic owners, sequence lifecycle, active-game Reset/resize and full author delivery remain open. Accepted packet mask stays zero and the full framework is unfinished. No game is active.

November 8, 15:52 UTC checkpoint. The unchanged IB2 DLL passes original BL and college BM for 30.062/30.091 seconds after copied-clock advance. Original/college retain and release 1,038/1,068 shader and 711/711 texture references with zero final owners or faults. Their 21/27 complete cache snapshot pairs and 18/21 physical material texture-owner pairs match through retirement; both restore all 99 files and release custody. Complete archives (2,198 files/177,823,188 bytes and 2,080 files/181,949,386 bytes), original metadata and rollback members are independently verified/fullfsynced externally. Only startup Reset is observed in the games; IC3 retains its separate three-Reset HAL component scope. ID2 adds the two nested material-VM call identities and passes Windows/sanitizer checks. IE passes its actual production observation body with matching entry/retire pairs and LastError preservation; observations never change refusal or grant packet authority. IF is building the combined 66-group candidate for paired native nesting evidence. Native material-effect admission, complete source/free/context, remaining semantic owners, sequence lifecycle, active-game Reset/resize and complete author delivery remain open. Accepted packet mask is zero and the full framework remains unfinished. No game is active.

November 8, 15:42 UTC checkpoint. IB2 passes 65 fresh Windows groups, eight fresh x86 DLL profiles, the AMD64 composer, 666 API regressions and 560 exact inputs. It integrates guarded resource-reference Reset handling and a denser bounded complete-candidate sampler. HZ4 passes eight actual-host/COM reset cases (791 checks), including held owners, failure/retry, foreign threads, driver reentry, context rebinding and terminal close; ASan/UBSan/TSan bank regressions also pass. IC3 independently passes 228 actual HAL checks: three real windowed Resets at 640x360, 800x450 and 2560x1080, GPU pixel readbacks and shader/default-pool texture closure, with controlled source headers, no game, unchanged 99 baseline files, exact job/task closure and release. IC2 failed with a stack-overflow exit; its job closed and released, and the fixture host was moved to the heap for IC3. HW7 passes 1,152 complete native material executions with 1,050,842 checks including the 1,224 leaf baseline, now covering the three-row initialized cache as well as the nine-handler layout. HY2 original BK failed the new native evidence gate: only menu-bound complete dependency records were observed, despite 47.962 seconds of live play. All 99 files restored, custody released, and its complete 2,159-file/187,479,739-byte archive and rollback are verified/fullfsynced. IB2 original BL is active; native cache admission, nested material execution, semantic/sequence coverage, active-game lifecycle and complete author delivery remain open. Accepted packet mask remains zero; the full framework is unfinished.

November 8, 15:08 UTC checkpoint. The unchanged HV2 DLL passes original BI and college BJ with 30.049/32.516 seconds after copied-clock advance. Both close physical shader/texture banks with zero final owners or faults, restore all 99 baseline files and release custody. College BJ retains/releases 1,008 shader and 713 texture references and records one match-bound material texture-owner copy/retire pair. Complete original/college archives (2,150/2,382 files, 178,725,218/182,187,691 bytes), metadata and rollback members are verified/fullfsynced externally. HW3 then passes the complete native material chain with 768 composed cases and 815,949 checks including 1,224 leaf regressions: native cache helpers, shader ready-query/bindings, VM, parameter dispatch and all nine leaves execute together; only the D3D driver and existing alias relocation are controlled. This is CPU execution evidence, not callback admission. Material cache/table environment, texture contents, complete source/free/context, semantic/sequence and device lifecycle/author delivery remain open. Accepted packet mask stays zero and the complete framework remains unfinished. No game is active.

November 8, 14:57 UTC checkpoint. HV2 passes all 62 fresh runtime groups and 666 API tests over 551 exact inputs. Eight native DLL profiles are reused byte-for-byte after exact source/compiler-command comparison; all CPU fixtures are rebuilt with controlled image placement. The previous HU2 fixture refused its own overlapping ASLR image without executing hooks; the isolated read-only diagnosis and failed freezes are preserved. The current material/texture production body passes 10,430 Windows checks; local ASan/UBSan/TSan owner models pass 10,185. Original BI then passes 30.049 seconds after copied-clock advance, with four bound texture-owner copy/retire pairs, 1,066 shader and 711 texture references fully released, zero final owners/faults, 99-file restoration and custody release. Its complete 2,150-file/178,725,218-byte archive is being verified before the college campaign. This accepts bounded material object-lifetime closure, not texture contents, native material effects, complete source/free/context, semantic/sequence or device lifecycle/author delivery. Packet capabilities remain zero and the full framework remains unfinished.

November 8, 14:28 UTC checkpoint. The unchanged HR DLL passes both original BG2 and college BH for 30.144/30.031 seconds after copied-clock advance. It retains/releases 1,064/1,044 shader and 715/711 texture references with zero final owners, pending/quarantined references or faults. Both joint close transactions acknowledge; both restore all 99 baseline files and release custody. Complete archives (2,398 files/183,084,607 bytes and 2,103 files/178,056,381 bytes), original metadata and rollback members are independently verified/fullfsynced externally. HS adds publication-bound descriptor allocation generations and independent texture owners to the material model: 9,991 Windows checks and local ASan/UBSan/TSan pass, including identical-address reuse, aliases and interrupted reads. That new model is not integrated into the native runtime yet. Texture content/write authority, native material effects, complete source/free/context, semantic/sequence, device Reset/resize and author delivery remain open. Packet capabilities stay zero, the complete framework is unfinished, and no game is active.

November 8, 14:20 UTC checkpoint. HR (60 Windows groups, eight x86 profiles, 666 API tests, 547 exact inputs) passes original BG2 for 30.144 seconds after copied-clock advance. The actual D3D driver retains and releases 1,064 shader and 715 texture references; both banks finish healthy/closed with zero owners, pending/quarantined entries or faults. Eight material captures retire; match-bound texture/source metadata is present. The joint close acknowledgement takes 56 ms. All 99 baseline files restore and custody releases at 14:19:32 UTC. The complete 2,398-file/183,084,607-byte run archive is being transferred and independently verified before the college campaign. This is bounded native object-lifetime and metadata evidence. Texture contents/writes, native material effects, complete source/free/context, semantic/sequence and device Reset/resize/lifecycle/author delivery remain open. The accepted packet mask remains zero and the full framework is unfinished.

November 8, 14:09 UTC checkpoint. HR passes all 60 Windows groups, eight fresh x86 DLL profiles, the AMD64 composer, 666 API tests and 547 exact inputs. The optional texture profile is compiled and bound to its exact DLL/map and 28 counters; HQ4 passes its real production texture-retention join and 121-site held OS debugger transaction. HP2/HR pass all eight combined shader/texture close cases. The original BG attempt stops at storage preflight before custody, staging or game start; the tested native texture DLL has not run in a game. Closed CPU outputs are preserved, restore-tested and retired, retaining the current DLL/map/inputs. Additional already archived capture binaries are undergoing exact identity/content checks before duplicate retirement. HE original BF and paired GY original BC2/college BE2 retain their earlier scopes. Native texture, material effects, complete source/free/context, semantic/sequence and lifecycle/author delivery acceptance remain open. Packet capabilities remain zero; no game is active.

November 8, 13:53 UTC checkpoint. The optional full texture DLL compiles with warnings treated as errors (HO2, 535 exact inputs, 28 resource routes); it is not installed. HL4 passes 196,522 checks across 7,680 captured native texture factory executions, preserving registers, flags, floating-point state, LastError and the original caller return word. HP2 passes all eight combined shader/texture cleanup processes (450 checks), including reentrant COM callbacks and independently held pins; acknowledgement waits for both banks to close. HN2 passes 21 existing Windows resource groups. HQ2 is checking the expanded production texture adapter and its 121-site held OS debugger transaction after completing a previously abbreviated fixture context helper. Native texture ownership, a fresh combined gate and paired original/college acceptance remain pending. HE original BF and paired GY metadata evidence retain their recorded scopes. Packet capabilities remain zero, the complete framework is unfinished, and no game is active.

November 8, 13:29 UTC checkpoint. HE passes 54 Windows groups, seven fresh x86 profiles, 666 API tests and 527 exact inputs. Its original BF campaign passes 30.107 seconds after copied clock advance, retains/releases 1,082 shader references with zero final owners/faults, restores 99 files and releases custody. Its complete 2,075-file/178,156,332-byte archive and rollback members verify/fullfsync externally. Three read-only samples capture the same runtime-decoded 101-byte texture classifier and expected 11-entry handler table. HI executes that complete classifier/factory/create/dimensions/subresource/release chain in 3,840 controlled Windows cases; HJ2 adds the actual COM retention backend/bank and passes 75,435 checks. HM passes 975 typed texture/source/context checks plus 1,330 shader regressions; the same model passes local address/undefined/thread sanitizers. Native texture retention is not installed. The new optional texture return route and epilogue plan are under machine-state verification; current sources have changed since HE and require a fresh combined gate. GY original BC2/college BE2 remains the last accepted paired metadata baseline; HE college acceptance is pending. Packet capabilities remain zero. Complete source/free/context, material effects/nested lists, semantic/sequence owners, full lifecycle and author delivery remain open. No game is active.

November 8, 12:56 UTC checkpoint. GY passes 54 Windows groups, seven fresh x86 profiles, the AMD64 composer, 666 API tests and 524 exact inputs. The same DLL passes original BC2 and college BE2 with 30.145/30.147 seconds after copied clock advance. Match-bound parameter graphs copy and retire actual texture/source metadata; the original has one such graph, college has four. All 1,048/1,062 retained shader references release with zero final owners or faults; both restore 99 files and release custody. College BD2 separately passed metadata but failed its 30-second duration requirement (22.339 seconds); its failure remains preserved. BE2 extends only the harness wall limit to 175 seconds. This accepts bounded metadata read/retirement, not texture COM ownership, native material effects or rendering. The full native getter/request/deferred-drain CPU proof is in progress; its actual state table separates ready4 return from state5 queue draining. Complete source/free/context, nested lists, semantic/sequence owners, lifecycle and author delivery remain open. Packet capabilities remain zero; no game is active.

November 8, 12:33 UTC checkpoint. GW passes54 Windows groups, seven fresh profiles,666 API tests and524 exact inputs, but original BA fails the match-bound native gate. BA advances the copied clock for15.105 seconds, retires one early epoch-zero metadata graph with six textures/six source headers, and records no bound cache/list/parameter snapshot. Its1052 shader references release with zero final owners/faults;99 files restore and custody releases. The full2077-file/171263163-byte archive and rollback members verify/fullfsync externally. This early graph is diagnostic evidence only; GT original AY2/college AZ2 remains the accepted paired native baseline. New capture-refusal traces identify the exact last dependency without extra native reads and pass7799 Windows production-body checks; local ASan/UBSan/TSan metadata checks pass7557. GY is rebuilding the54-group closure. Its next bounded harness requires30 seconds after copied clock advance, preserving all identity/generation/read/retirement guards. Native source/free/context, cache/material/list effects, semantic/sequence closure, full lifecycle and author delivery remain unfinished. Packet capabilities remain zero; no game is active.

November 8, 11:02 UTC checkpoint. GH passes50 combined Windows groups, seven x86 profiles,666 API regressions and508 exact inputs. Its DLL passes original AU and college AV with15.037/15.003 seconds after the copied clock advances:8/3 material snapshots captured and retired,1060/1052 physical shader references released, zero final owners/faults. Cleanup acknowledgements take4/27ms; each restores99 baseline files and releases custody. Complete run archives and rollback members verify/fullfsync externally. This accepts bounded material dependency reads, not callback/rendering or complete source/free/context admission. The next revision pins cache-mask allocations and publication bytes, rejecting stale reuse and changed contents;4100 local sanitizer checks and4230 Windows production-probe checks pass. The mask decoder passes2519 Windows checks; cache helpers/two constant leaves retain528 scenarios/126904 checks. GK is compiling a fresh51-group closure over511 inputs. Mutable cache-table effects, material/list/semantic/sequence closure, full lifecycle and author delivery remain open. Packet capabilities stay zero; no game is active.

November 8, 10:38 UTC checkpoint. FY passes49 Windows groups, seven x86 profiles,666 API regressions and503 exact inputs. Original AS/AT both refuse material acceptance:21/22 samples produce zero pinned/captured dependencies; AT records publication fence/read revocation. They retain/release964/1044 physical shader references with zero final owners/faults, restore99 baseline files and release custody. Both complete run archives and rollback members are independently hash-verified/fullfsynced externally. A new bounded metadata-copy window preserves publication-time identities, refuses active mutations and invalidates reentrant changes;3377 local sanitizer checks and GE3507 Windows production-probe checks pass. GC executes three complete cache helpers plus two native constant leaves in528 scenarios/126904 checks; remaining table leaves stay controlled. GD passes21 resource groups. GF is compiling a fresh50-group closure over508 inputs. Native material/window acceptance, complete source/context/free, material/list/semantic/sequence closure, lifecycle and author delivery remain open. Packet capabilities remain zero; no game is active.

November 8, 09:27 UTC checkpoint. FR passes48 Windows groups, seven x86 profiles, the AMD64 composer,666 API regressions and497 exact inputs. The same DLL passes original AQ and college AR after15.109/15.126 seconds from first copied clock advance. They retain/release978/962 physical shader references, with zero remaining owners/pending/quarantined references and zero ownership faults. Owning-thread close acknowledgements take32/21ms. Both restore99 files and release custody; full run archives, rollback members and metadata are hash-verified/fsynced externally. This accepts bounded retention/cleanup in both profiles, not device reset, complete source/free/context or renderer/callback admission. Material/list/semantic/sequence closure and full lifecycle/author delivery remain unfinished. Packet capabilities stay zero; no game is active.

November 8, 08:17 UTC checkpoint. FO passes48 Windows groups, all seven x86 profiles, the AMD64 composer,666 API regressions and496 exact inputs. The source-storage trace records up to512 constructor generations, their actual x86 thread stack bounds and matched destructor-entry/completed-return withdrawals. FN's21 Windows adapter groups verify a real stack constructor/destructor and duplicate-completion refusal;16 parser checks pass locally and on Windows. This grants no source storage authority. FM original AN independently places the first refused0277F8F8 header in an owned thread's stack below its later ESP, from three bounded38-thread held-event censuses; it passes15.380 moving-clock seconds and restores99 files/releases custody. The unchanged FM image-binding pair remains accepted in original AL/college AM2 with163 withheld kind1 entries each. AO's preflight correctly holds before acquisition for storage;37 exact closed compiler intermediates are hash-preserved/fsynced externally and restore-tested before retirement. AO2 is starting on FO for the complete temporary-lifetime audit. Persistent shader ownership, complete source/free/context coverage, material/semantic/sequence completion and full lifecycle acceptance remain unfinished. Packet capabilities remain zero.

November 8, 08:00 UTC checkpoint. FM passes48 Windows groups, seven x86 profiles, the AMD64 composer,666 API regressions and496 exact inputs. The same DLL passes original AL and college AM2 with15.296/15.420 seconds after copied clock advance,40/41 image-backed constructor bindings and zero ledger/binding/routing faults. Each restores99 files and releases custody. Both retain163 unbound kind1 headers; a read-only mapping query places the first at0277F8F8 in private memory, outside the sampled render-thread stack. No heap or stack ownership is inferred. A new held-event TEB32 bounds reader passes on a real owned x86 child, including its actual stack-local marker, forced selector upper bits,12-byte and time bounds, zero writes and child closure. Original AN is starting on the unchanged FM DLL with this bounded all-thread diagnostic. Complete source/free/context coverage, persistent shader ownership, material/semantic/sequence completion and full lifecycle acceptance remain pending. Packet capabilities remain zero.

November 8, 07:40 UTC checkpoint. FK passes48 Windows groups, seven x86 profiles, the AMD64 composer,666 API regressions and496 exact inputs. Original AK passes15.299 seconds after copied clock advance with zero source ledger, routing or binding faults, then restores99 files and releases custody. Its entry-time diagnostics identify204 unbound constructors:202 kind1, one kind6 and one kind10, with pinned embedded-header caller paths and no observed allocation. The first source addresses are candidates for main-image storage, not accepted allocations. New image-source tokens remain distinct from heap/backing storage; they reject allocator-pool/root overlaps, preserve constructor generations and pass1,262 local ASan/UBSan/TSan checks. FL is checking the real Windows image mapping and production constructor path. Native image-source qualification, complete source/free/context coverage, persistent shader ownership, material/semantic/sequence completion and full lifecycle acceptance remain pending. Packet capabilities stay zero; no game is active.

November 8, 07:26 UTC checkpoint. FI's same-DLL original AI and college AJ2 campaigns pass15.294/15.455 seconds after copied clock advance. They complete24,467/24,223 source constructor bindings, each with203 unbound entries, zero stale completions and zero binding faults. The allocator/module ledgers remain healthy; all99 files restore and both custody leases release. The repeated203-entry gap is under investigation. New entry-time refusal diagnostics preserve the existing allocation/ancestry checks and pass1,120 local ASan/UBSan/TSan checks; FJ Windows adapter verification is next. These observations do not qualify complete native source/free/context coverage or drawing admission. Persistent shader ownership, material/semantic/sequence completion and full lifecycle acceptance remain pending. Packet capabilities stay zero; no game is active.

November 8, 07:18 UTC checkpoint. FI passes48 combined Windows groups, seven x86 profiles, the AMD64 composer,666 API regressions and496 exact inputs. Original AI passes15.294 seconds after copied clock advance with24,467 completed constructor/allocation-generation bindings,203 unqualified entries withheld,11,213 destructor withdrawals and zero stale completions or binding faults. The allocator/module ledger stays healthy. All99 files restore and custody releases at07:17:20UTC. This is observed constructor/allocation identity, not complete native source read/free coverage or drawing admission. The same-DLL college constructor campaign is next; the FG allocator/module pair remains accepted in both profiles. The203 unbound cases require further source-storage work. Persistent shader ownership, complete source/context/free coverage, material/semantic/sequence completion and full lifecycle acceptance remain pending. Packet capabilities stay zero; no game is active.

November 8, 07:07 UTC active checkpoint. FH passes21 Windows adapter groups including an observed-allocation-backed real graphics constructor, an owned48-byte source snapshot, and destructor withdrawal that refuses a later read before the backing allocation is freed. The optional source probe records entry-bound constructor tickets; it withholds unrecognized allocations and rejects stale completion without issuing packet authority. The binding bank retains its fixed capacity through runtime-count record allocation. The expanded portable binding model passes1,091 ASan/UBSan/TSan checks, including equal-size parent/child ancestry, delegated holes, partial boundaries and address reuse. FI is compiling seven profiles and rerunning48 groups over496 inputs. FG original AG and college AH2 remain the accepted allocator/module observation pair; both restore99 files and release custody. Actual native constructor-binding acceptance, complete source/context/free coverage, persistent shader ownership, material/semantic/sequence completion and full lifecycle acceptance remain pending. Packet capabilities stay zero; no game is active.

November 8, 07:04 UTC active checkpoint. FG passes48 combined Windows groups, seven x86 profiles, the AMD64 composer,666 API regressions and496 exact inputs. The same DLL passes original AG and college AH2 with15.340/15.352 seconds after copied clock advance, a qualified writable main-image span and a healthy allocator-generation ledger; each restores all99 files and releases custody. The original has26 heaps,13,330 live allocations and12 backing regions at the final sample. These are allocator/module observations, not complete native source/free coverage or graphics admission. New constructor binding selects an observed allocation by recorded backing ancestry, refuses partial boundaries and delegated holes, and pins its entry generation through constructor return;1,091 local sanitizer checks pass. Its optional native adapter and guarded source-copy/destructor-withdrawal fixture are under FH Windows verification. Persistent shader ownership, complete source/context/free coverage, material/semantic/sequence completion and full lifecycle acceptance remain unfinished. Packet capabilities stay zero. No game is active.

November 8, 06:45 UTC active checkpoint. FF2 passes 21 Windows adapter groups with a real executable-image getter/mapping test, exact module-span joins and stale/partial provenance refusal. The portable module-storage model passes 991 ASan/UBSan/TSan checks and the allocator-event join passes 382; fixed image storage is represented separately from native heap objects and allocations. The pinned getter at0058D620 returns01665F28+00200000. The qualifier requires its exact bytes, the main x86 image identity and bounds, executable getter mapping and committed writable image storage. PE headers use32-byte reads, retaining the CPU reader's48-byte cap. FG is recompiling all seven profiles and all48 combined groups over496 inputs. The original AF refusal is restored/released and fully preserved externally. Actual-game qualification of this module span, source/free/context coverage, persistent shader ownership, semantic/sequence completion and full lifecycle acceptance remain pending. Packet capabilities remain zero; no game is active.

November 8, 06:23 UTC active checkpoint. FC2's source-memory join observes 24 heaps, 331 allocations and one backing region, then refuses a backing-adder return. Original-ae2 records 3,255,389 resource calls and zero routing, contract or allocator faults; it stops before moving-clock acceptance. All 99 baseline files restore and custody releases at 06:07:45 UTC. The complete 44-file/148,614,585-byte run archive is independently verified and fully fsynced externally; only its verified redundant ZIPs and five closed raw payloads were retired. FD3 passes 21 Windows adapter groups with versioned caller/span/result/parent diagnostics and a forced backing-provenance refusal; these readable snapshots grant no ownership. FE is compiling seven profiles and rerunning all 48 combined groups over 495 frozen inputs. Source/free/context qualification, persistent shader ownership, semantic/sequence completion and full lifecycle acceptance remain open. Packet capabilities remain zero; no game is active.

November 8, 06:07 UTC active checkpoint. FC2 passes 48 combined Windows groups, all seven x86 profiles, the AMD64 composer, 666 API regressions and 495 exact frozen inputs. The 118 patch windows remain exact and disjoint, with no raw near-branch interior candidates and seven reviewed entry aliases. The new opt-in source-memory profile joins typed allocator lifetimes to the generation ledger; production CPU callbacks and the held debugger barrier pass. Original-ae stopped because the monitor parsed the native log prefix as a counter; all 99 files restore and custody releases at 06:05:28 UTC. The corrected parser passes eight local and Windows replay/refusal checks. Original-ae2 is active on the same DLL under bounded restoration; native ownership/free qualification remains pending. Closed compiler intermediates are hash-verified and fsynced externally with new-path Windows restore proof; current binaries, maps, inputs and receipts stay unpacked. Persistent shader ownership, source/context/free coverage, semantic and sequence completion and full lifecycle acceptance remain unfinished. Packet capabilities stay zero.

November 8, 5:54 UTC active checkpoint. EY original-ac and college-ad both pass on the same DLL, each with at least 15 seconds of advancing copied clock and zero routing, contract, allocator or reference-cycle faults. College-ad records 8,158,359 resource calls, 37,160 balanced allocator calls and 1,040 same-thread reference cycles; all 99 baseline files restore and custody releases at 05:42:54 UTC. The opt-in source-memory join now passes FB2's 21 Windows adapter groups, including real captured constructor/destructor gateways, controlled allocation results, refusal cleanup and the 118-site held debugger barrier. EZ's copied native class view passes 39 Windows checks. FC is compiling seven profiles and rerunning 48 combined groups over 495 frozen inputs. Native source ownership/free coverage, persistent shader ownership, semantic/sequence completion and full lifecycle acceptance remain unfinished. Packet capabilities remain zero. No game is active.

November 8, 05:42 UTC checkpoint. EY passes 45 combined Windows groups, 489 exact frozen inputs, six x86 adapter profiles, the AMD64 composer and 666 API tests. The ready4 handoff now creates a distinct successor operation and pins older source records through late callbacks; 6,758 local sanitizer checks and the production native-instruction regression pass. A DWORD containing unused selector upper bits caused the debugger stack refusal; the checker now passes only the architectural 16-bit selector and keeps every TEB/stack/code-range check. Actual held-barrier fixtures force upper-bit changes on all inspected threads; the 118-site installer passes. Original-ac passes 15.357 seconds after clock advance, 6,570,594 resource calls, 37,412 balanced allocator calls, 97 balanced lifetime calls, 940 same-thread shader reference cycles and zero faults. All 99 files restore and custody releases at 05:35:59 UTC. The same-DLL college-ad campaign is active; paired acceptance remains pending. Native destructor/realloc paths remain unobserved. The allocator-event join has 336 Windows checks; a new copied-class view has 39 local sanitizer checks, with Windows validation and native wiring pending. Persistent shader ownership, native source/context/free qualification, material/semantic/sequence completion and full lifecycle acceptance remain unfinished. Packet capabilities remain zero and full framework acceptance remains false.

November 8, 05:13 UTC checkpoint. EV2 passes 44 Windows groups, 485 exact inputs, six x86 adapter profiles, the AMD64 composer and 666 API tests. Its 118 patch windows have exact bytes, disjoint spans, no raw near-branch interior candidates and seven reviewed vtable entry aliases; the ambiguous early-RET data value is preserved. Original-ab observes 51 constructors, 19 backing additions, 21 initializations and six bulk clears with balanced returns, zero allocator/routing faults and 922 shader reference cycles. The resource observer refuses a new ready4-to5 claim while the previous creation is still active (65552); this run is not accepted. All 99 files restore and custody releases at 05:03:42 UTC. The explicit successor-operation fix passes 6,758 local ASan/UBSan/TSan checks, including late finalizers and address reuse; its captured-native regression and updated combined/native campaigns remain pending. The source-memory model has 894 Windows/sanitizer checks. Its normalized allocator-event prototype has 260 Windows checks and an extended 336-check local sanitizer pass; neither supplies native source/free coverage. Persistent shader ownership, material/semantic/sequence completion and full native lifecycle acceptance remain unfinished. Packet capabilities remain zero; full framework acceptance remains false.

November 8, 04:40 UTC checkpoint. EQ3 passes 43 Windows groups, six x86 profiles, the AMD64 composer and 666 API tests. Original-z and college-aa pass advancing-clock forwarding with 37,126/37,137 balanced allocator calls and 1,020/972 same-thread shader reference cycles, with zero faults; both restore all 99 protected files and release custody. Native reallocations remain unobserved. ES2 passes 826 Windows and local ASan/UBSan/TSan checks for independently generated backing regions, parent-storage withdrawal and address reuse. ET2 passes 148 original/observed constructor/backing executions and 67,556 checks, preserving registers, flags, FX state, LastError and native caller words. The expanded 114-site installer is under CPU validation; these lifetime hooks are not installed in a game. Persistent shader ownership, source/free qualification, material/semantic/sequence completion and full native lifecycle acceptance remain unfinished. Packet capabilities remain zero and full framework acceptance remains false.

November 8, 03:45 UTC checkpoint. EH2 original-x and college-y2 pass advancing-clock forwarding plus 1,046/1,040 same-thread shader reference cycles, with zero routing, contract or cycle faults. Both restore all 99 protected files and release custody. The six complete allocator methods now pass 1,050 original/observed scenarios (2,100 executions), 19,387 native checks and 21,390 routing checks, preserving registers, flags, FX state, LastError and native caller words. Their 21-site v2 startup transaction passes a held OS debug barrier. The resize/constructor-generation model passes 643 Windows and local ASan/UBSan/TSan checks; failed system resize preserves the old generation, while native heap replacement requires exact allocation/free receipts. The enlarged 105-site combined installer and sixth native profile are being validated. Native persistent shader ownership, allocation/free bounds, material admission, semantic owners and sequence/visual lifecycle remain unfinished. All packet capabilities remain zero; full framework acceptance remains false.

November 8, 03:13 UTC checkpoint. EH2 passes 38 combined Windows groups with 466 exact frozen inputs, five x86 adapter profiles, the AMD64 composer and 666 API tests. Its new private reference-probe profile runs same-thread COM cycles only after retiring the observer frame and metadata lock. The complete shader factory passes 72,539 checks over 756 cases in four modes (3,024 executions), including reentrant native CAS calls, zero diagnostic reference counts and inaccessible-object refusal. The portable cycle/return-slot model passes 8,040 local sanitizer checks. The source-allocation/constructor model passes 440 Windows and local sanitizer checks; its entry token prevents rebinding to storage reused during construction. Separate EI3 executes four complete allocator methods over 738 controlled scenarios (5,817 checks); six method bodies are captured, while native allocation/free coverage remains unqualified. EH2 reuses only hash-verified identical compiled adapters from EH, rebuilds the update fixture and repeats every runtime/API case. EH had refused an occupied fixture address before hooks; the reservation guard remains intact. EB original-w passes advancing-clock forwarding with 8,147,615 calls and zero routing/contract faults, restores all 99 files and releases custody at 02:42:13 UTC. Its copied first/other device pointers match; the second path is unset. Native reference-cycle acceptance is pending original-x. All packet capabilities remain zero and full-framework acceptance remains false.

November 8, 02:25 UTC checkpoint. EB passes all 36 combined Windows groups with 458 exact frozen inputs, four x86 adapter profiles, the AMD64 composer and 666 API tests. The physical shader-reference bank defaults to same-thread retention/release, queues last-owner retirement, and calls its backend outside metadata locks. Its Windows bank/model gates pass, including 975 retained-reference/event checks. The complete pinned shader factory passes 44,704 checks across 756 scenarios executed in original, observed and retained modes (2,268 invocations), with identical machine state and controlled COM leaves. These remain CPU proofs; native physical ownership and callback admission are unaccepted. DW original-t/college-u and original-v pass advancing-clock forwarding; V records 7,894,698 calls with zero routing/contract faults and restores all 99 files/releases custody at 02:11:43 UTC. Its three allocator samples identify the system allocator, game mutex wrapper and delegated game heap; partial material censuses hit their candidate bound. The new source-memory allocation/generation model remains outside EB and has no native allocator/free qualification. All packet capabilities remain zero; full-framework acceptance remains false.

November 8, 01:44 UTC active checkpoint. DW passes 35 combined Windows groups with 455 exact frozen inputs, all four x86 adapter profiles, the AMD64 composer and 666 API tests. The expanded resource fixture passes 102,882 checks; the overlapping-upload model passes 6,471 checks under local ASan/UBSan/TSan. Original-s on DU2 refused a real kind1 texture CAS6->5 while an earlier upload was still running; routing stayed intact, all 99 baseline files restored and custody released at 01:25:13 UTC. DW reserves that exact upload before the original atomic operation and binds the later queue publication to it. Its Windows regression forces the earlier upload to finish before the overlapping CAS returns. Original-t is active under the bounded restoration supervisor, with new read-only, DLL-bound per-hook counter samples. Native acceptance of the fix remains pending. All packet capabilities remain zero; full-framework acceptance remains false.

November 8, 01:24 UTC active checkpoint. DU2 passes 35 combined Windows groups, all 455 frozen inputs, four x86 adapter profiles, the AMD64 composer and 666 API tests. Its shader reference model passes 713 Windows and local sanitizer checks. DR4 executes all 756 original shader-factory scenarios with and without the production observers: 26,414 checks preserve machine state across all five normal returns. The integrated adapter has 27 routes and 84 exact, disjoint patch windows. The update ABI fixture now relocates its controlled context/queue leaves, while retaining the original caller words; 140 scenarios pass 3,746 checks. DU2 reuses only hash-verified, byte-identical compiler artifacts from DU, rebuilds that changed fixture, and repeats every runtime/API check. Original-s is active under the bounded restoration supervisor; native shader acceptance is pending. All packet capabilities remain zero and full-framework acceptance remains false.

November 8, 01:09 UTC active checkpoint. DO3 passes all 33 combined Windows groups with 447 exact frozen inputs and 666 API tests. Original-q and college-r each pass at least 15 seconds after the copied clock advances, with 7.65M and 7.78M observed calls and zero routing/contract faults. Both close, restore all 99 baseline files and release custody. The immutable shader-reference model passes 486 Windows checks and local ASan/UBSan/TSan checks. DQ executes the complete pinned shader factory over 756 controlled x86 scenarios (9,314 checks), including cached AddRef and all five normal returns. These are CPU proofs with controlled COM leaves, not shader ownership or callback admission. DR is checking the production entry/exit observers. Current source has changed since DO3; the next combined/native gates remain pending. No game is active. Full-framework acceptance remains false and packet capabilities remain zero.

November 8, 00:39 UTC active checkpoint. DM passes all 33 combined Windows groups, 446 exact inputs, all four x86 adapter profiles, the AMD64 composer and 666 API tests. Its 26-route observer applies 78 independently checked patch windows. Original O passes bounded forwarding with 3.27 million observed calls and no contract/routing faults; this ends during the intro, without a moving clock. All 99 files restore and custody releases 00:09:35 UTC.

Longer original P observes a moving copied game clock, but the model refuses a concurrent native 3->2 request before its observed successful 2->3 claim is copied. Routing continues without fault; closure/restoration/release completes 00:16:11 UTC. DN corrects that pending-state check and passes all 17 Windows resource groups, including a deterministic native two-thread reproduction; local ASan/UBSan/TSan each pass 6,180 checks. The expanded native fixture passes 102,818 checks. The fix has not yet passed the next live-game run.

DO compiles all production profiles and passes the resource checks, but its separate update-exit fixture cannot reserve a fixed test address already owned by Windows. DO2 encounters another occupied address. No hooks execute on those setup failures. DK8 now distinguishes that precondition, never takes over the allocation, and bounds fresh-process acquisition attempts. All 140 ABI cases must still execute successfully. A storage guard stops its first attempt before stage creation. Closed predecessor stages are being preserved externally; current DM/DO/DN sources, games, saves and reference packages remain protected.

The next combined gate will reuse only compiler outputs whose production inputs are identical, rebuild the changed fixture, and run every runtime/API case. The next game harness also requires 15 seconds after the copied clock advances. All packet capabilities remain zero; 3 scoped accepted, 17 partial and 11 unavailable inventory entries remain. Shared shader-reference ownership, source/context/free coverage, material/list closures, semantic owners, native sequences, full original/college lifecycle and author delivery are unfinished. Network packages remain deferred.

November 8, 00:02 UTC active takeover checkpoint. DJ remains the last complete 32-group combined Windows gate (440 inputs, source 95bbb674...). DL3 now passes all 17 resource groups, including the actual production adapter, repeated native vertex/index buffer updates and the held debugger transaction. DK6 executes the actual entry/exit observers over 140 pinned x86 update paths: 3,740 checks preserve caller words, registers, flags, FX state and LastError. The new adapter has 26 routes and 78 patch windows. DM freezes 446 inputs (source 452f15ee...) for the expanded 33-group combined gate; it is running. These changes are not yet installed in a game. Packet capabilities remain zero.

Original-game N closes/restores all 99 files and releases custody at 23:44:12 UTC. Its corrected read-only capture follows the complete chain: pass reference -> [reference+14] shader object -> object enabled +10 / source +18 -> optional kind-7 aliases -> native kind-9 header. L and M shader candidates remain unqualified; both earlier shortened chains were incorrect. N contains 768 shader observations across 384 partial material candidates: 766 state4 and two state1 observations, 84 distinct sources, and 18 shared physical-backing groups. All 2,190 archived run files and overlapping snapshot spans verify. This is pointer-layout evidence, not ownership or callback admission.

All 14 takeover staging attempts are closed; 12 launched game processes. Closed CPU archive14 preserves all 866 paths/metadata and verifies a new-path Windows restore before retiring redundant copies. Current candidates and games are excluded. The full framework remains active: 3 scoped accepted, 17 partial, 11 unavailable, with all 31 active and 22 retained native categories in scope. Coalesced/direct updates, shared shader references, source/context/free, material/list closure, semantic owners, sequences, original/college lifecycle, Reset/resize/wide acceptance and third-party delivery remain unfinished. Network-specific packages remain deferred.

November 7, 23:21 UTC takeover checkpoint. DH freezes 440 inputs and passes 32 Windows groups: four complete x86 adapter profiles, the AMD64 composer, resource/packet fixtures and all 666 public API tests. Every current frozen input is exact. All 73 patch windows remain exact and disjoint against the pinned image; independent raw scans find no overwritten interior or literal-pointer candidates. This is CPU evidence; packet capabilities remain zero.

The indexed resource model now also refuses a backing already held by an uncertified live source, through both synchronous and queued creation paths. DG2 passes all 17 resource groups; exact-source local ASan/UBSan/TSan checks pass 5,847 checks per run. DF2 independently executes the original x86 parameter dispatcher against 39 copied streams: all 260 records agree, with 658 checks. Its table leaves are stubbed; no material callback or resource ownership is admitted.

Original-game L reaches its confirmed requested match and captures 192 partial material candidates from three read-only debug events, with 21.8–28.2 ms attach and pause intervals. Each sample stops at its 64-candidate bound. Eleven distinct pass records and 39 streams are copied. The first shader sampling chain used wrapper+14; pinned E080E0 instead requires enabled+10/source+18. Those L shader candidates remain unqualified. The sampler is corrected with a regression check; its next native harness must be re-frozen before use.

Twelve takeover staging attempts are closed; ten created game processes. All 99 DW baseline files are restored and unchanged, redundant rollback ZIPs are verified externally, and the latest readback shows no game, GPU lock or owned CPU worker. DH has not been installed in a game.

All 31 active and 22 retained categories remain in scope: 3 scoped accepted, 17 partial and 11 unavailable. Full framework acceptance remains false. Existing resource updates/direct uploads, source/context/free, materials and list closure, semantic owners, native sequences, original/college lifecycle, Reset/resize/wide verification and third-party delivery remain unfinished. Network packages stay deferred. See the takeover STATUS and per-candidate receipts for exact evidence.

November 7, 19:44 UTC combined checkpoint. CN source d5c19981... passes29 Windows CPU groups with277 exact frozen inputs, three complete x86 /W4 /WX adapter profiles and the AMD64 composer. The shared native resource fixture now preserves caller return words for all22 paths, including waiter, queue, creator and release tails. Native cancellation waits for queued work; the source address is reconstructed before its old creator returns without certifying or changing the new resource. Eight simultaneous producers pass Windows and local ASan/UBSan/TSan checks after short registry transactions were serialized. The expanded resource fixture passes84,680 checks.

No new game/GPU/install/custody change or callback admission occurred. Packet caps remain0; all31 active/22 retained categories and3 scoped accepted, 17 partial,11 unavailable remain. Source/context/free qualification, material/callback/list admission, semantic owners, native sequence/lifecycle, full original/college visual verification and third-party closeout remain in scope. Network packages stay deferred. See the takeover STATUS and CN receipt.

November 7, 18:51 UTC takeover checkpoint. CD2 source d87c8f7a... passes28 combined Windows CPU groups, three complete x86 /W4 /WX adapter profiles and the AMD64 composer. All268 frozen files remain exact. The shared native resource event joiner now uses actual observed constructor, creation, destruction, publication and release tokens. Native upload/retry and exact pool-node reuse before a producer returns pass. Source tombstones and operation slots remain pinned through the original destructor. Caller return words stay intact across constructors, copy, destruction, CAS, update and both drains.

See analysis/presentation-framework75/takeover/README.md and STATUS.json. This is CPU implementation evidence: no new game/GPU/install/custody change. Packet caps remain0; all31 active/22 retained categories and3 scoped accepted, 17 partial,11 unavailable remain. Source/context/free and remaining tail/queue integration, callback admission, semantic owners, sequence/lifecycle, full native verification and third-party closeout remain unfinished. Network packages stay deferred. The18:59 UTC readback again verifies99 DW baseline files unchanged.

November 7, 18:05 UTC takeover checkpoint. Current BY source ce96760c... passes 28 Windows CPU groups, three complete x86 /W4 /WX DLL profiles and the AMD64 software composer. The interrupted queue-exit checks are recovered. Distinct tail transfer, real-SEH stack reuse, depth/reentrancy and outer/abandoned-inner cases pass. All ten original constructors additionally use a native exit observer in the owned CPU fixture, retaining their original return words and exact registry/Mutation bindings. The optional queue adapter has exact-code and early single-thread guards and remains disabled by default.

Read analysis/presentation-framework75/takeover/README.md and STATUS.json. No new game/GPU/install/custody change occurred. The 18:08 UTC readback verifies all 99 DW baseline files unchanged. Packet caps remain0; all 31 active/22 retained scopes and 3 scoped accepted,17 partial,11 unavailable remain. Resource and callback integration, remaining owners, sequence/lifecycle/resize/Reset and native combined acceptance are still unfinished. ESPN/network work stays deferred. Same-chat automatic continuation is active and bounded; delete its recorded heartbeat on user stop/pause or actual completion.

Current task, November 6 local / November 7 UTC: complete the entire neutral framework before building network packages. ESPN artwork is deferred, with all 31 active inventory entries, 22 retained native categories and native lifecycle requirements retained. See the work75 record for current changes, preservation and verification. Historical work74 checkpoints below retain their original evidence scope; they do not grant full completion.

November 7, 15:58 UTC checkpoint. Packet-bound early publication and prepared pixels pass the AZ4/AZ5 Windows gates, including both full x86 native build profiles and actual software-composer output through the Host fixture. The reviewed BB3 registry retains exact creation/release/domain identity. BD3 executes the actual native buffer create/update queue synchronously in a CPU fixture. BE5's seven groups qualify local captured-code entry/return detours and independent corrupted-stack/exception handling; they do not install game hooks. See the current work75 record and BE5's exact scope.

Resources/material callbacks, asynchronous native lifetime integration, remaining semantic owners, lifecycle and Reset/resize/ultrawide game acceptance remain unfinished. Accepted packet caps are zero; no new game/GPU run occurred. All 31 active/22 retained categories remain. Coverage is 3 scoped accepted, 17 partial, 11 unavailable, 22 retained native; full completion is false.

The independently reviewed 75a college run has 2,970 schema checks, two equal acknowledged result consumers and restored 99-file DW baseline. Half labels, producer recovery and minimize/restore have scoped pixel evidence. The trace still has 123 main-scorebug fallback frames after prior replacement; no stable main FINAL capture was obtained. The 75b draw-demand continuation is undergoing combined-source headless checks and has no native-game acceptance. Full neutral framework acceptance remains false; see work75 for the remaining gates.

November 7, 05:54 UTC checkpoint. Current candidate75h freezes the corrected plane scheduler over the exact F native/renderer binaries and API dev57 inputs. All 197 Windows presentation tests and 16 managed regressions pass. F's actual HAL fixture passes 964 checks. G's stopped original-rules segment independently shows all ten public on-court names and 2,596 fully replaced lineup draw frames, but exposes manual-overlay starvation behind inactive prepared native planes. H fixes that scheduling defect without changing the eight-plane/32 MiB bounds, freshness or native context guards. H original/college acceptance is in progress; no new owner or full completion is accepted from these checks alone. See original75g-flat-scoped-review.json, checks-h.json and PLANE-SCHEDULING.md.

November 4, 16:53 UTC checkpoint. DW's frozen comparison-table diagnostic is running as p74-live-table-original-dw, acquired16:52:42UTC with a17:12:42UTC cap. All384 sources, DY harness and99 restored baseline files verified; private probe DLL3393e032 is selected only in the isolated game. Preflight free space was4,085,370,880bytes. Table semantics/ownership are pending actual image review. Consult fresh custody before any competing work.

The local compact-overtime formatter correction passes two portable checks and20 Windows scene/software-pixel tests against the captured START OF 2OT fields; no production DLL or installed game worker changed. The first isolated check failed at Python import setup before tests; EA2 corrects that helper and passes. See overtime-heading-ea-review.json and overtime-heading-ea2-checks.json. This is software-pixel evidence, not a new native-game pass. Own closed evidence/private-state retirement receipts DZ/EA/EB preserve independently hash-verified external archives; current games, game-before backups, source and peer paths remain untouched.

November 4, 16:43 UTC checkpoint. DY completes natural MIL8-PHI10 after double overtime with 4,960 schema checks and two identical acknowledged result deliveries. Independently reviewed first/second-half, OT/2OT, nonfinal period-end and FINAL scorebug pixels pass; all10 scorebug draw frames in the FINAL capture fence are replaced. The committed odd frame counter48791 confirms the diagnostic-reader correction. Producer recovery/minimize-restore pass, all99 files restore, the task is removed and custody releases at16:35:14UTC. The179-file external review archive verifies SHA558a8a5e387d8539c15895aca3e01c947483f76797481e72d302703486fc049f. See live-finish-dy-review.json, period-pixels-dy-live.json, period-brackets-dy.json and coverage-dy-live.json.

This is scoped period-label acceptance:246 scorebug fallback frames remain after prior replacement outside the deliberate producer outage. Native owner32 falls back during the reviewed2OT-start/FINAL captures. A concrete formatter bug rejects the copied native heading START OF 2OT; the local fix and two portable formatter tests pass, with Windows pixel validation pending. No production DLL changed. DW's frozen table probe remains unlaunched; use fresh shared custody and adequate storage headroom. All31 active entries,12 unavailable owners, smooth transitions, actual resize/ultrawide and native sequence semantics remain in scope; full framework acceptance is false.

November 4, 16:17 UTC checkpoint. The unchanged DV production DLL completes DX's natural MIL4–PHI11 result with two reconciled consumers, 3,042 schemas and recovery/minimize checks. All99 baseline files restore and custody releases16:05:30UTC. FINAL pixels remain unaccepted because an evidence-only parity check incorrectly refused legitimate odd frame counters. DY fixes that diagnostic rule and reads frame metadata under its nonblocking mutex; native ownership/context/age rules and production inputs are unchanged. Windows reader tests and the actual x86 channel/64-bit producer fixture pass. The bounded DY retry is active through a16:37:42UTC cap; fresh custody governs. Full framework acceptance, remaining owner coverage and transition smoothness remain unfinished.

November 4 — presentation Dell work reauthorized by Brent. “Continue with the Dell available!” supersedes the October 29 hold below for this task. Fresh custody, storage and process guards still apply. See analysis/presentation-framework74/dell-resume-20261104.json; the historical hold remains unchanged. DV2's missing FINAL/result acceptance and all remaining inventory gaps remain open.

October 29, 21:07 UTC — all Dell work held by Brent. Do not access, build, launch, collect evidence or reacquire the Dell until Brent explicitly reauthorizes. This supersedes every earlier queue promise and prepared run below. All presentation GPU/headless queues are cancelled; DW never launched. See analysis/presentation-framework74/dell-hold-20261029.json.

DV2 was stopped on request: exact owned game/broker closure at 21:06:20 UTC, all renderer/watchers confirmed closed, all 99 isolated baseline files restored and hash-checked at 21:06:37 UTC, exact task removed and lease released at 21:06:51 UTC. Final shared readback at 21:07:26 UTC shows all 26 projects released, no lock and 5,464,788,992 bytes free. The coordinator and MVP received the exact release; no peer resources were changed. Current isolated baseline is restored DV dependency DLL a6a0d54b; installed harness is DV2.

DV2 reached a finished state after 2,917 schema checks, with reviewed first/second-half and recovery images, but was stopped before reconciled result delivery. The ending captures failed stability qualification; FINAL visual acceptance remains false. Complete trace/archive collection is pending and must respect the Dell hold. See live-finish-dv2-stopped-review.json. All 31 active entries remain in scope; full framework acceptance remains false. Local records and all prior verified archives are preserved.

October 29, 20:56 UTC checkpoint. The unchanged DV production retry p74-live-finish-college-dv2 acquired at20:56:20UTC with a21:16:20UTC hard cap after fresh all-project/storage/process guards. All384 frozen sources, the installed DV2 harness and99 restored baseline files verified; dependency DLL remains a6a0d54b and preflight free space was6,048,624,640bytes. Native FINAL and natural-result acceptance are pending. MVP's exact run2 release was received; its short navigation retest is next after this run. Notify it after exact closure/restoration/release and hold further presentation acquisition. Consult fresh custody rather than this dated record.

October 29, 20:51 UTC checkpoint. DV remains closed/restored/released with partial native evidence only. MVP confirmed and closed its two stalled read-helper processes at20:44:02UTC; no presentation action touched peer processes or system settings. Fresh read-only shared status at20:51:34UTC shows6,767,210,496bytes free and MVP's bounded uniform-components-2 run active. Presentation waits for its exact final release. The unchanged DV production candidate and DV2 harness are prepared for p74-live-finish-college-dv2 to complete the missing natural result and FINAL checks.

DW's local private probe now observes the practice free-throw UI and manual replay controls without attempting one-frame omission trials. Original localization distinguishes both operational resources from the still-unresolved broadcast entries. Production inputs remain DV; no new native owner or full-framework acceptance follows. The frozen DW probe passes175 Windows presentation tests,16 managed regressions and138,385 exact x86 controller/queue checks; it has no game acceptance. Source SHA2f377675ff504fdcee53e66686ee41ef08844136c1c88bf54bae8ab53e0def9b; probe DLL3393e032. See replay-controls-scope-dw.json, free-throw-resource-scope-dt.json, framework74dw-source-review.json and native-finish-dv2-run-plan.json.

October 29, 20:43 UTC checkpoint. DV visibly verifies the custom first-half, intermission and second-half labels, producer recovery and minimize/restore. Its nonfinal phase11 capture shows1st at0.0 with a copied finished=false value; all12 scorebug draw frames inside that capture fence are fully replaced. The broader trace retains212 fallback frames after earlier scorebug replacement, excluding the deliberate producer outage. This is scoped evidence, not continuous smoothness or complete framework acceptance.

A fresh storage guard stopped DV at42 seconds in the second half after2,150 schema checks. No natural final/result delivery or FINAL pixels were reached. The owned game/broker/renderer close at20:39:39UTC; all99 files restore at20:40:02UTC, task is removed and custody releases20:40:07UTC. Current isolated baseline is DV dependency DLL a6a0d54b; installed harness is DV2. See live-finish-dv-review.json, period-pixels-dv-live.json, trace-review-dv.json and coverage-dv-live.json.

The external review ZIP verifies146 payload files, SHA463709fd0b946fe6da7a94f3c17909a5bf9b564a4bdd7fa081a1286e11cb5aea. Fresh20:40:33UTC shared readback has all projects released/no lock and1,929,469,952 bytes free; later readback fell further. Read-only accounting found large PowerShell processes using MVP read-helper paths; their identities were sent to that owner, with no peer process stopped. MVP's bounded retest remains next once fresh storage/shared/process guards pass. Presentation holds further acquisition. All31 active and22 retained entries remain in scope, including12 unavailable active owners; FINAL validation and actual game resize/ultrawide remain open.

October 29, 20:32 UTC checkpoint. DV passes 175 Windows presentation tests, 16 managed-launch regressions and 138,385 exact x86 controller/queue checks. ScoreFacts.2 adds a copied finished bit and same-sample phase identity; native_scoreboard.v4 can distinguish nonfinal period-end from FINAL without changing the public API or legacy ScoreFacts.1 layout. Source SHA 06480d6d468cea9a5f68fc99d7008fd0f2d6970e8e362efa9109071dc66544a0. These are headless results; native transition pixels remain pending.

The focused college run p74-live-finish-college-dv acquired custody at 20:31:11 UTC with a hard 20:51:11 UTC cap. All 384 frozen sources, the installed DV2 capture harness and 99 restored DT baseline files verified before selecting dependency DLL a6a0d54b6e182523a327d5c260cf9b0d686b2eec630e9928b8944c3786081f07. Fresh preflight free space was 4,210,307,072 bytes. DV2 supersedes the uninstalled first DV capture harness; its generic notice capture matches DS2 exactly. Consult fresh custody and closure records before acting. MVP released its hold, then requested its bounded retest after DV; notify it after exact restoration/release and hold further presentation acquisition for that window.

Two additional verified external archives preserve only closed own after-game/private-state files: closed-runs-du.zip has 304 files (SHA ebf0b6705df731e0c806cc4b3f0efd60b843536d9b3d64ecce4664a029c6d0cf), and closed-runs-dv.zip has 312 files (SHA 56730209dffbede67fde8cbea5ec0ee8d365d0b74634a34364d8aedd02e44732). Their exact redundant Dell copies were retired after verification; corresponding restore helpers require new destination paths. Current game/game-before/source and peer paths remain protected.

All 53 inventory rows remain in scope: 31 active and 22 retained native, with 12 active owners still unavailable. Full framework acceptance remains false; see coverage-dv.json. DU's reserved table-occurrence probe is included in DV's diagnostic DLL, but this production run does not use that probe and cannot establish table-owner coverage.

October 29, 20:10 UTC checkpoint. DT diagnostic completes natural MIL13-PHI5 with4,657 schema checks and two reconciled result consumers. Its16 numeric predicate branches each immediately pair with the matching original stop and clear. All83 accepted stop records have parent provenance;32 refused records remain excluded, with zero queue contention/overflow. Native highlight playback and the return to the final menu were reviewed. This is private observation evidence, not new production sequence authority. See live-dt-review.json, controller-predicate-live-dt.json and coverage-dt-live.json.

All99 isolated files restore at20:02:17UTC, the owned task is removed and custody releases20:02:32UTC. Current isolated baseline is restored DT probe DLL796ae8b1; installed harness remains DS2. Further presentation GPU acquisition is held for MVP's coordinated bounded window and final release. MVP's first fresh guard yielded to Fertitta; neither that queue nor these dated records replace fresh shared/storage/process guards.

The DT review archive verifies290 payloads, SHAadddf325fd153defadbb24431e50c5273236415bed9b3227adf5003564fdc67f. Only88 verified redundant raw files were retired, recovering333,414,400allocated bytes; use restore_closed_evidence_du.py into a NEW directory. Separately, closed-runs-dt.zip preserves206 files from three older closed own runs, SHA8119a5723b888e42810e97dd12e42cceb9b1691ec2eb6342af13be789c12daad; restore_closed_runs_dt.py restores new paths only. Current game/game-before/source/peer paths remain protected.

The free-throw resource previously watched is practice UI, confirmed by native localization/dispatch; ordinary-match free-throw graphics remain unresolved. DU currently adds only two reserved diagnostic captures for six submitted comparison-table fields under the existing scorebug root; it is unbuilt and has no native acceptance. DS phase11 fallback, missing owners, smooth content changes and actual game resize/ultrawide remain unfinished. All31 active and22 retained-native entries remain tracked; full framework acceptance is false.

October 29, 19:48 UTC checkpoint. DT private controller-predicate observation is active as p74-live-predicates-original-dt, acquired19:47:48UTC with a hard20:07:48UTC cap. The99 restored DS files passed before selection of exact probe DLL796ae8b129f16ca8288ab3ad31c044e340270ad1aef5e9d6ea1516660d924053. Current installed harness is DS2 in diagnostic mode; game PID512 starts19:48:14UTC. No DT result or new sequence authority is yet accepted. MVP granted this bounded window during headless compilation; notify it directly after exact closure/restoration/release. Always consult fresh custody.

The closed DP/DQ/DS post-run saves and private API state are now externally preserved in closed-runs-ds.zip:255 files,743,595,951logical bytes, SHA608d9656d313485a643178b642f287a06129e516dd3d5f4fe1265741b6e93df2. Every file was verified before its redundant Dell copy was retired; use restore_closed_runs_ds.py into a NEW directory. DS raw offload and duplicate reviewZIP retirement separately preserve the verified167-file review archive. Current game/game-before/source/peer paths remain protected.

October 29, 19:44 UTC checkpoint. DS completes college MIL7–PHI4 with2,857 schema checks and two identical reconciled result deliveries. Exact DLL31138f6e visibly shows1st during the first half/intermission and2nd during second-half play; producer restart and minimize/restore recover with zero host failures. All99 files restore at19:36:06UTC and custody releases19:36:23UTC. The external archive verifies all167 payload files (SHA8571717c18f571ff5cb2a50df2e2df35f1fb17b315534a96e43c686b155783cc). See live-period-ds-review.json and coverage-ds-live.json.

This is scoped period-label acceptance. The trace retains152 scorebug fallback frames after replacement; its longest19-frame/297ms gap is phase11, for which copied data deliberately lacks finished-state authority. No smooth-transition or complete-framework acceptance. The native final menu retains its quarter-column layout. Current isolated baseline is restored DS and installed harness is DS2.

DT is frozen and headless-verified, with private probe DLL796ae8b129f16ca8288ab3ad31c044e340270ad1aef5e9d6ea1516660d924053. It has not launched. MVP authorized the next bounded diagnostic during its headless compilation, but fresh19:42:53UTC custody showed Bramlage running; presentation remains released. Recheck all shared/storage/process guards before acquisition. Only60 verified redundant closed DS raw files were retired,240,343,282logical bytes; restore_closed_evidence_ds.py restores the external archive into a new directory. No current game/save/source or peer files were retired.

October 29, 19:26 UTC checkpoint. The bounded DS college period-label run p74-live-period-college-ds has acquired presentation custody at19:25:38UTC, with a hard19:45:38UTC cap. Fresh verification matched all378 frozen inputs, the DS2 harness and all99 prior DQ baseline files before installing dependency DLL31138f6ed250e7d32403acd2c526ff0223dd97886f2fc2a469cff37b024bc565. The task is scheduled; native period-label acceptance and restoration are pending. Consult fresh custody and the run's acceptance/closure records before acting. Allen's final release has arrived. DT remains headless-only; no full-framework acceptance or new owner coverage is claimed.

October 29, 19:13 UTC checkpoint. DT's private predicate observer passes171 Windows presentation tests,16 managed regressions and138,385 exact x86 controller/queue checks. The new fixture executes the original224-byte predicate unchanged and with the shared hooks across1,024 combinations; its two true-return branches retain identical results and controller state. No DT game or production sequence authority is accepted. See controller-predicate-dt-review.json; frozen source SHA853ad2ef71df9e4043e7841ae1ab67ebf919bd9ef188eca34f91755a9ec85120.

The exact DS period-label candidate and DS2 capture/recovery harness pass headless checks. The planned college run requires stable reviewed first/second-half captures and natural reconciled results; it does not require or claim unrelated award/substitution occurrences. See native-period-ds-run-plan.json. Allen's final details-c run currently holds custody; presentation remains released and waits for its final restoration/release before fresh guards. Installed harness remains DP; current isolated baseline remains restored DQ.

All53 inventory rows retain their prior scope in coverage-dt.json:31 active and22 retained native, including12 active entries with unavailable owners. Full framework acceptance remains false. Verified external DP/DQ archives now replace129 redundant raw Dell evidence files and two duplicate ZIPs,654,905,375logical bytes total. Saves/current game/source/peer paths remain; restore_closed_evidence_dt.py restores either review archive into a NEW directory. DT's first headless attempt refused storage before stage creation; its subsequent guarded build passed. Fresh storage checks remain mandatory.

October 29, 18:54 UTC checkpoint. DQ's exact diagnostic DLL6af7282a completes original MIL10–PHI9 with5,055 schema checks and two reconciled result consumers. All99 isolated files restore; the task is removed and custody releases18:46:21UTC. The external archive verifies all289 payload files (SHAfed3711773d0ac42ca7f5d3dc0712b41b26e4079caf991bb183ffa20f1259214). See live-dq-review.json and controller-parent-review-dq.json. Of82 accepted stops,79 have copied parent provenance and3 remain explicitly unknown. Nineteen reach the checked completion-predicate return:17 cached states are consistent with DONE;2 necessarily use the zero-visibility condition. This does not establish universal natural animation completion, cancellation, or new production ownership.

DR adds the exact recovered third-manager lookup prefix/helper and passes128,866 x86 controller/queue checks; no DR game has run. DS adds opt-in native_scoreboard.v3 clock.period with mandatory phase dependencies and unchanged native identity/freshness guards. Its171 Windows presentation tests,16 managed regressions and128,866 controller checks pass; native period-transition acceptance is pending. Phase11 deliberately falls back because the copied observation lacks the finished flag. See manager-third-parent-dr.json and native-period-ds-review.json.

The current isolated baseline is restored DQ; the installed harness remains DP. Allen R8 has the next coordinated GPU window, so presentation acquisition is held until its final restoration/release. All31 active inventory entries remain in scope. Full framework acceptance, missing owners, smooth content transitions and actual game resize/ultrawide remain unfinished. Prior DP active-resource Reset acceptance stands; published team1 remains unchanged.

October 29, 18:24 UTC checkpoint. Exact DP DLL7505e572 passes the focused active-resource native Reset test: fullscreen and timed windowed reversion on the same device both refresh mode metadata, generation3→5, zero host failures and2,247 schema checks. Custom scorebug rendering resumes after both resets. All99 isolated files restore; task removed and custody released18:16:36UTC. The external review ZIP verifies146 payload files (SHA4cc892a263cdbadc2edad2633d2ac2b74e633ef1a3de5686242bf5fdfd311a6e). See live-reset-dp-review.json. The continuation image still shows native scorebug fallback at the second-half transition; this is not smooth-transition or full-framework acceptance. Actual game resize/ultrawide remains unaccepted; the848-check D3D9 fixture resize is separate evidence.

DQ's frozen diagnostic and128,455 x86 controller/queue checks are ready for a bounded native parent-caller observation. Its production semantics remain unchanged; no DQ game or native completion/cancellation authority is claimed. The current isolated baseline is restored DP. All31 active inventory entries remain in scope, with coverage-dq.json recording the unresolved owners and partial visual coverage. Fresh custody/storage guards apply before the next run.

October 29, 18:09 UTC checkpoint. DP passes848 actual D3D9 checks, including effective window/mode refresh after same-device Reset and a subsequent fixture resize. Its GPU task is closed/removed and custody released18:07:57UTC; see gpu-reset-dp-review.json. Exact DP DLL7505e572 is now active in the focused native Video Settings run p74-live-reset-college-dp, acquired18:08:13UTC with a hard18:28:13UTC cap. Native DP mode readback is not yet accepted; inspect fresh custody and acceptance before acting.

DQ is a separate diagnostic-only successor: original manager prefixes and source arguments are recorded for two checked stop paths, with unknown/unreadable paths explicitly unavailable. Its Windows build,169 presentation tests,16 managed regressions and128,455 exact x86 controller/queue checks pass. No DQ game or sequence-authority claim. Frozen source SHA9f1507df2f9e147e90c62415f44e5e748b7e287bb32348526968fb13371624fe. coverage-dq.json retains all31 active inventory entries:3 scoped owner entries,5 partial scorebug entries,11 partial visual scopes and12 unavailable native owners, plus22 retained-native entries. Full acceptance remains false.

Storage: closed-runs-dq.zip preserves1,328 individually verified files from15 closed own runs, including after-game saves and private API state (3,900,487,495logical bytes; ZIP SHAe8a6198af03447767b1112d6ed3c789243bd5e084f49b109edce6a12efc326f0). Only their verified redundant Dell copies were retired; use restore_closed_runs_dq.py into a NEW directory before replay. The retirement finished18:03:37UTC with7,854,116,864bytes free. Supported LZX compression of1,801 historical own build artifacts recovered178,718,920allocated bytes with content/identity/mtime preserved; see compression-builds-dq.json. Current game, game-before baseline, current stages and peer paths remain protected. Fresh storage/custody guards still apply.

October 29, 17:52 UTC private diagnostic preparation. Current source adds parent-call provenance for the two checked native stop-manager paths. The probe verifies original manager prefixes and diagnostic stack cleanup before installing; unknown manager paths and unreadable/overflowing stack reads remain explicitly unavailable. A Windows x86 fixture executes the original prefixes through the virtual stop call and checks the actual parent return and source arguments. This is untested source preparation, not sequence authority or native acceptance; frozen DP remains unchanged for its pending Reset checks. See controller-completion-candidates-dq.json for the timeout/completion-predicate static leads.

October 29, 17:37 UTC storage and test queue. DP remains headless-only: the D3D9 fixture and focused native Reset-mode readback are waiting for Allen's final coordinated release. No presentation game or GPU fixture is active. The current isolated game remains restored DG.

Closed DN/DN2/DO saves are now preserved in closed-saves-dp.zip (SHA1fcf37ebb51faeee3e5307e1b952b5cd99368b5730aaade15d09c04fe03f2e53):93 individually hash-verified files,169,855,463bytes. Only those redundant post-run copies were retired; restore_closed_saves_dp.py restores into a new destination. Three separately verified duplicate review ZIPs were also retired from Dell,141,401,077bytes. Supported LZX compression of68 closed own database files recovered989,262,232allocated bytes, preserving content/identity/mtime and every path; see compression-closed-databases-dp-review.json for the exact narrower scope than its inherited raw-receipt description. Last compression readback17:35:57UTC had5,395,918,848bytes free. Recheck live custody and storage before testing.

Existing CX traces have85 accepted activations/85 resource-ready boundaries and82 paired stop/clear events. They have no resource-end or restart event. Their immediate callers and finite/FLT_MAX durations do not establish completion versus cancellation; controller-sequence-review-dp.json records the narrower parent-manager caller/argument investigation needed next. No production semantics were broadened.

October 29, 17:17 UTC checkpoint. Frozen DP fixes stale window/mode telemetry after a successful same-device Reset. The Windows AMD64/x86 build,169 presentation tests,16 managed regressions and127,641 exact x86 controller checks pass; its focused harness also validates. Source ZIP SHA fbeddd34acaa3ad7688da66effb045fbeec59a44c328e0034bcff3fb7c2508ca; dependency DLL SHA7505e5721242d04065c401297efb0dda2e4784db8458b1f343fa62a6bb957afb. The new actual D3D9 window-change regression and exact-DP game mode readback remain pending Allen's coordinated GPU window. The isolated game remains DG; no DP installation or GPU acceptance is implied. See framework74dp-checks.json, dp-source-difference.json and harness-dp-checks.json.

The closed DN/DN2/DO external archives are verified;119 redundant Dell raw captures/logs were retired, recovering434,913,214bytes. Small closure/restoration/release records and current game/saves/source remain. Historical raw evidence is restorable from the recorded external ZIPs; see closed-evidence-offload-dp.json. Fresh custody/storage guards still apply.

October 29, 16:56 UTC checkpoint. The focused native menu run p74-live-reset-college-do proves two successful active-resource Resets on unchanged DG DLL439f1c89: fullscreen at16:54:25UTC and native timeout reversion to windowed at16:54:40UTC. Custom graphics and gameplay resume; host generation3→5, zero host failures,1,620 schema checks. All99 isolated files restore and custody releases16:56:13UTC. The external review archive verifies all132 payload files (SHAf979bbe00e5dcca484805728618688b31c6e8273e123a6ab64300b30e5c76573); see live-reset-do-review.json. This is bounded Reset evidence at1280×720, not resize/ultrawide or full framework acceptance. DN and DN2 menu reviews timed out before applying settings; both are restored/released and independently archived.

The run exposes stale windowed telemetry during the fullscreen interval. Current local native/presentation/host3.hpp now refreshes actual swap-chain mode/window after successful Reset; that edit is not yet frozen, built or tested. Production remains frozen DG. The focused DO harness is validated and installed; it omits already-completed recovery exercises and permits780seconds for separately reviewed native-menu inputs. Allen has the next guarded GPU window. The complete31-entry inventory, sequence semantics, content-transition smoothness and resize/ultrawide remain unfinished; team1 is unchanged.

October 29, 15:51 UTC checkpoint. Exact DG production DLL439f1c89 passes reviewed original and college rendering, producer recovery and minimize/restore in DG/DM. DG finishes7–9 after overtime (4,950 schema checks); DM finishes7–5 (3,413 checks), with two reconciled consumers each. Both restore all99 baseline files and release; DM releases15:31:07UTC. Full acceptance remains withheld: DG lacks the required substitution occurrence, DM's postgame menu exploration expires, and content/context fallbacks remain. See live-dg-review.json, live-dm-review.json and per-Present trace reviews.

Static native analysis now identifies why our Video Settings menu was missing: config index35, ADJUST_VIDEO_SETTINGS, is zero in the isolated baseline, and the native menu callbacks b6fc20/b6fbe0 hide that entry at zero. DN is a tested harness-only change that enables that one byte before a guarded exact-DG in-match menu test, preserves the original config and restores the99-file baseline afterward. Windows harness/authoring/input-refusal/config tests pass; no DN game or active-resource Reset acceptance yet. The unchanged production DLL is still framework74dg.

DN storage maintenance verifies four closed external archives before retiring only redundant own raw captures/logs and ZIPs. A further12-run postgame-save archive now preserves1,205 independently hash-verified files in closed-saves-dn.zip (SHA58b6b3b8eb70d66344c110c3713bc06c5b92a5cdf07bb7cf409ceb13a42bb585). Retiring only those redundant Dell copies recovered2.8GB. restore_closed_saves_dn.py restores into a new directory; keep the manifest, verification and retirement receipts together. No current save/game/source or peer path is targeted. Fresh shared custody and the3GiB floor remain mandatory. Dell wake is solved; public team1 is unchanged. All31 inventory entries, native content/sequence semantics, smooth content changes and active-game Reset/resize/ultrawide remain active work.

October 29, 14:49 UTC checkpoint. DD passes 168 Windows presentation tests and adds immutable raster/pixel reuse. Its DE college run proves substitution visual slots and minimize/restore, but expires during native menu review before a final result. DF completes naturally MIL8–PHI6 with 3,631 schema checks, two reconciled consumers, reviewed substitution/halftime/award/sponsor imagery and zero host failures. Both restore 99 baseline files. DF releases14:32:05 UTC; its required break-wipe occurrence is absent, and content/context fallbacks remain. See live-de-review.json, live-df-review.json and trace-review-df.json.

DG removes repeated immutable shared-payload copies and reuses GPU textures only after exact pixel comparison. It keeps all identity, content, generation and freshness guards. 169 Windows presentation tests,16 managed regressions and127,641 actual x86 controller/queue checks pass. Its real D3D9 fixture passes843 checks for unchanged/changed pixels, placement, mixed reuse/upload failure, fallback and Reset; custody releases14:48:31 UTC. Production game acceptance is next. This fixture does not establish active-game Reset/resize/ultrawide.

Historical post-run save copies from12 closed presentation runs are now preserved in closed-saves-dk.zip:1,310 independently hash-verified files, external ZIP SHA332949e0bfa1584a5bc76034e8242830cd9e6b84318b5756be24de4f9a948bd6. Only those redundant Dell after-game copies were retired, recovering about2.8GB. Use restore_closed_saves_dk.py into a new directory before historical replay; manifests, metadata and retirement receipts stay together. Current game/save baselines, sources, private API databases and peer paths remain. Always recheck live storage/custody.

The complete31-entry inventory, native sequence/content semantics, smooth transitions and active-game Reset/resize/ultrawide remain unfinished. Dell wake is solved. Public team1 is unchanged.

October 29, 13:35 UTC checkpoint. DA exposed a real producer-crash recovery defect: an interrupted upload was safely withheld, but an abandoned mutex was classified as a GPU failure and sequence reset could delay a new producer. DB corrects both paths without relaxing identity, content, generation, or freshness checks. Its 164 Windows presentation tests and 16 managed-launch regressions pass. DC changes only the GPU test reader from Trace.1 to Trace.2; all production inputs match DB. The actual D3D9 fixture passes 751 checks, including abandoned-upload pixels, fallback, restart, and reset. See gpu-recovery-dc-review.json.

The exact DB original game (p74-live-recovery-original-db, DLL d42448b9aadca6730513c5bdb145488a227cb4689200f9e75c4c0c3f53f911b8) is active under a 13:42:16 UTC cap. Producer recovery passed with zero host failures; final award/sponsor review and closure are pending. Current local DD adds bounded immutable pixel reuse to reduce measured repeated pixel-copy work; it is frozen but has no Windows or game acceptance yet. Shared game and published team1 remain unchanged.

Transparent NTFS compression of closed, receipt-verified CR/CT/CV/CX/DA evidence preserves bytes, names, identities and modification times and recovered 459,920,067 allocated bytes (compression-closed-history-db.json). Rupp r10-b independently restored/released at 13:10:26 UTC. Full inventory coverage, smooth content changes, sequence semantics, and active-game Reset/resize/ultrawide remain unfinished; always recheck live custody. Dell wake is solved.

October29 12:52 UTC checkpoint. CX original diagnostic finishes naturally4–7 with4,835 schema checks; all99 baseline files restore and custody releases12:39:25UTC. Its reviewed triplet proves the corner Jordan sponsor can be hidden independently while native highlights continue. The full-screen award transition still requires replacement. Generic controller callers are recorded, but completion versus cancellation is unresolved. See analysis/presentation-framework74/live-cx-review.json and sponsor-owner-review-cx.json.

CZ passes164 Windows presentation tests,16 managed-launch regressions and127,641 actual x86 controller-hook/queue checks. It adds production award/sponsor owners, interrupted-publication protection and Reset caller receipts; no CZ production game acceptance yet. The authoring harness passes13 declared groups within8 simultaneous planes. A final harness correction is preparing reviewed postgame menu entry for production owner acceptance. The corrected rotation schedule holds bench intent through a real substitution opportunity.

The external CX archive verifies all281 files. Scoped closed CT/CV/CX raw duplicates were retired only after external archive/member verification, preserving200 files and reclaiming about779MB; see closed-evidence-offload-cx.json. Current game/save/source/peer paths are untouched. Rupp owns the next coordinated GPU window; recheck live guards. Full31 coverage, content-transition smoothness, stronger sequence semantics and actual active-device Reset/resize/ultrawide remain unfinished. Dell wake is solved; team1 is unchanged.

October29 12:02 UTC checkpoint. CT original finishes naturally11–9 with4,353 schema checks, reviewed FULL/SHORT/FOULS data and foul notices, paired advancing display captures, producer recovery and minimize/restore continuation. All99 baseline files restore and custody releases11:41:19UTC. This is partial acceptance: substitution owner512 never occurs, per-Present fallback remains, and the observed successful Reset is during startup only. See live-ct-review.json and trace-review-ct.json.

CV passes163 Windows presentation tests,16 managed regressions and127,038 actual x86 controller-hook/queue checks. It adds the independently observed break-transition root, separate Trace.2 refusal detail, and private controller boundary hooks. The exact CV college diagnostic is active under a12:15:22UTC hard cap; consult fresh custody. Two-minute halves and checked rotation intents improve event coverage. A single postgame Enter may be sent only after reviewing the exact displayed menu and recording its image/DLL identity. No native callback or sequence timing is replaced by these diagnostics.

Local next-candidate work removes redundant pixel concatenation and duplicate director observation; interrupted publication clears its sequence before any new pixel copy. Windows and native validation remain pending. Full31-entry coverage, smooth content changes, stronger controller semantics and active-device Reset/resize/ultrawide remain unfinished. Dell wake is solved; team1 remains unchanged.

October29 11:40 UTC checkpoint. CR college completes naturally5–4, passes2,534 schema checks and the copied timeout/alternating-possession comparisons, and restores99 files/releases11:12:43UTC. It is incomplete: no substitution owner512 occurred, its neutral TO/20 heading is rejected, and per-Present content/context fallbacks remain. See live-cr-review.json and trace-review-cr.json.

CS corrects the labels but its package identifier test refuses uppercase IDs before any game. CT fixes those IDs and passes162 Windows presentation tests plus16 managed regressions. The exact CT original run is active through11:45:36UTC; its minimize/restore continuation passes, but final closure and reviewed image acceptance remain pending. Consult fresh custody, not this dated note.

CU is frozen for the next headless check. It adds separate Trace.2 refusal reasons without changing replacement guards, and private native-controller write-boundary observation hooks. These hooks preserve original state writes/callbacks and have a shared x86 register/FPU preservation fixture; no native controller semantics are accepted yet. They are excluded from the production DLL.

Storage: closed-evidence-offload-cs.json verifies retirement of640 closed own raw diagnostic duplicates after complete external ZIP/member hash verification. Free space rose from3,010,535,424 to6,443,126,784 bytes. Restore historical members from each recorded external_zip into a new directory before replay. Games, saves, source, current stages and peer paths are unchanged. Dell wake remains solved; team1 is unchanged. The full31-entry framework, content-transition smoothness and native Reset/resize/wide remain unfinished.

October29 11:15 UTC continuation. CP2 original finishes naturally13–10 after overtime, passes4,342 schema checks and reviewed substitution/studio-wipe, halftime, paired GDI/DXGI and producer fallback/recovery images. All99 baseline files restore; custody releases10:49:35UTC. Its134-file external archive is fully hash-verified. Brief content/context fallbacks remain (trace-review-cp2.json); this is not smoothness or complete31-entry acceptance.

CQ/CR add separate typed ScoreFacts.1 plus native_scoreboard.v2: full/short timeout counts, qualifying fouls, applied-profile bonus and native alternating-possession team. The facts join the exact same-draw Content.2 revision without pretending to be native UI text. CR passes162 Windows presentation tests,16 managed regressions, 29 actual x86 fact checks and47 content opacity/revision checks. Its production college run p74-live-facts-college-cr is active through11:23:12UTC; consult fresh shared status. Visible facts/arrow and recovery are observed; final closure, review and substitution-banner occurrence remain pending. The current local neutral package corrects an erroneous TO/20 label to FULL/SHORT, and local native source adds Reset begin/end logging; neither is in frozen CR. Next freeze CS.

Redundant closed review ZIPs were offloaded only after their complete contents and hashes verified on the canonical external SSD;27 Dell duplicate ZIPs retired, raw run folders untouched, about797MB reclaimed. See review-zip-offload-cq.json; each ZIP is restorable from its recorded external_path. Full31 inventory, content-transition smoothness, controller lifecycle semantics and actual native Reset/resize/wide remain active work. Dell wake is solved; do not ask Brent to wake it again. Team1 remains unchanged.

October29 10:40 UTC continuation. CO college completes naturally5–2 with 1,723 schema checks, reviewed lineup/halftime/final images, paired advancing GDI/DXGI captures and producer recovery. All99 files restore and custody releases 10:19:46UTC; its86-file external archive is hash-verified. The Observation.2 shot-expiry policy passes original-x86 and rendered-pixel checks; no natural negative-shot sample occurred in CO. Content-change fallback remains open (score17, watermark22, lineup135 after-replacement frames outside the deliberate producer fault). See live-co-review.json and trace-review-co.json.

CP passes158 Windows presentation tests and adds exact native substitution visual slots plus two studio-wipe roots. No incoming/outgoing-player semantics or native sequence completion is inferred. The first CP harness omitted new capability names and refused before any game/broker; all99 files were independently verified unchanged and custody released10:33:28UTC. CP2 corrects and validates the actual preflight expression. Its same-DLL original test is active under a10:54:09UTC hard cap; consult fresh shared status before any other GPU work. Full31 coverage, scorebug parity, content-transition smoothness and real-game Reset/resize/wide acceptance remain active. Authorized wake works; do not ask Brent to wake the Dell again. Team1 remains unchanged.

Latest October29 10:10 UTC checkpoint: CL original finishes naturally12-9 with 3,718 schema checks, reviewed native lineup bindings, named halftime, advancing paired display captures and producer fallback/recovery. All99 files restore; custody releases09:47:56UTC. Its108-file external archive verifies. Per-Present trace exposes a4.219-second fallback from a native shot-clock undershoot at expiry; lineup content latency remains open. See analysis/presentation-framework74's live-cl-review.json and trace-review-cl.json.

CN passes151 Windows presentation tests and16 managed-request tests, including bounded control-pipe shutdown and reduced producer copying. No CN game was run. CO adds Observation.2 for the original formatter's bounded shot expiry(-0.1,0), with raw clocks unchanged and legacy Observation.1 strict. Exact original x86 rounding and rendered-pixel regression checks are running. Full31-entry coverage, sequence/content fidelity and real-game Reset/resize/ultrawide remain active. Authorized Dell wake works; no user wake is needed. Team1 remains unchanged.

October 28, 2026: Brent requested the complete reusable presentation framework and its validation before the ESPN scorebug. Neutral diagnostic graphics exercise the same package, data, scene and rendering interfaces intended for other authors. ESPN artwork and its exact typography/motion acceptance remain separate.

This is active implementation work. A checked source build, offscreen renderer or fixture is not native-game acceptance. The published team1 package is preserved.

Latest October29 update: draft3 dependency and Content.2 transport passes708 actual D3D9 checks in BT, including Reset/resize and producer failure/recovery. BU's corrected notice panel is reviewed in the original profile, with a natural 6–9 result,3,740 schema checks,29,880 equal independent native-content reads and zero drawing failures. All99 files restore and custody releases06:52:56UTC. See the scoped review. Remaining per-Present fallbacks, college notice verification, new native card adapters, full31-entry coverage and real-game lifecycle are active work.

Earlier October29 update: typed scene data is implemented in private package version4 and passes120 Windows presentation tests,124 core checks and52 x86 runtime checks in BC. Its corrected college capture campaign now accepts four explicit typed cards, a natural7–1 result and producer recovery; the first failed capture campaign remains preserved. Automatic native content and final cards remain open. AS/AT paired displayed-image verification resolves the earlier capture blocker. Their per-owner traces, remaining coverage failures and AU's expanded diagnostic findings are recorded in the current checkpoint. The older acceptance rows below retain their exact candidate scope; none of these additions closes all31 requested replacement/hiding entries.

Scope

The framework owns broadcast graphics and their resource/data/visibility lifecycle. Native simulation, operational menus, replay playback, cameras, audio, 3D scenes and arena artwork retain their existing owners. The 53-component inventory remains the coverage index; its 20 own-2D, eight hybrid and three hide-branding entries require explicit adapter support before a package may replace them. An unknown native owner or cue is unavailable, never inferred from a filename or phase.

Acceptance matrix

Area Required evidence Current state
Runtime packages Load different layouts/resources without recompiling; strict versions, bounds, asset hashes and paths; atomic replacement Framework74i headless pass; original-i production reload passes
Composition Text/images/alpha/clipping, layer ordering, transitions, long/unknown data, 16:9/21:9/32:9 AK passes 106 presentation tests, 112 core checks, 45 actual x86 runtime checks and 22 legacy pixel checks; accepted production game captures use 1280x720
Copied live data Same public HTTP contract as independent clients; frozen identity, clock fidelity, reconnect/stale/context guards College-m passes public identity/metadata plus optional native copied clocks; two running-play captures match native tenths, with0–32ms observation ages in1401 ready status samples
Scene/cue lifecycle Scoped monotonic intents, cancellation, overlap/priority, interruption and explicit source provenance Headless guards and five original-d manual intent receipts pass; native sequence cues unavailable
Game drawing Exact x86 production host, saved/restored D3D state, no per-frame network/file access or blocking wait AF passes 498 actual x86 HAL checks, natural college 4–5 and original 12–10 matches with zero drawing failures
Native ownership Supported layer negotiation; complete native/hidden/restored captures; no UI/gameplay/sequence damage Scorebug and game watermark have both-profile acceptance. AF college also proves team-wipe replacement and clearing while native replay playback continues. Other owners, alerts and sequence APIs remain open
Failure recovery Missing assets, invalid frames, producer exit/stall, required/optional policy, lost/reset device and resize AK HAL covers four independent owners and four deliberate draw failures with recovery, plus actual device Reset/resize. AF original verifies producer termination during an active team wipe, visible original-animation fallback, replay continuation and producer recovery. AM's real minimize/restore has no accepted image evidence
Original/college integration Real launch, gameplay/sequences, exact-match results, closure and restoration AC original 6–7/4,270 checks and college 2–3/1,842 checks pass, including real pause/resume in original. AF college 4–5/2,145 checks adds the third owner. Each delivers identical reconciled results to two consumers and restores 99 files
Author handoff Neutral example, package/transport contract, capabilities and precise verification limits Draft runtime contract below; complete replacement handoff pending native ownership and visible acceptance

No native ownership is granted by a successfully composed or uploaded image. The current per-component evidence is indexed in coverage-ak.json. It distinguishes accepted owners, partial main-scorebug coverage, unavailable owners and the retained native scope; a successful shared renderer test does not certify all53 inventory entries. AA proves the team-wipe visual boundary through before/hidden/restored images. AF subsequently accepts its production replacement in the college profile: 26 captured frames bracket a natural occurrence, with the custom card appearing only at the native boundary and clearing as native replay playback continues. The same production DLL also passes original12–10/4,436 checks. Captures show the replacement card before deliberate producer termination, the restored native animated team logo during the same wipe, then continuing native replay. Two consumers reconcile the natural result and all99 files restore before release. The capture named final.png is a late free-throw view, not final-screen evidence. The exact sequence API and authored motion remain unavailable. See ../analysis/presentation-framework74/owned-college-af-review.json and ../analysis/presentation-framework74/owned-original-af-review.json.

AK adds expanded version3 packages and the separately diagnosed backend-wipe root. It passes106 Windows presentation tests and570 actual D3D9 HAL checks, including four independent draw-failure/recovery cases. Its college run finishes 3–0 with2,535 schema checks, two reconciled consumers and the intended controls, but production visual acceptance is withheld: desktop captures remain on the initial GDI loading cover. AG's direct backbuffer captures prove its scoped backend-wipe boundary, while its ordinary screenshots show the same cover. The display/capture discrepancy is under investigation; successful draw/Present counters do not resolve it. See expanded-college-ak-review.json and diagnostic-ag-review.json in the evidence directory. AK restores99 files and releases custody at01:07:00UTC. Team1 is unchanged. The framework74s bounded probe proves complete native scorebug disappearance and restoration at its tagged drawing boundary, with other native visuals and gameplay retained. This fixed-window trial does not establish negotiated production ownership/fallback. See ../analysis/presentation-framework74/diagnostic-s-review.json. Required unsupported ownership refuses admission; optional failures retain native visibility. Graphics clocks never advance authoritative game clocks. Any required clean-view mode must separately prove its reveal/fallback lifecycle.

Evidence and immutable build/run identities belong in analysis/presentation-framework74/. Builds and executable tests run on the Dell. Game/GPU work needs a fresh atomic shared lease and exact closure/preservation.

Runtime authoring contract (private draft1)

framework/make_diagnostic.py creates a neutral eight-layer package. It uses an explicit supplied font and a small alpha-test PNG; this is not ESPN artwork. The package folder contains framework.json, its SHA-bound layout.json, and SHA-bound TTF/OTF/PNG resources. Authors can change the layout and resource files without recompiling the DLL, updating their manifest hashes together.

The manifest format is openhoops.presentation.framework, version1, with exactly format, version, layout, layout_sha256, resources, ownership and scenes. The loader rejects duplicate keys, unknown fields/versions, symlinks, escaping resource paths, mismatched hashes and missing declared resources. Limits are 128 nodes, eight layers, 32 scenes and 64 resources; each resource is at most32MiB and combined resources at most128MiB. Layout authoring remains the validated draft contract in authoring/compile_package.py.

Each scene specifies id, layers, trigger, priority, exclusive_group and duration_ms. Current triggers are manual, observed_scorebug and always_with_valid_data. A manual scene has a finite lifetime up to30seconds. Exclusive groups choose the highest-priority eligible scene, with later intent and stable scene ID breaking ties. Native replay/award/wipe triggers refuse admission because the adapter has not recovered their sequence authority.

The independent Python worker reads the public HTTP snapshot and match endpoints. It requires matching run/match/fingerprint, a fresh advancing source sequence, and a SHA-verified frozen roster context whose roster and metadata IDs match the snapshot reference. It copies values; it never advances game clocks. Disconnect, source stall, identity change and expiry clear pending graphics.

This HTTP route is not synchronized to the native display frame. The current native API publishes at a100ms minimum interval, the broker polls at50ms and the presentation source polls at50ms plus request time. Original-i's gameplay capture shows a24.8-second copied clock while the native graphic shows24.6. That is an observed transport delay, not a clock extrapolation or frame-accurate clock acceptance. Framework74m's optional copied-clock path subsequently passes actual college/overtime acceptance. Running-play captures show30.1 and19.0 on both the native and framework game clocks;1401 ready status observations use local copies aged0–32ms (median15ms). This establishes improvement in the tested run, not exact synchronization of every native display frame.

The draft channel's original72 reserved bytes at offset184 now carry an optional version1 Observation (see native/presentation/transport.hpp). An independent 32-bit odd/even seqlock protects one game-thread writer and copied client reads. Version0 means unavailable, preserving older hosts. The record includes native run nonce/match epoch, capture tick, public sample number, phase/period, validity, pause/replay/simulation flags, scores and two clocks. The adapter captures from its established application-dispatch entry without advancing the public sample counter or running rule/command handlers more frequently. Executable addresses remain private to the adapter; clients use the documented copied record.

The worker accepts observations at most100ms old only when their run, native epoch, phase, period and scores agree with the fresh public state. Pause, replay, invalid state or a new identity/transition hides obsolete graphics. An unavailable or old observation retains the HTTP path. Neither path extrapolates a clock or refreshes public metadata/cue expiry. clock_source and observation age identify the selected source in the worker status. This is an asynchronous observation, not a native rule command, completion cue or claim of exact frame synchronization.

The worker's owning process sends newline-delimited JSON on stdin (64KiB per line; 32 queued requests). A manual request is {"type":"cue","request":{"run_id":"...","match_id":"...","epoch":1,"source_sequence":123,"intent_sequence":1,"scene":"card","show":true}}. Use the current identities emitted by the worker's status stream. The source revision must have been observed within750ms and remain fresh; intent sequences must increase. show:false cancels that scene. These are presentation intents, not game commands or native sequence completion notifications.

To reload the same package root, send {"type":"reload","request":{"expected_package_sha256":"..."}} using the worker's current manifest digest. The loader validates the complete replacement and creates new native resources before committing. Failure retains the old renderer and its private hash-checked resource copies. Success clears old cues.

Launch the managed game with a fresh random32-hex OPENHOOPS_PRESENTATION_CHANNEL, then run presentation.framework.worker with --library, --package, --channel, --game-pid, --api-url and --token-file (optional --labels). The game creates the mapping; the worker only opens it. The mapping is tied to the exact host PID/creation time and admits one producer. Its draft1 layout is in native/presentation/transport.hpp and framework/channel.py. It carries bounded premultiplied BGRA pixels and a deadline, not executable code, API credentials or game addresses. The game never waits on a busy producer; stale/invalid frames are discarded. Game-side rendering saves/restores D3D state, targets and depth surfaces.

Draft1 hosts expose no native ownership. Required ownership causes admission failure in that version. Draft2's separately negotiated scorebug capability is described below. Optional unsupported ownership retains native visuals. Frame.ready, an accepted cue, a published pixel frame and a successful D3D call each describe only their own stage; none establishes visible replacement or native suppression.

Evidence and open gates

The precise original-d review is original-d-review.json. Its raw acceptance.json reports the automated data/control/result checks; the separate review explicitly rejects complete presentation acceptance after direct inspection of all owned-window screenshots. No ESPN asset or public release was changed. The game, renderer, broker and owned tasks closed; all99 mutable files in the independent game were restored and custody released.

Framework74g/h add Present/window diagnostics and separately build a private probe DLL. The probe is not the production artifact: it observes native UI draw dispatch without suppressing anything, and takes three bounded before/after backbuffer readbacks. p74-live-diagnostic-h completes its bounded 210-second observation with 1,288 schema observations, five accepted manual intents, package reload, producer recovery and zero host failures. Directly inspected backbuffers and owned-window gameplay show the neutral framework and correctly copied scores/clocks. Present reports success to the expected game window.

The earlier startup-cover failure did not reproduce; its cause is not established. This diagnostic was cancelled deliberately and supplies no natural final result or production-DLL acceptance. All99 mutable files were restored, both owned tasks removed and custody released at17:11:02.044239UTC. See diagnostic-h-review.json. Production p74-live-original-i subsequently repeats visible output without any probe/readback hooks, using DLL SHA256 735e144d807060f2daa45316121eccb8f19bf008068e5b27ea5c1a8c6e86c07f. It finishes naturally away10-home2, delivers identical reconciled results to two independent consumers, checks3,798 public snapshots, accepts five manual cues, reloads the package and visibly recovers from producer exit. Host failures remain zero. The final native summary agrees with the results. All99 mutable files are restored, both owned tasks removed and custody released at17:41:44.269676UTC. See original-i-review.json. This accepts the tested production drawing path; its copied-clock delay remains explicit above. The next native work must establish safe visual owners/cues and complete exact-candidate original/college and synchronization checks. The complete framework is not ready for ESPN production.

Capability boundaries for current authors

Version3's31 declarations are package capacity, not31 accepted native owners. The current internal adapter can negotiate only the following named boundaries:

Capability Modes Native evidence
scoreboard.main replace Both profiles; individual alert and indicator content remains incomplete
branding.game_watermark hide, replace Both profiles, independent from scorebug visibility
transitions.team_wipes replace Both profiles; original active-wipe producer failure visibly restores the native animation
transitions.backend_wipe replace Private one-frame visual boundary and AK HAL checks; production visual acceptance pending

The backend capability covers only the hash-checked 2kbackendwipe.iff root. It does not include every replay transition. No capability offers authoritative sequence-start, completion or duration events. Packages must preserve required versus optional admission rules: required unsupported ownership fails admission; optional unsupported ownership remains native and contributes no replacement pixels. The main scorebug boundary includes its original attached notices, so a package must not claim full informational parity until those data bindings are implemented and checked.

Production p74-live-college-m finishes naturally away6-home8 after two60-second halves and one60-second overtime. It uses framework74m/DLL 56f880e886eae3108cce24d42c6e090d786a42bb1865f941749ef8dd70f04199, without probe hooks. Both independent consumers receive identical reconciled results,3112 snapshot schemas pass, five manual intents/reload/producer recovery pass, and drawing failures remain zero. Captures verify second-half/OT labels, matching game-clock tenths, producer expiry/recovery and the native final score. All99 mutable files restore, six owned tasks are removed and custody releases at 18:54:39.715249UTC. See college-m-review.json.

The earlier college-i attempt stopped before gameplay when Windows denied a managed-request receipt replacement. M adds bounded file-operation retries and passes15 request tests, including actual held-reader and concurrent reader/writer cases. Receipt persistence never retries a native Start command. These private changes do not modify the published team1 package or grant native visual ownership.

Framework74n exposed a remaining intermittent Windows CRT read error: errno13 without a winerror attribute. Framework74o handles that form with the same bounded retry, passes16 managed-request tests and the unchanged76 presentation tests. Its first diagnostic stops before gameplay on a test-harness expiry assertion; no native acceptance follows. The harness timing correction and a repeat observation are tracked in diagnostic-o-review.json.

The corrected O2 diagnostic subsequently completes all five cues, reload and producer expiry/recovery. P connects the known scorebug owner to 6,216 deferred draw records on a separate rendering thread, with zero unparsed spans. Both paired gameplay captures retain native visuals and show matching 57.0 clocks. See diagnostic-o2-review.json and diagnostic-p-review.json. These deliberately cancelled probes do not grant native ownership or production acceptance. The next private Q probe passes 70 actual x86 marker roundtrip checks and the existing headless suite. It reversibly marks scorebug command counts and restores their exact values before D3D drawing; all native draws still forward. Its native proof remains pending. Native suppression, visibility fallback and sequence cues are still open gates.

Diagnostic-j subsequently observes the known scorebug UI root during actual gameplay, but none of its nested calls reach the four intercepted D3D drawing methods or the observed vertex commit routine. The native scorebug remains visible in the paired backbuffer. Therefore this UI traversal is not yet a verified suppression boundary. The observer preserves all original calls. The bounded run completes1,292 schema observations, five manual intents, reload and producer recovery with zero host failures, then closes/restores/releases at 18:04:32.854650UTC. See diagnostic-j-review.json.

Independent component transport (private draft2 candidate)

Framework74w passes85 presentation tests and61 x86 plane-contract checks, including a real64-bit producer/x86 mapping reader and delayed-device startup. The original draft1 remains available. Draft2 passes260 actual x86 D3D9 HAL checks and original-profile production main-scorebug replacement/fallback in a natural14-9 match. See ../analysis/presentation-framework74/gpu-w-review.json and ../analysis/presentation-framework74/owned-w-review.json. The latter's 5,009 snapshot checks, two reconciled result consumers, five manual cues, reload, worker recovery and zero host failures are scoped to that exact W DLL. All99 independent-game baseline files restore and custody releases21:23:20UTC. Same-W college finishes7-5 with2,221 schema checks, two reconciled consumers, five cues/reload/recovery and zero adapter/host faults. Eight reviewed captures show first/second-half labels, matching clock readings, full scorebug replacement and native fallback/recovery. See../analysis/presentation-framework74/owned-college-w-review.json. All99 files restore and custody releases21:42:47UTC. Continuous native visibility through ordinary sequences, other native owners and exact sequence cues remain open. No public package or stable hook is promoted by this checkpoint.

Manifest version2 keeps the existing top-level keys. Each ownership entry adds mode (replace or hide) and a nonempty layers array to component and required. Components cannot share owned layers. Unassigned layers compose an independent overlay. Unsupported required components refuse admission; optional unsupported components' assigned layers stay hidden while native work remains. At most seven declared native components and one unowned overlay fit this draft. All component runtimes preload before a package reload commits; a failed reload retains every prior runtime. Combined duplicated resources remain within128MiB.

The historical draft2 ceiling and its full-inventory gap are recorded in ../analysis/presentation-framework74/capacity-audit-af.json. Version3 below separates declared package capacity from concurrently prepared surfaces. No extra native owner becomes supported merely by increasing that capacity.

Channel version2 uses Local\OpenHoops.Presentation.2.<token>, a640-byte header, eight bounded plane descriptors and32MiB combined premultiplied pixels. Each plane has a positive canvas-contained rectangle, exact contiguous payload range, and either one negotiated owner bit or zero for the unowned overlay. Replace bits exactly equal supplied owned planes; hide and replace masks cannot overlap. The game validates capabilities against private values, never shared fields that a producer could forge. The worker's --transport 2 accepts manifest version2, 3 or4; version4 additionally declares typed scene content. No bearer token, executable address or command authority travels in this channel.

Expanded package declarations (private version3)

AJ passes105 Windows presentation tests,1,438 x86 command-marker checks and the unchanged45 x86 runtime-pixel/107 plane-contract checks. Version3 retains the manifest keys and ownership fields but permits31 native component declarations, 256 global layers,4,096 total nodes and256 scenes. The layout file is bounded to4MiB. Each component and the unowned overlay still has at most8 local layers and128 nodes. Loading remaps each group's global layers to local0–7 before the unchanged native software renderer sees them. All runtimes and maps commit together only after preload succeeds. Resource limits remain unchanged.

The unchanged channel permits at most8 simultaneously ready planes and32MiB of combined pixels. Since BF, a package with at most8 negotiated groups prepares all eligible components before their first native appearance. Larger packages use the previous-Present visual mask to select work; missing/invalid observations prepare no owned planes for that larger inventory. This is scheduling, never a sequence cue or suppression permission. The host still checks the actual native owner, capability, current match context and freshness at its draw call.

The producer selects whole ready planes by priority within both limits. Omitted planes are reported in capacity_fallbacks and keep their original native drawing; the producer continues. The host independently rejects malformed or over-budget publications. Optional unavailable owners remain native even when their package layers and scenes are valid. BF's preparation behavior has121 Windows test coverage; live acceptance remains pending.

Private count markers retain the original seven encodings byte for byte; reserved bytes e1–f8 encode owner bits7–30. Bit31 remains unavailable. This is private adapter transport and does not assign public capability names or recover new native roots. AJ passes498 HAL checks; AK's four-owner successor passes570. Exact-game version3 visual acceptance remains pending the display/capture discrepancy described above, despite AK's passing automated college result.

The device identity stays zero while device hooks are still initializing. After all supported hooks are installed, the adapter publishes the actual device identity as its capability-readiness boundary. The draft2 worker waits at most 45seconds for this boundary, checking that the exact game process remains alive. Only then does it accept or refuse required components. A ready device with a zero capability mask still refuses required replacement. This startup wait performs no game/render-thread work and never changes native visibility.

A replacement frame carries the exact copied native run nonce, epoch, public sample, phase, period, score, live flags and capture tick used by the composer. At the actual native draw, the host compares a coherent private observation and requires both observations to be at most100ms old. A label/card plane being ready cannot authorize omission of a missing scorebug plane. Busy/abandoned/expired or invalid frames, changed identity and failed drawing retain the native draw.

Current replacement capabilities are scoreboard.main (bit1), branding.game_watermark (bit2) and transitions.team_wipes (bit4). The watermark also has a separately negotiated hide capability. Each requires original code bytes, exact controller/cache identity and its original resource hash (newscorebug.iff, gooeyoverlaysstatic.iff or teamwipe.iff). One unavailable resource does not grant or revoke another component. W/Z/AC establish the scorebug and watermark; AF establishes the team wipe in the college profile. The adapter tags owned deferred counts before publication and restores those counts at the known D3D interpreter callers. It executes the original dispatcher, pre/post callbacks, updates, allocations and resource/state commands. A component's custom pixels must successfully draw on the primary backbuffer inside the existing native scene before its native pixel calls can be omitted for that frame. Complete D3D state/targets are restored, with one consistent component decision per Present. Unowned overlays draw at Present. The observed native_visual_mask describes prior-frame native drawing; it is not a replay/wipe entry or completion cue.

Framework74y adds an opt-in draft2 native_draw scene trigger. Every scene layer must belong to a declared native component; draft1 and mixed/unowned scenes refuse it. It makes a component eligible while the actual native draw remains the visibility boundary. It supplies no sequence entry, completion or cancellation authority. Ordinary overlays retain conservative visibility and disappear immediately on a copied pause/replay boundary.

In this candidate, valid active simulation observations may include pause/replay bits. The composed frame and current native observation must have identical flags as well as the existing run/epoch/phase/period/score and100ms freshness checks. A previously composed ordinary-play frame cannot cross a new pause state. Draft1 copied clocks keep their stricter policy. Y passes90 presentation tests, 107 native plane checks and the existing headless regressions. Its actual x86 D3D9 HAL fixture passes326 checks, including matching pause/replay native-boundary pixels, and releases21:43:39UTC. See../analysis/presentation-framework74/gpu-y-review.json. Game acceptance remains pending; W acceptance does not certify these changes.

Z combines that policy with the independent watermark owner.92 presentation tests,238 count-marker checks,107 plane checks and existing regressions pass. Its actual x86 D3D9 HAL fixture passes422 checks: both owners replace independently, watermark hiding remains independent of missing scorebug pixels, stale contexts retain native drawing, failed drawing retains undrawn owners, and fresh frames recover both. Two deliberately injected device-method failures are expected; this is not a driver-outage claim. See../analysis/presentation-framework74/gpu-z-review.json. Original Z production acceptance finishes naturally 13–8 with 3,716 schema checks and identical reconciled results for two independent consumers. Nine reviewed captures show complete main-scorebug replacement, independent watermark hiding, native fallback and recovery. Host drawing failures remain zero; all 99 independent-game files restore and custody releases at 22:10:29 UTC. See ../analysis/presentation-framework74/owned-z-review.json. Same-Z college finishes naturally6–1 with2,428 schema checks, two identical reconciled consumers, five manual cues and producer recovery. Ten reviewed captures show independent watermark replacement, first/second-half labels and native fallback. Its1,503 ready copied-clock observations are0–32ms old (median15ms). All99 files restore and custody releases22:34:41UTC. See ../analysis/presentation-framework74/owned-college-z-review.json.

Z still shows a copied shot-clock value when the stock graphic hides it. Static original-formatter evidence establishes that shot>=game or game<=0 suppresses that field, explaining the recorded original11.2/25 and college 13.1/31 pairs. The source correction applies this comparison to both public and fresh local clock pairs, using an inapplicable optional binding while keeping the raw API/native context unchanged. AC passes95 Windows presentation tests and 45 actual x86 runtime pixel checks, proving the optional field can disappear without removing the scorebug. All108 native source inputs match Z. Game verification of the correction remains pending; see headless-ac-review.json in the framework74 evidence directory. Other native formatter predicates and alternate-clock selection are unresolved; this does not establish a free-throw-specific hide rule or complete display parity. Actual manual-pause gameplay and stock alert flyouts remain unaccepted.

This candidate has not accepted every scorebug flyout, every native inventory entry, native sequence timing or every display frame's clock. Those remain explicit game/integration gates. ESPN fonts, assets, geometry and motion fidelity remain separate from the neutral framework tests.