Highlights below: 8 clips around score locks that are in the signed receipt, plus stills from each moment. Full footage stays on the player's PC. Some stills are redacted copies to hide account names. No on-screen names or controller timelines are on this page.
The verdict is math anyone can re-run. The QorTime verifier checks the signed receipt in your browser, with the referee's public key pinned.
Do it from a terminal
- Download receipt.json. Its sha256 is c35f4fd739b25cc61a82e98afb7addb4ed7467e40a0fb8cbe68f0443fdd7579c.
- Run the open verifier with the referee key pinned. It should print VERIFIED. Without the pin it prints PARTIAL (referee_key_not_pinned), which is expected.
- Look the receipt up in the referee's public log. The logged body_sha256 is SHA-256 of the receipt's canonical JSON without signatures.
git clone https://github.com/ConWan30/qortime && cd qortime && pip install -e . qortime verify receipt.json --referee-pub b283cffb9279f2f90859268c64ef21861ef7e5d9fc773bdb8061af0e80624889 curl https://qortime-referee-production.up.railway.app/v1/log/inclusion/8a3a6d60e498c607fef79e013afab333b760fb7c9e58e16fa9912683c98cb71b
AI explains, it never decides. Meta Muse Spark (muse-spark-1.3) was given only the verifier's result and allowlisted facts. A deterministic guardrail checked the text before it was kept. The verdict above comes from the verifier, not from this text. Signed with the separate AI advisory key 2d5de84a3dcdeaec, which cannot make or change a verdict.
VERIFIED
This Madden27 single match on PS5 for home 21 away 7 in Q4 was confirmed from the end menu re-read, which matches the claimed final.
What passed: bundle structure, capture chain, challenges, final re-read, input integrity, motion, and video bytes. Referee qortime-online-s2 is trusted. No errors and nothing missing.
Video was spot-check: 270 of 878 segments checked.
Controller input advisory did not pass as human proof, with alignment at chance level. It is advisory only.
NOT proven: who was holding the controller, that laptop countersignatures were made live, that footage was not fabricated before QorCap hashed it except where anchored, or content of unchecked video segments.
The referee re-reads the end of the recording itself. Normally it finds a fourth-quarter scorebug in the last part of the video and checks the end-of-game menu against that read.
In this match that normal path did not apply. The referee looked at 486 frames at the end of the recording and found no fourth-quarter scorebug among them: all 486 were end-of-game menu frames. It recorded this as “no Q4 scorebug read before the end-of-game menu”.
So the referee used its end-menu fallback. It read the end-of-game menu and checked it against the last locked score in the bundle's signed timeline: away 7, home 21, Q4. They match, so the final re-read check passed. The referee's own note, word for word:
What this means: the final comes from the end-of-game menu plus the last scorebug lock in the signed chain. It does not come from a fresh fourth-quarter scorebug read at the very end of the video.
During the match the referee sent 3 official challenges (kickoff, halftime, end): a short button pattern to enter on the pause menu. The referee checked all three itself (rule 2): latencies 1464 ms, 1486 ms, 1550 ms.
Challenge Echo is a separate, information-only cross-check. Meta SAM 3.1 tracks the pause-menu highlight in the video, and deterministic code checks that each logged D-pad press shows up on screen. It never raises a status and never changes the verdict. At most it could suggest lowering VERIFIED to PARTIAL. Here it found all three challenges consistent.
| Challenge | Echo | Median press-to-screen |
|---|---|---|
| kickoff | consistent | 1637 ms |
| halftime | consistent | 1626 ms |
| end | consistent | 1662 ms |
3 of 3 consistent. Press-to-screen is how long a D-pad press took to show up in the captured video. That is a different measure from the referee's challenge latency. Echo scores UP and DOWN moves and checks that LEFT and RIGHT leave the highlight still. It does not score SQUARE. Run with qortime-ai Challenge Echo after PR #7 (menu-aware wrap test). Echo's per-press files are not published.
Show Your Hands checks that on-screen play starts have a controller press behind them. It is lower-only: it can lower VERIFIED to PARTIAL, never raise a status or fail a match. No Muse Spark play-start filter was run for this match, so no play start was confirmed and nothing was scored. hands.json holds counts and per-clip summary fields only, no per-frame controller data, and is bound to this receipt (receipt_jcs_sha256 95b9c9961c9d…). It is not signed with the AI advisory key yet, so the verifier does not show it as a panel. Controller signals never show who was holding the controller.
- kickoff0–0Q1 · 0:31 in the recording · 10 frames
- touchdown (+6, try not yet on the scorebug)0–6Q1 · 4:13 in the recording · 18 frames · clip hl00
- extra point (PAT kick, +1)0–7Q1 · 4:26 in the recording · 10 frames · clip hl01
- quarter change · Q20–7Q2 · 4:36 in the recording · 10 frames
- touchdown (+6, try not yet on the scorebug)6–7Q2 · 7:11 in the recording · 18 frames · clip hl02
- extra point (PAT kick, +1)7–7Q2 · 7:22 in the recording · 10 frames · clip hl03
- touchdown (+6, try not yet on the scorebug)7–13Q2 · 10:17 in the recording · 18 frames · clip hl04
- extra point (PAT kick, +1)7–14Q2 · 10:28 in the recording · 10 frames · clip hl05
- quarter change · Q37–14Q3 · 10:53 in the recording · 10 frames
- touchdown + extra point (+7; the +6 state was never read)7–21Q3 · 11:29 in the recording · 18 frames · clip hl07
- quarter change · Q47–21Q4 · 16:24 in the recording · 10 frames
- end of game7–21Q4 · 18:52 in the recording · 12 frames · the final was confirmed by the referee from the end-of-game menu (see above)
