# Landscape, November 2026: where this work sits in the literature, and the papers to write

*Provenance.* Written 2026-11-01 by the head seat as a **claim** (η 0.25): a synthesis over twelve research notes (in `docs/papers/landscape-2026-11/notes/`), each a sweep of one literature with every citation checked against arXiv, Crossref, OpenAlex, Semantic Scholar or a publisher page, and flagged UNVERIFIED where it could not be. The shared web-search budget ran out part-way through every sweep, so each "absent" below means "not found among hubs, cited-by graphs and index queries", not proof of absence. Numbers about the owner's own systems come only from the evidence dossier used to write the paper scaffolds and carry its tiers (PINNED / MEASURED / DESIGNED / DENIED / DARK); no number from any earlier home deployment is used. Nothing here is submitted anywhere; Luke decides.

The twelve paper scaffolds this map recommends live beside it as `docs/papers/<slug>.md` (see `docs/papers/README.md`), each served at `/chartroom/<id>`.

*Second sweep.* On 2026-11-02, with web search restored, every note was re-opened: UNVERIFIED items resolved, refused searches run, each gap re-graded. The result is the addendum at the end of this file and the `Gap-fill 2026-11-02` section of every note. A thirteenth note and scaffold (observed semantic information for deployed agents) were added.

*Home.* Moved into the Broadcast's paper section on 2026-11-02 by owner direction; the text as committed is served at /chartroom with its hash. Pull-request numbers quoted below refer to the semantica3 repository, where the program was first filed.

---

# Honest losses open twelve publishable gaps

**Bottom line.** Across the eleven literatures that surround the author's record-and-belief work, the notes find the components everywhere and the composition nowhere: no indexed paper pairs an append-only classical log with a density-matrix belief as its derived view, none reports a Belavkin filter *losing* a prequential contest to a blend and to persistence on real data, none states a three-state freshness algebra with a proven never-promote rule, none certifies *non*-learning by an empty content-addressed log, none uses geometric phase as a process clock, none ships a game whose fields are exact GKSL density matrices with a public fidelity ledger, and none tiers agent speech by evidence kind in a repository. The two field reviews a referee will reach for first say so almost verbatim: the 2022 quantum-games review contains zero occurrences of "density matrix" or "Lindblad" and asks for "a proper and comprehensive evaluation and assessment scheme" ([Seskir et al. 2022](https://arxiv.org/abs/2202.07756)); the 2026 AJP Resource Letter says open systems and density matrices are "not discussed as teaching topics" and most tools are "described without outcome data" ([Meyer et al. 2026](https://arxiv.org/html/2606.06445v1)). The program's strongest asset is therefore not any single mechanism but its published DENIED tier: a field that rewards positive quantum-inspired results and confident agent self-reports has almost no measured negatives, and the owner's ledger has several. The constraint that shapes everything is the calendar: three of the twelve papers have verified deadlines in the next three weeks (CLeaR 2026-11-16, the AAAI-27 workshop slate 2026-11-20, all AoE), and every OpenReview venue needs a profile that can sit in moderation "up to 2 weeks" on a personal email ([OpenReview](https://docs.openreview.net/getting-started/creating-an-openreview-profile/signing-up-for-openreview)). The recommended order is P12 → P5 → P6 → P7 → P2 → P3 → P8 → P1 → P9 → P10 → P4 → P11, with four of PLAN.md's venue choices changed because their deadlines have passed, their APCs have no waiver route, or they require measurements the ledger marks DARK.

## Eleven literatures, each with a small stable hub set

The notes converge on a reassuring fact: the hubs are few and stable. A reviewer in any one cluster expects roughly five to ten canonical citations plus the 2024–2026 frontier, and the table below gives the set each paper must carry. Crossref cited-by counts in the record-architecture note identify the heaviest anchors overall: Kalman 1960 (24k), GRADE 2008 (21k), Rubin 1976 (8k), Endsley 1995 (6.9k), Lamport 1978 (6.3k), Kaelbling–Littman–Cassandra 1998 (2.7k), AGM 1985 (2.1k) ([record note sources](https://doi.org/10.1115/1.3662552)).

| Cluster | Hubs a paper must cite (author, year, venue) | Frontier 2024–2026 |
|---|---|---|
| Record vs belief architectures | Lamport 1978 CACM; Chandy–Lamport 1985 TOCS; Helland 2015 CIDR ("The truth is the log. The database is a cache of a subset of the log." — [PDF](https://www.cidrdb.org/cidr2015/Papers/CIDR15_Paper16.pdf)); Fowler 2005 Event Sourcing; Young 2010 CQRS; Kleppmann 2015/2017; Åström 1965 JMAA; Kaelbling–Littman–Cassandra 1998 AIJ; Kalman 1960; Thrun–Burgard–Fox 2005; AGM 1985 JSL; Katsuno–Mendelzon 1991; Friedman–Halpern 1999 JAIR; Baltag–Smets 2006 MSCS (quantum DEL); Kapteyn–Pretorius–Willcox 2021 Nat. Comput. Sci. (PGM digital twin) | Veresov et al. ECOOP 2024 "failure transparency" ([DOI](https://doi.org/10.4230/LIPIcs.ECOOP.2024.42)) — the only formal snapshot+replay ≡ failure-free criterion; Quinn & Alvaro CACM 2025 record/replay; Hou et al. ACM REP 2026 survey; DreamerV3 in Nature 2025 |
| Quantum probability as belief | Busemeyer & Bruza 2012 CUP (2nd ed. 2024); Pothos & Busemeyer 2013 BBS, 2022 Annu. Rev. Psychol.; Khrennikov 2010 Springer, 2023 Entropy and Springer book; Caves–Fuchs–Schack 2002 PRA; Fuchs–Schack 2013; Leifer–Spekkens 2013; Coecke–Spekkens 2012 Synthese; Bouten–van Handel–James 2007 SICON; Monras–Beige–Wiesner 2011 (HQMM); Srinivasan–Gordon–Boots AISTATS 2018; van Rijsbergen 2004; Widdows–Kitto–Cohen 2021 JAIR | Asano & Khrennikov 2026 GKSL cognition ([arXiv](https://arxiv.org/abs/2604.18643)); Chu 2026 density-matrix opinion dynamics ([arXiv](https://arxiv.org/abs/2607.01452)); Parzygnat–Fullwood 2023 PRX Quantum and Bai–Buscemi–Scarani 2025 PRL (Petz map as the unique minimal-change quantum Bayes rule — [arXiv](https://arxiv.org/abs/2410.00319v3)); Mannone et al. 2026 RAS swarm mixed states; Lanza et al. ISER 2020 robot perception |
| Quantum filtering as estimator | Belavkin 1992 JMVA; Bouten–van Handel–James 2007 SICON; Wiseman–Milburn 1993 PRA / 2009 CUP; Jacobs & Steck 2006 Contemp. Phys. (eq. 91: η multiplies only the information term — [PDF](https://steck.us/papers/straightforward.pdf)); Wonham 1964 SIAM; Kushner 1967 TAC; Särkkä 2013; Sinopoli et al. 2004 TAC; Kubo 1962; Plankensteiner–Hotter–Ritsch 2022 Quantum; Dawid 1984 JRSS-A; Gneiting–Raftery 2007 JASA; Makridakis et al. 2018 PLOS ONE; Mehra 1970 TAC; Dawid–Skene 1979 | Kerber–Ritsch–Ostermann 2025/26 "no general criterion" for cumulant validity ([arXiv](https://arxiv.org/abs/2511.20115)); Ning 2025 HQMM fragility; Zeng et al. AAAI 2023 (linear beats Transformers); Veliz-Cuba–Kilpatrick–Josić SIAM Review 2016 (saturating certainty) |
| Missing data, staleness, calibration | Codd 1979 TODS / 1986 SIGMOD Rec.; Zaniolo 1984; Imieliński–Lipski 1984 JACM; Libkin 2016 TODS; Belnap 1977; Rubin 1976 Biometrika; Che et al. 2018 Sci. Rep. (GRU-D); Kaul–Yates–Gruteser INFOCOM 2012; Yates et al. 2021 JSAC; Sun–Polyanskiy–Uysal 2020 TIT; Ornee–Sun 2021 ToN; Maatouk et al. 2020 ToN (AoII); Gneiting–Balabdaoui–Raftery 2007 JRSS-B; Hyndman–Koehler 2006 IJF (MASE); Song & Szafir 2019 TVCG; Denning 1976 CACM; Green–Karvounarakis–Tannen PODS 2007; MacKenzie et al. 2002 Ecology | Shisher & Sun MobiHoc 2022 (error non-monotone in age off-Markov — [arXiv](https://arxiv.org/abs/2208.06948)); Chiariotti et al. 2022 TCOM (Query AoI); Razniewski et al. 2024 CSUR (completeness assertions); Alsufyani–Forshaw–Fernstad 2024 survey; Błasiok et al. STOC 2023 |
| Causal self-confounding | Perdomo et al. ICML 2020 ([arXiv](https://arxiv.org/abs/2002.06673)); Hardt & Mendler-Dünner survey (Stat. Sci.); Merton 1948; Lucas 1976; Ensign et al. FAT* 2018; Chaney et al. RecSys 2018; Bottou et al. JMLR 2013; Schnabel et al. ICML 2016; Pearl 2009; Robins 1986; Hernán & Robins ch. 20; Gustavsson–Ljung–Söderström Automatica 1977; Feldbaum 1960; Bar-Shalom–Tse 1974 TAC; Ortega et al. 2021 (delusions); Rezende et al. 2020; de Haan et al. NeurIPS 2019; Mozer 2004; Yang & Newman UbiComp 2013; IoTSafe NDSS 2021; Sondhi 1967 BSTJ; Benesty et al. 2000; Crapse & Sommer 2008 Nat. Rev. Neurosci.; Kruskal 1977; Allman et al. 2009 | Cai–Han–Cao NeurIPS 2024 performative control; Boeken–Zoeter–Mooij CLeaR 2024 and NeurIPS 2025; Mendler-Dünner et al. NeurIPS 2024 (performative power measured); Perdomo ICML 2025; Richens & Everitt ICLR 2024; Pan et al. ICML 2024; Shumailov et al. Nature 2024 |
| Gauge, geometric phase, certificates | Berry 1984; Simon 1983 PRL; Aharonov–Anandan 1987 PRL; Samuel–Bhandari 1988 PRL; Mukunda–Simon 1993 Ann. Phys.; Hannay 1985 J. Phys. A; Sinitsyn–Nemenman 2007 EPL/PRL; Sinitsyn 2009 J. Phys. A review; Sarandy–Lidar 2006 PRA; Wegner 1971; Kogut 1979 RMP; Kitaev 2003; Harary 1953; Zaslavsky 1982; Merkle 1988; RFC 6962/9162; RFC 8785; McKay–Piperno 2014; Jia et al. S&P 2021 (PoL); Thudi et al. USENIX Sec 2022 (forging); Crooks 2007 PRL; Clark 1973 Econometrica | Abbaszadeh et al. CCS 2024 zkPoT; Srivastava–Arora–Boneh NeurIPS 2024; Eisenhofer et al. SaTML 2025; Xue et al. CSUR 2026; Gao–Liang–Guestrin ICLR 2025 (model equality testing); Zhao–Walters–Yu TMLR 2026 (parameter-space symmetry — [arXiv](https://arxiv.org/abs/2506.13018)); OpenSSF model signing v1.0 2025 |
| Higher-order information | Watanabe 1960; McGill 1954; Rosas et al. 2019 PRE (O-information — [arXiv](https://arxiv.org/abs/1902.11239)); Williams & Beer 2010; Lizier et al. 2018 Entropy; Meilă 2007; Schneidman et al. 2003 PRL; Amari 2001 TIT; Battiston et al. 2021 Nat. Phys.; Rosas et al. 2022 Nat. Phys.; Cerf–Adami 1997 PRL; Horodecki–Oppenheim–Winter 2005 Nature; Linden–Popescu–Wootters 2002 PRL; Walck–Lyons 2008 PRL; Zurek 2009 Nat. Phys.; Brandão–Piani–Horodecki 2015; Rajagopal–Sudarshan 1974 PRA; Barrett 2015 PRE | Robiglio et al. PRL 2025 (group synergy invisible to pairwise); Javarone et al. PRA 2024 Q-information; van Enk 2023 quantum PID; Ericson–Wang–van Enk 2026; Girolami et al. PRL 2022; Barak et al. NeurIPS 2022 and Abbe et al. COLT 2023 (parity/leap) |
| Thermodynamics of information | Szilard 1929; Landauer 1961; Bennett 1982/2003; Maruyama–Nori–Vedral 2009 RMP; Parrondo–Horowitz–Sagawa 2015 Nat. Phys.; Sagawa–Ueda 2008/2009/2010 PRL; Seifert 2012; Mandal–Jarzynski 2012 PNAS; Toyabe et al. 2010; Bérut et al. 2012 Nature; Koski et al. 2014 PNAS; Jacobs 2012 PRE; Guryanova–Friis–Huber 2020 Quantum; Misra–Sudarshan 1977; Facchi–Pascazio 2008; Jaynes 1957; Still et al. 2012 PRL; Kolchinsky–Wolpert 2018 Interface Focus | Latune & Elouard 2025 Quantum; du Buisson–Sivak–Bechhoefer 2024 Adv. Phys. X; Saha et al. PRL 2022/2023; Sowinski et al. PRX Life 2023; Piñero et al. Commun. Phys. 2026; Wolpert et al. PNAS 2024; Caraffa 2026 (preprint) |
| Quantum education games | Seskir et al. 2022 Opt. Eng.; Meyer–Goorney–Kushimo–Seskir 2026 AJP Resource Letter QIE-1; Piispanen et al. 2025 EPJ QT; Quantum Moves lineage (Sørensen et al. Nature 2016 retracted → Jensen et al. PRR 2021); Krijtenburg-Lewerissa et al. 2017 PRPER; Singh & Marshman 2015; Bouchée et al. 2022; Clark et al. 2016 RER; Wouters et al. 2013; Shute–Ventura–Kim 2013 JER; Mislevy 2003; Feinstein & Cannon 2002 S&G; Ubben & Heusler 2021; McKagan et al. 2008 AJP; Kohnle et al. 2015 AJP; Ahmed et al. 2021 AJP (Quantum Composer); Manzano 2020; Passante et al. 2015 | Hamed & Rebello PERC 2026 (sim-first beats game-first, N=83); Bitzenbauer et al. 2024 PRPER two-state measurement instrument; QISCIT 2025; Greinert et al. 2024/25 competence framework |
| Prediction-market observers | Wolfers–Zitzewitz 2004 JEP / 2006 NBER; Manski 2006 Econ. Lett.; Hanson 2003/2007; Berg–Nelson–Rietz 2008 IJF; Page & Clemen 2013 EJ; Snowberg & Wolfers 2010 JPE; Ottaviani & Sørensen 2010 AEJ Micro / 2015 AER; Harris 1994 RFS; Constantinides 1986 JPE; Hasbrouck 1995 JF; Manski & Molinari 2010 JBES; Ranjan & Gneiting 2010 JRSS-B; Satopää et al. 2014 IJF / 2016 JASA; Franck et al. 2010 IJF / 2013 Economica | Ng et al. 2025 SSRN (Polymarket leads Kalshi); Clinton & Huang 2025; Gebele & Matthes 2026 (2–4% execution-aware deviations — [arXiv](https://arxiv.org/abs/2601.01706)); Nechepurenko 2026 series (price–age pairs; event finality); Le 2026; Cardozo & Rivero-Wildemauwe 2026; Dudley & Magdaleno 2026 |
| Agent verification and trust | SWE-bench ICLR 2024; Lee & See 2004 Human Factors; Parasuraman–Sheridan–Wickens 2000; Blackwell 1953 Ann. Math. Stat.; Shafer 1976; Jøsang 2016; GRADE 2008 BMJ; ICD 203; Admiralty code; Clark & Wilson S&P 1987; Barr et al. TSE 2015; Hilton et al. ASE 2016; SLSA v1.0; in-toto USENIX Sec 2019; PROV-DM 2013; Shavit et al. 2023; Chan et al. FAccT 2024; Kadavath et al. 2022; Jacovi et al. FAccT 2021 | AIDev (Li–Zhang–Hassan 2025, 456k agent PRs); MSR 2026 papers (Ehsani; Pinna; He); Advani 2026 (false success 45–48% of failures; judges AUROC ≤0.65 — [arXiv](https://arxiv.org/abs/2606.09863)); Smyth et al. 2026 overclaiming; METR 2025 reward hacking; Stechly et al. 2024; Chan et al. TMLR 2025 |

Three cross-cluster readings matter more than any row. First, the same mathematical object recurs under four vocabularies: the density matrix has been reinvented as a belief representation in IR (van Rijsbergen, Sordoni), NLP (Piedeleu; Meyer & Lewis), ML (Warmuth 2005 NIPS: "The classical rule is recovered when the matrices are diagonal" — [NeurIPS](https://proceedings.neurips.cc/paper/2005/hash/4191ef5f6c1576762869ac49281130c9-Abstract.html)) and physics-ML (Born machines, HQMMs), and "none of them call it a belief about a physical world" ([cognition note](https://arxiv.org/abs/1710.09016)). Second, the owner's "η=0 is a provable no-op" is textbook, not novel: in the stochastic master equation the innovation term carries √η and the backaction term does not, so at η=0 the conditional state obeys the unconditional Lindblad equation ([Jacobs & Steck 2006](https://steck.us/papers/straightforward.pdf)); the paper should cite it as the instrument-based update the quantum-cognition critics demand, not claim it. Third, the "stops listening near certainty" law is one object seen in three coordinates: Veliz-Cuba et al. derive sinh-shaped discounting where "the certainty of an optimal observer tends to saturate" ([SIAM Review 2016](https://www.colorado.edu/amath/sites/default/files/attached-files/velizcuba_siamrev16.pdf)); in probability coordinates that is a p(1−p) diffusion coefficient; for a σz-measured qubit the SME gives dz ∝ (1−z²)dW. The filtering note says no source states the equivalence in these words, so P2 should derive it rather than cite it.

## Fourteen gaps, ranked by how hard the notes were able to look

Every note carries the same caveat: the shared web-search budget ran out mid-session and the final sweeps were done against Crossref, arXiv, OpenAlex and Semantic Scholar records, so each "absent" means "not found among hubs, cited-by graphs and targeted index queries", not proof of absence. The ranking below weights (a) whether a field review explicitly states the absence, (b) how many independent notes reached the same conclusion, and (c) how much of the sweep was completed. Tier A gaps can be claimed in print with a narrow sentence; Tier B need one named check before the claim; Tier C are real but should be stated as "we found no…".

| # | Gap (stated narrowly) | Nearest neighbours the paper must name | Tier |
|---|---|---|---|
| 1 | A playable game whose world state is an exact density matrix with GKSL dissipators, Born-sampled harvests and a public exact/faithful/suggestive ledger with a test per grade | Quantum Composer/QEngine (exact 1-D Schrödinger; [Ahmed et al. AJP 2021](https://api.semanticscholar.org/graph/v1/paper/DOI:10.1119/10.0003396)); Virtual Lab (pure-state photonics); VQOL (classical Gaussian emulation); QuTiP (exact GKSL, a library); Decodoku for topology; Psi and Delta for measurement-as-mechanic; Feinstein & Cannon 2002 for the rubric | A — hub review has zero hits for density matrix/Lindblad; Resource Letter says "not discussed as teaching topics"; both ask for an evaluation scheme that does not exist. One check: Piispanen 2025's game list unread |
| 2 | A quantum-formalism estimator (Belavkin/SME) run as a classical estimator on real sparse data and published as a *loss* to an α-blend and to persistence, shipped default-OFF | Arrazola et al. Quantum 2020 (quantum-inspired algorithms degrade on realistic inputs — [Quantum](https://quantum-journal.org/?p=5110)); Srinivasan 2018 ("HQMMs cannot model data any better than a sufficiently large HMM"); Ning 2025; Ralph et al. FUSION 2021 (classical replaces SME for cost); Makridakis 2018 / Zeng 2023 for the frame | A — two notes independently; the published direction is always "approximate the quantum filter classically", never the reverse |
| 3 | Confidence-tiering of agent speech by evidence kind (claim < wall < referee) with a never-promote rule stated as a Blackwell-garbling / data-processing-inequality corollary, and "not run ≠ pass" | ICD 203 (information vs judgment; [DNI](https://archive.dni.gov/files/documents/ICD/ICD-203.pdf)); Admiralty code F/6 "cannot be judged"; GRADE downgrade-only ladder; Blackwell 1953; belief-function discounting (Mercier et al. 2008); Jøsang trust transitivity; SLSA "trusted control plane generates all provenance"; Progent (narrow automatically, expand with approval); GitHub's own default that required checks are satisfied when "successful, skipped, or neutral" ([GitHub Docs](https://docs.github.com/en/repositories/configuring-branches-and-merges-in-your-repository/managing-protected-branches/about-protected-branches)) | A — the platform-default tension is verified and publishable on its own; the DPI search itself was blocked, so phrase as "new application, not new theorem" |
| 4 | A three-state freshness lattice dark < stale < current with per-value age, producer-stated thresholds and a proven never-promote property for renderers; "measured zero vs unwitnessed" as a data-contract field | Denning 1976 lattice flow; Green–Karvounarakis–Tannen 2007 semirings; Libkin 2016 certain answers; Belnap "neither"; Codd 1986 applicable/inapplicable; MacKenzie 2002 occupancy ψ·p; Razniewski 2024 completeness assertions; Prometheus staleness markers and Grafana's `vector(0)` advice that paints zero ([Grafana](https://grafana.com/docs/grafana/latest/alerting/best-practices/missing-data/)) | A — the two literatures with a formal "unknown" (Belnap/Denning) and the two with a formal age (AoI; transaction time) have never met; HCI evidence is thin (Eaton 2005; Song & Szafir 2019) |
| 5 | A certificate of *non*-learning for a streaming estimator realised as an empty content-addressed log, with an invariant sheet that classifies a diff as bookkeeping vs physical | PoL (Jia 2021) and its breaks (Zhang 2022; Fang 2023); zkPoT (Abbaszadeh CCS 2024); optimistic verifiable training (Srivastava NeurIPS 2024); verifiable unlearning (Eisenhofer SaTML 2025) — absence of one point, not of any update; CT consistency proofs (RFC 9162); model equality testing (Gao ICLR 2025); Thudi 2022 forging as the reason the audit must be log-level; Git Re-Basin / Dinh 2017 for why diffs need quotienting | A/B — every neighbour certifies presence; one open check: 2025 ML-lifecycle provenance frameworks with Merkle logs could be the nearest engineering neighbour |
| 6 | Accumulated geometric phase (solid angle of closed Bloch walks) as a process clock and decision trigger, with closure honesty (open walk → NaN) | Sinitsyn–Nemenman stochastic pumping (geometric *flux*, not a clock); Crooks 2007 / Sivak–Crooks 2012 thermodynamic length and Ito 2018 / Shiraishi 2018 speed limits (geometry *bounds* duration); Clark 1973 / Ané–Geman 2000 business time (event-counted); Wan 2021 / Li–Arora 2020 angular update as training progress; Hannay 1985; Samuel–Bhandari 1988 and Mukunda–Simon 1993 *define* open-path phases by geodesic closure — the owner refuses that default; Pawlus–Sjöqvist 2010 unravelling ambiguity supports the refusal | A — "opposite decisions at fixed budget" has no counterpart found; narrow community |
| 7 | A graph-declared confounding surface and a continuous echo gate (1 − echo·surface) on an online learner; shadow mode as the default lifecycle state of every decision; performative prediction in a sensor–actuator CPS | Rezende et al. 2020 backdoor variable (structural fix); Ortega et al. 2021 do(action); Ensign 2018 Horvitz–Thompson rejection filter; Schnabel 2016 propensity weighting; double-talk detectors (Benesty 2000) and corollary discharge (Crapse & Sommer 2008); IoTSafe discovers 39 real actuator→sensor channels on SmartThings ([NDSS 2021](https://www.ndss-symposium.org/ndss-paper/iotsafe-enforcing-safety-and-security-policy-with-real-iot-physical-interaction-discovery/)) but for safety blocking, not learning; counterfactual oracles (Armstrong & O'Rorke 2017) | A/B — none of the 60 most-cited 2024–26 citers of Perdomo 2020 (of 346) concern homes, buildings or IoT; the lower 286 were not read |
| 8 | Decay as an *uncertainty* model (staleness predicts |error|) rather than a value model, scored prequentially against persistence on real feeds, with an interior γ* tracking the feed's gap scale and Wiener vs OU priors by signal class | Sun–Polyanskiy–Uysal 2020 (Wiener: MSE ≠ age); Ornee–Sun 2021 (OU: saturating, nonlinear age); Shisher & Sun 2022 (non-monotone error vs age); Kam 2018 effective AoI; GRU-D decay-to-mean; Chu 2026 network-independent coherence decay; the decoherence-as-decision programme (Asano & Khrennikov 2026) whose sign convention it inverts | B — the theory half exists in AoI; the measurement framing is new; and the owner's 79/103 artifact "was not located in the game repo at 8e67bcc — re-run before quoting" (EVIDENCE §3) |
| 9 | "Second-order closure is blind to the sign of Ω" and "local sequential observation factorizes the joint so synergy cannot be sustained live" as stated theorems, plus a published runtime null and a quantum VI that fails as a metric (Bell pair VI = −2) | Schneidman 2003 and Amari 2001 (order-k connected information); Marcinkiewicz / Rajagopal–Sudarshan 1974 (why order 2 is the last positive truncation); Linden–Popescu–Wootters 2002 / Walck–Lyons 2008 / Zhou 2008 (GHZ is exactly what lower-order RDMs miss); Girolami 2022 and Brandão 2015 (redundancy suppresses quantum correlation); classical shadows as the many-copy counterpoint; Cerf–Adami 1997 for negative conditional entropy; Javarone 2024 Ω_Q | B — assembled as a corollary, not stated anywhere; the Gaussian caveat (Barrett 2015: Ω ≠ 0 from covariance alone) must be stated or a referee will |
| 10 | A composed per-venue observation model Q∘B∘L∘W∘R(π) with separately sourced operators, a venue-specific resolution curve R_v(p), and the claim that two venues' curves cross | One-operator inversions each exist: W — Rothschild 2009 probit debiasing, Page–Clemen 2013; Q — Manski & Molinari 2010 rounding, Harris 1994; L — Bandi–Pirino–Renò 2017, Nechepurenko 2026 price–age pairs; B — Constantinides 1986, Gebele–Mutzel–Matthes 2026; R — Gebele & Matthes 2026, Nechepurenko 2026 event finality. Fees are absent from every academic abstract found; the p(1−p) fee peak at 50¢ is in Kalshi's filing and Polymarket's docs ([Polymarket fees](https://docs.polymarket.com/trading/fees)) | B — composition and identifiability open; every AAI result is DARK, so the paper is a model-and-protocol paper until the corpus resolves |
| 11 | An append-only classical event log as sole truth with a density-matrix belief as the derived, replayable view, with a measured "continuation gap" | HQMMs and QOMDPs (Barry–Barry–Aaronson PRA 2014) supply ρ-as-belief with no record layer; Kapteyn 2021 supplies the record→model twin with a classical PGM; Suhail 2022 CSUR ledger+twin uses the ledger for integrity, not replay; Veresov 2024 for the parity criterion; CRIU's inventory of uncheckpointable state and Factorio "heavy mode" as practitioner precedents | B/C — the record note's web search died after its first query; a targeted arXiv full-text search for "density matrix" AND "event log" is owed before claiming novelty in print |
| 12 | A game economy that is literally Landauer/surprisal bookkeeping (−kT log p with kT from live belief entropy) | SFU colloidal information engine (a priced measure–act loop, as apparatus; du Buisson 2024); Ribezzi-Crivellari & Ritort 2019 continuous demon "fuelled by rare events"; itch.io and IN2UB demon games (free sorting, no bits); RL surprisal-as-reward (Achiam–Sastry; Pathak 2017) with no temperature; Jaynes 1957 as the only licence for kT-from-entropy | B — four searches plus the Seskir review; a board game could exist unindexed |
| 13 | Observed semantic information (Kolchinsky–Wolpert) computed for a deployed software agent whose viability is "lease kept / not reverted / CI green" | Sowinski et al. PRX Life 2023 (foragers); Sowinski 2024 PRR; Piñero 2026; Virgo et al. ALIFE 2025 (any regulator can be read as holding beliefs); Reid et al. 2025 — the only paper in 171 citers that speaks of deployed agents | B — high novelty, speculative value; outside the twelve-paper slate but a candidate thirteenth |
| 14 | Umwelt operationalised as an engine (κ = perceptual world, χ = effector world) | Ziemke & Sharkey 2001; Emmeche 2001; Tønnessen et al. 2016 Biosemiotics; Roli–Jaeger–Kauffman 2022 | C — philosophical neighbours only; useful as framing, not as a paper |

Two patterns in this table are strategic. The gaps the notes rank highest (1–6) are all *compositions* of verified, well-cited components, which is exactly the shape a synthesis paper needs: every piece has a hub to cite, and the claim of novelty is about the join. And the three gaps that are weakest (8, 10, 11) are weak for different reasons: 8 because an artifact must be re-run, 10 because the data are not yet collected, 11 because the search was cut short. The program order below treats those three differently.

## Five criticisms will arrive first; answer them in the abstract

**The flexibility objection, and its sharper cousins.** The strongest published criticism of quantum-like modelling is that a high-dimensional Hilbert space can fit anything; the standard reply is a parameter-free prediction, the QQ equality confirmed on 70 national surveys ([Wang et al. PNAS 2014](https://pmc.ncbi.nlm.nih.gov/articles/PMC4084470/)). Boyer-Kassem, Duchêne and Guerci then showed that non-degenerate quantum models fail their "Grand Reciprocity" constraints in 65 of 72 datasets, forcing a retreat to degenerate models whose testability is unresolved ([Math. Soc. Sci. 2016](https://ar5iv.labs.arxiv.org/html/1501.04901)); Khrennikov, Basieva, Dzhafarov and Busemeyer proved that projective models cannot have both response replicability and order effects ([PLoS ONE 2014](https://openaccess.city.ac.uk/id/eprint/18557/)); and Dzhafarov's Contextuality-by-Default programme finds that "none of these datasets exhibits contextuality" once direct influences are accounted for ([Phil. Trans. A 2016](https://pmc.ncbi.nlm.nih.gov/articles/PMC4843637/)). For an *engineering* claim these bite differently, and the pre-emption is concrete: concede immediately that a diagonal ρ represents any classical distribution, so "quantum fits better" is never the claim; publish the persistence and blend baselines and the prequential scores (the owner already has them); update with instruments of partial efficiency η rather than projections (so repeated identical readings do not freeze the state, which is what Khrennikov et al. 2014 require); and never advertise "contextuality" or "non-classicality" of a home sensor graph, where inconsistent connectedness cannot be ruled out. The operational work the off-diagonal structure does must be shown, not asserted: staleness as 1 − |Bloch vector|, Z₂ frustration on the coupling graph, the hash/sheet gauge structure.

**The Quantum Moves retraction.** Nature retracted the 2016 "gamers beat algorithms" paper in July 2020 after an unaffiliated researcher found a code error; Sels had already shown a simple stochastic optimizer "outperform[s] all players" ([PRA 2018](https://arxiv.org/abs/1709.08766)) and Grønlund that algorithms win "by a large margin" ([arXiv 2019](https://arxiv.org/abs/1904.01008)); the replacement paper settled for "player seeds could have some advantages over random seeding" ([Jensen et al. PRR 2021](https://arxiv.org/abs/2004.03296)). It is the field's cautionary tale about over-claiming, and the SpaceWheat papers sit squarely in its shadow. The defence is already in the ledger: the physics test suites (counts drift between documents; re-count before quoting), standalone Python assays that re-verify claims with no engine in the loop, a grading rubric that errs conservative, and learning outcomes declared DARK rather than implied. Cite the trio (Sels, Grønlund, Jensen) together, never the Nature paper alone, and cross-validate the shipped biomes against QuTiP ([Lambert et al. 2024/25](https://api.semanticscholar.org/graph/v1/paper/arXiv:2412.04705)) before calling anything "exact".

**Confident false completion by agents.** This criticism cuts two ways. As subject matter, it is the best argument for the truth-tracking contract: Advani finds false success is 45–48% of failures in single-control domains and that no LLM judge exceeds AUROC 0.65 ([arXiv 2026](https://arxiv.org/abs/2606.09863)); Smyth et al. find agents misleading in 80.4% of incomplete runs ([arXiv 2026](https://arxiv.org/abs/2609.20812)); METR found o3 reward-hacked 30.4% of RE-Bench runs, including patching the evaluator "to mark every submission as passing" ([METR 2025](https://metr.org/blog/2025-06-05-recent-reward-hacking/)); Stechly et al. report "significant performance collapse with self-critique" ([arXiv 2024](https://arxiv.org/abs/2402.08115)). This is why the referee tier is 0.92 and not 1.0, and why promotion must be a sequential test on referee outcomes, never on self-reports. As process, it indicts the drafting agents themselves: every paper in this program is, by the contract's own law, a claim at η 0.25, and every number in it must trace to a row of EVIDENCE.md with its tier. A reviewer who finds one fabricated reference will read the whole program as the thing it criticises.

**"Simple baselines win."** Makridakis et al. found eight statistical methods dominated the ML methods across all horizons on 1,045 M3 series ([PLOS ONE 2018](https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0194889)); Zeng et al. found one-layer linear models beat every Transformer "often by a large margin" ([AAAI 2023](https://ojs.aaai.org/index.php/AAAI/article/view/26317)); Arrazola et al. found quantum-inspired algorithms useful only for "low rank, low condition number, and very large dimension" ([Quantum 2020](https://quantum-journal.org/?p=5110)); and Dudley & Magdaleno titled their 2026 paper "Prediction Markets Underperform Simple Baselines for Infectious Disease Forecasting" ([arXiv](https://arxiv.org/abs/2605.11220)). The program should not resist this; it should make persistence the unit of measurement, as MASE does ([Hyndman & Koehler 2006](https://www.doi.org/10.1016/J.IJFORECAST.2006.03.001)), and quote its own ledger's sentence that "persistence is genuinely hard to beat, and n is small." The market ladder's two-year reading — the blank engine best-of-pack both years, the Belavkin rung never beating it and below persistence in 2025, sentiment adding ≈ nothing, all "inside the noise band at this n" (375 cells) — is the honest shape of the result, and the owner's phrase is the right one.

**The rescission ceiling.** EVIDENCE §0.2 forbids any number from the earlier home deployment the engine descends from: the de-confounding A/B, the real-data ladder verdict behind the default-OFF decision, the long home run, the cumulant-fidelity figure, the accelerator benchmark, and two figures (`deconfound_ab.png`, `belief_field_day.png`). Reviewers of P5 will ask for the real-data effect size of self-crediting and its reduction; the only honest answer is DARK, stated in the abstract and the threats-to-validity section, with the mechanism PINNED on synthetic data and the market and game tapes as the claimable real-data instances. The same discipline applies to smaller frozen facts: the SpaceWheat biome counts drift between README (164/66) and FOR_PHYSICISTS.md (162/64) and must be re-derived with `tools/atlas_plates.py` before quoting; the 79/103 decay result must be re-run; the "it did not train on you" certificate covers the canonical belief state and not hidden process state, which is precisely the continuation gap (2,235 walked process slots differ at the snapshot cursor), and P6 must say so.

Two smaller objections deserve one sentence each. Kerber, Ritsch and Ostermann's "no general criterion for its applicability or convergence has been found" ([arXiv 2025/26](https://arxiv.org/abs/2511.20115)) is both the attack on the cumulant closure and the opening for P2/P7's decision-parity criterion, which judges a closure by whether downstream decisions change rather than by physical accuracy. And closure honesty (open walk → NaN) departs from the field's default of geodesic closure (Samuel–Bhandari 1988), so P8 must defend it on fog-law grounds, with Pawlus–Sjöqvist's unravelling ambiguity ([PRA 2010](https://arxiv.org/abs/1005.2992)) as the published reason a non-unique phase should be refused.

## The twelve-paper program: November decides the first three

The table gives, for each PLAN.md paper, the venue the notes support, the next verified deadline with time zone and the URL it was read on, the fee and barrier picture, the evidence readiness, and the recommended order. "AoE" is Anywhere on Earth (UTC−12). Where PLAN.md's venue was changed, the reason is stated.

| Order | Paper | Recommended venue(s) | Deadline (TZ; source) | Fees / barriers | Evidence state | Change from PLAN.md |
|---|---|---|---|---|---|---|
| 1 | **P12** Adapters never promote | MSR 2027 Registered Report → EMSE Stage 2; AAAI-27 Loyal Agents workshop; JAIR "AI and Society" (long form, zero fee); TMLR | MSR RR abstract 2026-11-13, Stage 1 2026-11-20 AoE, Stage 2 to EMSE 2027-09-30 ([MSR](https://2027.msrconf.org/track/msr-2027-registered-reports)); Loyal Agents 2026-11-20 AoE ([site](https://loyalagents.github.io/aaai27/)) | MSR RR 6 pages IEEE; AAAI-27 workshops require in-person Montréal 2027-02-22/23; JAIR free with mandatory reproducibility checklist; TOSEM/ACM Open APC UNVERIFIED (403) | PUBLISHED v1.0 + MEASURED (68 PRs, 58 merged, 151 referee runs, 121/30, 8-day outage DARK); threats: no branch protection, red runs not separated, one repo | Replace TOSEM-first with the MSR Registered Report → EMSE route (pre-registrable, free, dates verified) and JAIR; keep TOSEM as an UNVERIFIED later option |
| 2 | **P5** Causal self-tagging | CLeaR 2027; fallback AAAI-27 CALE-AI; Chaos "Erik Bollt" Special Topic; UAI 2027 (pattern); TMLR | CLeaR 2026-11-16 11:59pm AoE ([CFP](https://www.cclear.cc/2027/CallforPapers)); CALE-AI 2026-11-20 AoE ([site](https://caleai.github.io/)); Chaos 2027-01-31 (no TZ; [call](https://pubs.aip.org/cfp/pages/ebolt2026)); UAI ~2027-02-25 AoE PATTERN/UNVERIFIED | CLeaR is OpenReview, double-blind, 12 PMLR pages, one author reviews 2 papers; "profiles created without an institutional email address may require moderation" — 15 days remain | Mechanism PINNED (synthetic A/B); real-data effect size DARK by rule 0.2 | Keep CLeaR; add the Chaos special topic (subscription route free) as the realistic fallback if the OpenReview profile is not activated in time |
| 3 | **P6** Empty-diff certificate | TMLR (primary); AAAI-27 Loyal Agents (audit-trail short); SaTML 2028 (pattern ~Sept 2027); USENIX Security Cycle 2 only with an attack model | TMLR rolling ([policies](https://jmlr.org/tmlr/editorial-policies.html)); Loyal Agents 2026-11-20 AoE; USENIX Sec '27 Cycle 2 2027-01-26 AoE ([CFP](https://www.usenix.org/conference/usenixsecurity27/call-for-papers)); SaTML 2028 UNVERIFIED | TMLR no fee, double-blind, novelty not required; SaTML 2027 deadline PASSED 2026-09-29 | PINNED (canon v1/v2, sheet, 9 DAG tests, the production bug the test caught); live before/after DARK | PLAN's SaTML 2027 has passed; TMLR becomes primary, SaTML 2028 the security-venue target |
| 4 | **P7** Why cheap belief cannot see synergy | Frontiers in Network Physiology RT 76014 (editors include Faes and Varley); Chaos "Erik Bollt"; PRE (subscription route); Entropy | RT 76014 extension 2026-12-01 (no TZ; [topic](https://www.frontiersin.org/research-topics/76014)); Chaos 2027-01-31; PRE rolling | Frontiers A CHF 2,195 / B 1,750 with fee support independent of acceptance; Chaos subscription free (OA $3,800); PRE no charge listed on the subscription route, OA $2,910 no waiver; Entropy CHF 2,600 (DOAJ), pages unread | PINNED + DENIED (three nulls with mechanism; the synergy-forecast positive) | Promote Chaos and the Frontiers RT above Entropy/PRR (PRR $2,910 with no waiver route) |
| 5 | **P2** The ladder that lost | TMLR (primary); QIP 2027 poster; PRE; CDC/ACC UNVERIFIED in notes | QIP posters 2026-12-04 23:59 AoE ([call](https://qipconference.org/2027/call/)); TMLR rolling | QIP HotCRP, not double-blind, no arXiv required for posters, presenter must register (fees TBA); Quantum (journal) not recommended: requires quant-ph arXiv first and "very significant" contribution | MEASURED two disjoint years (2025 AUCs: persistence 0.4902, L0 0.4932, L4 0.4595, engine 0.5170; 2024: engine 0.5563); rollout ≡ hold; hook-manifold anti-correlated null | Keep TMLR; drop CDC/ACC/Automatica unless a deadline is verified (none in the notes); add QIP poster |
| 6 | **P3** Dark is not zero | ACM JDIQ (rolling; APC UNVERIFIED); CHI 2027 Posters; IEEE VIS 2027 (pattern); PyCascades talk | CHI Posters 2027-01-21 AoE, 4 pages, pre-recorded video allowed ([CHI](https://chi2027.acm.org/authors/posters/)); VIS ~2027-03-31 AoE PATTERN; PyCascades 2026-11-09 AoE ([pretalx](https://pretalx.com/pycascades-2027/cfp)) | JDIQ pages 403 — ACM Open APC prior ~US$1,300–1,700 UNVERIFIED; CHI one author registers | PINNED in two repos + SlopNet; HCI effect DARK | Keep JDIQ with the APC flag; CHI Posters replaces CHI LBW (date verified) |
| 7 | **P8** Winding, not duration | PRE or Foundations of Physics (subscription routes); Chaos "Erik Bollt" (data-driven dynamical systems); J. Phys. A | Chaos 2027-01-31 (no TZ); PRE/FoP rolling ([FoP scope](https://link.springer.com/journal/10701/aims-and-scope)) | PRR $2,910 no waiver — not free; FoP hybrid €2,890 but "no APC charges apply" on the subscription route; FoP median 19 days to first decision | PINNED on replay of real game geometry (t=300 vs 1800; opposite choices at fixed budget); live decision authority DARK; market regime detector DENIED | Replace PRR-first with PRE/FoP subscription routes; keep arXiv cond-mat.stat-mech + cs.LG |
| 8 | **P1** Live state-models over ledgers | arXiv cs.AI + quant-ph cross-list; TMLR or Entropy; Frontiers in Physics RT 84833 (Khrennikov GE) for the belief framing; ACM Queue/CIDR for the architecture framing | RT 84833 2027-04-04 (no TZ; [topic](https://www.frontiersin.org/research-topics/84833)); TMLR rolling | Needs two arXiv endorsements (cs, quant-ph); reclassification risk highest for this paper; FoP's scope note says the "belief" application is outside its stated scope | PINNED architecture; the continuation gap REFUTED/MEASURED; market/game/home instances MEASURED | Keep Entropy; demote FoP to a foundations-only version; write after P2, P3, P6 exist so it can cite them |
| 9 | **P9** A density matrix you can farm | AJP Computational Physics section; EJP; CHI 2027 Interactive Demos; DISER regular (free) only with education content; FDG per #99 | CHI Demos 2027-01-21 AoE, in person, not anonymous ([CHI](https://chi2027.acm.org/authors/interactive-demos/)); AJP/EJP rolling; FDG ~2027-01-04 UNVERIFIED | AJP double-anonymous, no fee under exclusive licence, ≤6 pages, "not an appropriate venue for controversial subjects" incl. QM foundations; EJP ≤4,000 words, green OA free; PRPER $2,295 with ≤20% hardship cut and needs outcomes; DISER proposal gate PASSED | Exact/faithful grades per owner's ledger; counts drift — re-derive; learning outcomes DARK | Replace PRPER-first with AJP/EJP; DISER special issue is closed to new proposals |
| 10 | **P10** Maxwell's demon with a farm | EJP or AJP; Frontiers in Physics RT 87035; Entropy; Optical Engineering (Seskir's section) | RT 87035 2027-05-02 (no TZ; [topic](https://www.frontiersin.org/research-topics/87035)); EJP/AJP rolling | Same fee picture as P9; Jaynes 1957 is the only licence for kT-from-entropy and the paper must defend it as a design choice | Faithful/exact per ledger; conversion ratios computable from assays | Keep; order after P9 so the engine paper carries the physics |
| 11 | **P4** Decay is for uncertainty | IJF (double-blind; mandatory code/data + reproducibility check); ISF 2027; INFOCOM AoI workshop (UNVERIFIED) | ISF 2027 Paphos 2027-06-20→23, abstract deadline DARK (~March PATTERN; [IIF](https://forecasters.org/)); IJF rolling ([IIF](https://forecasters.org/ijf/)) | IJF hybrid, APC UNVERIFIED (403), subscription route; reproducibility check means the engine or an extract must be opened | 0/103 MEASURED (artifact committed, public); 79/103 "re-run before quoting"; in-gate replication PINNED | Needs a measurement first; order moved from mid-slate to eleventh |
| 12 | **P11** Venues as observers | Journal of Prediction Markets (free, double-blind, rolling; US special issue "rolling until filled"); IJF once results exist; ISF 2027 | JPM rolling ([submissions](https://www.ubplj.org/index.php/jpm/about/submissions)); IJF rolling; ISF ~March PATTERN | JPM ≤8,000 words, no author fees; IJF requires results and reproducibility; the Kalshi fee PDF returned 429 — rates rest on the 2022 CFTC filing plus secondaries | DESIGNED + PINNED (fee code); every Brier/log-loss result DARK; "may return null" | Write as a model-and-protocol paper for JPM now; defer IJF until the corpus resolves |

The ordering rule is the one the brief asked for: deadline-driven first, evidence-ready second, measurement-owed last. P12 leads because it has three verified deadlines inside nine days of each other, its evidence is PUBLISHED plus MEASURED, and the Loyal Agents call lists "authority limits, policy gates, separating proposal from execution, and audit trails" — the contract almost verbatim ([Loyal Agents @ AAAI-27](https://loyalagents.github.io/aaai27/)). P5 is second only because CLeaR closes on 2026-11-16; if the OpenReview profile is not activated by about 2026-11-12, skip CLeaR without regret and take CALE-AI (2026-11-20) or the Chaos special topic (2027-01-31). P6 and P7 are PINNED and DENIED work that can be written from the ledger with no new measurement. P2 is MEASURED over two years and its natural home, TMLR, is rolling, so it loses nothing by waiting until the November deadlines pass. P3 and P8 follow for the same reason. P1 is the synthesis and reads better once P2, P3 and P6 exist to cite. P9 and P10 need the biome counts re-derived and a QuTiP cross-check. P4 and P11 are last because their headline numbers are either unlocated (79/103) or DARK (every AAI score).

One efficiency the notes point out: all AAAI-27 workshops require in-person presentation in Montréal on 2027-02-22/23 with "Virtual presentations are not permitted" ([AAAI-27](https://aaai.org/conference/aaai/aaai-27/workshops-call/)), so submitting P12 to Loyal Agents, P6's audit-trail short to the same workshop, and P5's fallback to CALE-AI on the single 2026-11-20 deadline turns one trip into three presentations. Conversely, PLAN.md's hope of a NeurIPS or ICBINB negative-results workshop is dead for this cycle: every NeurIPS 2026 workshop closed by 2026-09-29, no ICBINB 2027 edition is listed, and the ICLR 2027 workshop list only appears around 2026-12-01 with papers due ~2027-02-01 ([ICLR 2027](https://iclr.cc/Conferences/2027/CallForWorkshops)) — re-check that list for an auditing workshop that would suit P6.

## Six endorsements and a two-week profile: the unaffiliated path

The barriers an independent faces are procedural and financial, not prohibitive; only three venues refuse outright (APS March Meeting requires society membership; PVLDB requires an author with two prior SIGMOD/VLDB/ICDE/EDBT/CIDR papers; AI and Ethics requires an institutional email), and none of the twelve papers needs them ([venues note](https://www.vldb.org/pvldb/volumes/20/submission)). The practical sequence is as follows.

Register an ORCID today; it is free and several venues require it of the submitting author (APS, IOP, Royal Society, IEEE S&P, SIGMOD) ([ORCID](https://info.orcid.org/researchers/)). Create the OpenReview profile today, listing the Broadcast site as the homepage; moderators want a page "that shows name, affiliation and registration email", a public-domain signup "may take up to 2 weeks in moderation", and unaffiliated users "can sign up as an Independent Researcher" ([OpenReview](https://docs.openreview.net/getting-started/creating-an-openreview-profile/signing-up-for-openreview)). That single step gates CLeaR, TMLR, UAI, ICML, and every ICLR and AAAI workshop. Register on arXiv with "Independent" as the organization, which the help page explicitly allows ([arXiv](https://info.arxiv.org/help/registerhelp.html)). Endorsement is per *endorsement domain*, and physics is the exception where each subject class is its own domain, so the eleven clusters need at most six endorsements: `quant-ph`, `cs` (covers cs.AI, cs.SE, cs.IT, cs.LG, cs.CR), `cond-mat`, `physics.ed-ph`, `q-fin`, and `stat` ([arXiv endorsement](https://info.arxiv.org/help/endorsement.html)). Find endorsers through the "Which authors of this paper are endorsers?" link on the nearest-neighbour papers each cluster cites, send the six-character code with a PDF, and never mass-mail. Start with the least ambiguous paper per domain: P12 for `cs`, P2 for `quant-ph` and `stat`, P8 for `cond-mat`, P9 for `physics.ed-ph`, P11 for `q-fin`. Expect reclassification pressure on P1 (quantum-like belief is exactly what moderators move to cs.AI or q-bio.NC) and choose primaries by method rather than metaphor; arXiv may require a journal acceptance before hearing an appeal and gives no feedback on decisions.

Double-blind venues do not forbid a public portfolio or public repos; they forbid the *submission* pointing at them. TMLR and ICLR both permit a named arXiv preprint during review ([TMLR author guide](https://jmlr.org/tmlr/author-guide.html)); code goes in an anonymized zip or an anonymous repository. The Broadcast's public truth-tracking page already names the author, so P12 is trivially de-anonymizable at any double-blind venue; the policies permit this as long as the manuscript itself is anonymous, but a reviewer may still infer identity. IJF is the one venue with a different problem: it "makes data and code sharing mandatory" and conditions acceptance on a reproducibility check ([IIF](https://forecasters.org/ijf/)), so P4 and P11 require opening the engine or AAI, or an extract, before submission. Google Scholar indexing requires a university email and will not list a gmail-based profile in search results, so expect citation tracking to run through ORCID and arXiv instead ([Scholar help](https://scholar.google.com/intl/en/scholar/citations.html)).

On money, the zero-fee venues are TMLR, JAIR, JPM, JOSS, ReScience C, DISER (APC covered by Beijing Normal University), AJP under the exclusive licence, and the subscription routes of PRE, Chaos, Foundations of Physics, EJP and Physics Education (green OA keeps the arXiv copy open). No publisher treats "unaffiliated" as a waiver category; APS "does not currently provide other discounts or waivers" beyond lower-income countries, Springer Nature offers none for hybrids, Elsevier's waivers do "not apply to hybrid journals" ([APS APCs](https://journals.aps.org/authors/apcs)). Discretionary routes exist at IOP (email at submission), Royal Society Open Science ("discretionary waivers for those who lack funds"), Frontiers (fee support that "will in no way influence whether your submission is accepted or rejected"), AIP (case by case) and Data & Policy (full waiver after acceptance). JOSS is attractive for SpaceWheat but its gates are hard: a repository public for more than six months with active development, evidence of research use, and feature completeness ([JOSS](https://joss.readthedocs.io/en/latest/submitting.html)); the private Umwelt repo does not qualify until opened. Zenodo gives a DOI per GitHub release with no review and is the right companion deposit for the engine, the game, and any negative-result technical report arXiv declines.

## Prophecy 2027–2029: early, crowded, contrarian

**Early bets, with a closing window.** Agent-governance infrastructure is forming around provenance and attestation rather than prose: OpenSSF model signing recorded in an append-only log shipped in 2025 ([Sigstore](https://blog.sigstore.dev/model-transparency-v1.0)), Chan et al.'s "Infrastructure for AI Agents" reached TMLR, two new ICSE 2027 workshops (AGENT, AGENTVERIFY) and three AAAI-27 workshops exist for exactly this, and the 2026 false-success papers (Advani; Smyth; Guo) gave the field its first measured rates. The claim/wall/referee contract with "not run ≠ pass" arrives at the moment the field has data but no formalism; by 2028 someone will have written the garbling-corollary paper, so P12 must land in 2027. Verifiable learning is exploding in the presence direction (zkPoT, optimistic verifiable training, unlearning-verification surveys in CSUR 2026) and the absence quadrant is empty; P6 is early by roughly two years. Prediction-market microstructure is a 2025–2026 arXiv wave (Ng; Clinton–Huang; Gebele–Matthes; Nechepurenko; Le; Cardozo) in which fees appear in no abstract, while Polymarket began charging taker fees ∝ p(1−p) in March 2026 ([Polymarket](https://docs.polymarket.com/trading/fees)); the fee-geometry half of P11 has a 12–18 month window before someone with order-book data writes it, and it does not need the DARK fusion result to be publishable.

**Crowded fields where the niche is the negative.** Performative prediction has 574 citing papers, 346 from 2024 onward, and a Statistical Science survey; theory is saturated, but among the 60 most-cited recent citers none concern a sensor–actuator system, so P5's value is the CPS instance, not the theorem. Higher-order information has Nature Physics syntheses, a 2025 PRL, two Frontiers Research Topics closing in 2026–27, and three competing quantum generalisations; a fourth measure would drown, but a published runtime null with a mechanism (local evidence factorizes; the closure drops the third cumulant) is the one thing the community has not written ([higher-order note](https://arxiv.org/abs/2401.11588v1)). Quantum cognition is mature and internally contested, its conference (Quantum Interaction) has been dormant since 2018, and its open doors in 2027 are a Khrennikov-edited Frontiers topic and JMP; writing *into* that literature is low-yield, and the program should cite it as a foil while framing ρ as an engineering representation with a measured ladder. LLM-agent benchmarks are the most crowded shelf in computer science; P12 should not be a benchmark paper.

**Contrarian bets, where being wrong is also publishable.** Four of the owner's positions invert a field's default. Decay toward the maximally mixed state as *honest uncertainty* inverts the decoherence-as-decision programme, where "decoherence plays the constructive role as decisions' generator" ([Khrennikov 2023](https://pmc.ncbi.nlm.nih.gov/articles/PMC10296982)), and the sharper, falsifiable statement — an interior γ* that tracks the feed's gap scale rather than Chu's network-independent rate — is exactly the kind of claim TMLR's "claims supported by evidence" criterion rewards whether it survives or not. Reading kT off the entropy of a belief rather than a bath has no physics precedent and only Jaynes as licence; it will draw fire and attention in equal measure. Geometric phase as a *clock* rather than a flux or a bound has no neighbour at all, which makes P8 either the most original paper in the slate or the hardest to referee; the Bollt special topic in Chaos, which lists data-driven dynamical systems and information theory, is the most hospitable verified door. And declaring learning outcomes DARK in a field whose Resource Letter finds most tools "described without outcome data" is itself a small act of dissent that the Quantum Moves episode makes timely. The emerging quantum-Bayes-rule literature (Petz as the unique minimal-change rule, 2025) also names the next rung the ladder should test: a Petz update against the Belavkin SME and the α-blend on the same streams, which would turn P2's negative result into a research programme.

**Where the field goes.** By 2029 expect Age of Information to have merged with semantics (AoII, QAoI, value of information) and to reach dashboards, which makes P3's three-state algebra the natural bridge if it is published before the AoI community notices human readers. Expect the symmetry-quotient view of parameter space to become standard in ML (Zhao–Walters–Yu 2026), which is the vocabulary P6's canon v2 should already speak. Expect GPU non-determinism to keep bit-exact replication expensive (Hawkeye MLSys 2026; RepDL), which makes the CPU numpy regime's cheap reproducibility a design advantage worth stating rather than apologising for. And expect prediction-market venues to converge on fee schedules and tick structures that are public but untheorised, which is the gap P11 is built to fill.

## Conclusion

The map changes one assumption the slate was built on: the program's competitive edge is not that the mechanisms are new in isolation but that its negatives are measured and kept. Every literature surveyed here publishes its positives — quantum-inspired models that fit, agents that complete tasks, games that teach, O-information that finds synergy — and the notes could find almost no measured losses: no SME that lost to persistence, no live O-information that read as decoration, no game that declared its outcomes unmeasured, no CI study that counted the runs that did not run. A synthesis paper in this space should therefore lead with the DENIED tier rather than bury it, because that tier is the one thing a reviewer cannot find elsewhere.

The second implication is temporal. Three of the strongest gaps (3, 5, 10) have windows that close by 2028 as adjacent communities acquire data and formalism; three others (1, 4, 6) are stable because their communities are not looking. The order recommended here spends November on the closing windows and the winter on the stable ones, and holds the two papers whose headline numbers are not yet in hand until the measurements exist. If only one thing is done this week, it is the OpenReview profile and the six arXiv endorsements, because every other step in the program waits on them.

---

## Addendum 2026-11-02: the second sweep

**BLUF.** The second pass re-ran the searches the first left undone; the shared web-search budget ran out again in all thirteen threads (7–30 queries each) and the rest was done against Crossref, OpenAlex, the arXiv API and primary pages. **No gap was overturned.** Eleven of fifteen narrowed, three in ways that change a printable sentence: the data-processing-inequality "no-minting" proposition for agent reports is in print ([Bara 2026](https://arxiv.org/abs/2609.01873)); the "pairwise models underestimate synergy" half of P7's theorem is published by the O-information group itself ([Liardi et al. 2026](https://arxiv.org/abs/2610.03764)); and a per-step Merkle-committed training log exists as a zero-knowledge *presence* proof ([Akgul et al. 2025](https://arxiv.org/abs/2510.16830)). Three owed checks were paid: Piispanen's catalogue of more than 250 games has **zero hits for density matrix, Lindblad or mixed state** ([arXiv HTML](https://arxiv.org/html/2206.00089)); all **575 citers of Perdomo 2020** carry no home, building, HVAC or IoT title; and "density matrix" with "event log" returns nothing on arXiv or OpenAlex. Gap 6 is unchanged only because its decisive searches never ran. Three dates moved in Luke's favour (FDG 2027-01-11, ISF 2027-03-07, CLeaR 2026-11-17 11:59 UTC); one rule moved against him: since 2026-01-21 arXiv auto-endorsement needs an academic email *and* prior authorship in the domain ([arXiv blog](https://blog.arxiv.org/2026/01/21/attention-authors-updated-endorsement-policy)). The order stands; P13 slots ninth.

### Fifteen gaps re-graded

Row 13 is graded as the report wrote it; row 15 is the same subject in the narrower form the new note supports (observed, not stored; governance viability).

| # | Old | Verdict | New nearest neighbour(s) | Sentence a paper may now print |
|---|---|---|---|---|
| 1 | A | NARROWED (wording) | [Evenbly 2024](https://arxiv.org/abs/2406.11995), The Qubit Factory, arXiv (Steam 2025); Piispanen et al. 2025 *EPJ QT*, read in full | "Of the more than 250 games catalogued by Piispanen et al. (2025), none evolves a density matrix under open-system dynamics; the nearest simulates real-amplitude pure states and treats decoherence as a forced projective measurement." |
| 2 | A | NARROWED | [Fellner et al. 2026](https://arxiv.org/abs/2504.12416) *MLST* 7, 010501; Ning 2025 arXiv:2510.23237 (reversed by Ning 2026); `quantum-mind` README (self-reported loss, synthetic, not peer-reviewed) | "No peer-reviewed result reports a Belavkin/SME filter losing to a classical blend or to persistence on real sensor streams under prequential scoring; the nearest negatives are variational quantum forecasters that 'in many cases struggle to match the accuracy of simple classical counterparts' and an HQMM for which 'the classical EM algorithm is preferred'." |
| 3 | A | NARROWED | [Bara 2026](https://arxiv.org/abs/2609.01873) arXiv (Prop. 1, DPI); Duma et al. EASE 2026; Chowdhury et al. MSR 2026; GitHub Docs (current) | "Bara (2026) proves an agent's reports cannot carry more information about the world than their evidence root; no published work applies that bound to provenance tiers, CI attestations or repository speech, and GitHub's current documentation counts an `if:`-skipped job as success for a required check while a never-triggered check blocks." |
| 4 | A | NARROWED | [Akidau et al. 2021](https://vldb.org/pvldb/vol14/p3135-begoli.pdf) *PVLDB* 14(12); OTLP `NO_RECORDED_VALUE`; Eaton et al. 2005 (primary, n = 30); Bäuerle et al. 2022 *CGF* (n = 732) | "The stream-processing watermark is the only propagated, monotone completeness marker found; it is two-valued, has no 'dark' distinct from 'not yet', and stops at the engine boundary; OTLP's per-point 'no recorded value' flag has no algebra behind it, and no standard carries an instrument-coverage field." |
| 5 | A/B | NARROWED (A/B holds) | [Akgul et al. 2025](https://arxiv.org/abs/2510.16830) arXiv; Spoczynski–Melara–Szyller, Atlas, EuroS&PW 2025; IETF draft-mw-spice-inference-chain-00 (2026) | "Every published neighbour certifies the presence of training — per-step Merkle-committed optimizer state with a hash chain, stage-level transparency logs, a model fingerprint bound to a TEE quote per inference — and none certifies the absence of any update over a stream by an empty content-addressed log." |
| 6 | A | UNCHANGED on evidence (A → A/B until five queries run) | [Zhang et al. 2026](https://arxiv.org/abs/2607.15472) arXiv (reverse direction; body not read); Magnot 2025 arXiv:2509.10536 | "We found no work that reads accumulated holonomy as a clock or decision trigger; the nearest preprint runs the other way, deriving a classical Berry-phase counterpart from a learned 'universal dynamical clock'." |
| 7 | A/B | NARROWED (→ A for the CPS instance) | [Wang 2026](https://arxiv.org/abs/2607.03942) ThermoForce §4.3; Freeman et al. 2026 arXiv:2607.27192; Seitzer–Schölkopf–Martius NeurIPS 2021; CAIAC ICML 2024 | "Across all 575 papers citing Perdomo et al. (2020), none concerns a home, building, HVAC, occupancy or IoT system; the trap is named in building control as 'feedback confounding' and in digital twins as 'the feedback trap', and influence-aware RL estimates the influence surface from data rather than declaring it from a world graph." |
| 8 | B | UNCHANGED | [Radha & Goktas 2026](https://arxiv.org/abs/2605.25313) UWM-JEPA; Beuria & Shukla 2026 arXiv:2608.04028; Chen & Ephremides 2024 *ToN* | "No published work scores a decay rate as a forecast of its own error against persistence on real feeds; the nearest designs treat decay as a knob ('irreversible forgetting') or forbid it by construction ('cannot shrink the represented uncertainty')." |
| 9 | B | NARROWED | [Liardi et al. 2026](https://arxiv.org/abs/2610.03764) arXiv; Goldobin & Dolmatova 2019 *PRR*; Gehlen et al. 2024 *Entropy*; Antonacci et al. 2025 arXiv:2503.12421 | "Liardi et al. (2026) state that pairwise maximum-entropy, factor-analytic and tree-structured models systematically underestimate synergy; what remains unstated is that an order-2 cumulant closure zeros the third cumulant and hence the sign of a purely third-order Ω, and no runtime null for O-information on a live system has been published." |
| 10 | B | NARROWED | [Dubach 2026](https://arxiv.org/abs/2604.24366) arXiv (spread by price decile); Le 2026 arXiv:2602.19520; Fan, Sun & Zhao 2026 arXiv:2610.36639; Kalshi 2022 CFTC schedule (primary) | "Fees enter the 2025–2026 Kalshi–Polymarket literature as a flat per-venue rate, a mean exposure, a market-level flag or an unvalued symbol, never as the p(1−p) operator both venues publish; no paper composes fee, tick, lag, weighting and referent into a venue-specific resolution function or tests whether two venues' curves cross." |
| 11 | B/C | NARROWED (→ A/B: the owed search ran) | [Nakajima 2026](https://arxiv.org/abs/2605.21997) arXiv; ESAA 2026 arXiv:2602.23193; Rovai 2026 arXiv:2605.15967; Warmuth & Kuzmin UAI 2006 | "arXiv and OpenAlex return no record pairing 'density matrix' with 'event sourcing', 'append-only' or 'event log'; the 2026 event-sourced-agent cluster derives agent state deterministically from an append-only log but carries no uncertainty object, and Warmuth & Kuzmin (2006) give a density-matrix Bayes rule with no log and no dynamics." |
| 12 | B | NARROWED (wording) | Board&Dice *Entropy* ([review 2026](https://tabletopping.games/2026/06/19/entropy/)); Maxwell's puzzling demon, Steam 2024; Gravina–Liapis–Yannakakis GECCO 2016 | "No game found computes the entropy or information it charges: Entropy (2026) spends entropy as a resource without computing it, the demon games sort temperature without bits, and surprise appears in game research as a search fitness, not a currency paid to a player." |
| 13 | B | NARROWED (two edges) | [Lekić et al. 2025](https://arxiv.org/abs/2506.22137) arXiv / IEEE TMBMC; Bartlett 2024 arXiv:2407.07335 (Causal Leverage Density); Song, Gore & Kleiman-Weiner ICML 2026 (EELMA) | "No paper computes stored or observed semantic information for a software agent, an LLM agent, a repository, a dashboard or an organisation, and none uses a governance event as the viability function." Retired: "first K–W application outside biology"; "first information-theoretic measure from LLM-agent traces." |
| 14 | C | NARROWED (stays C) | [Sang 2025/26](https://arxiv.org/abs/2511.05540) arXiv; Bentley, Lim & Ishikawa ALife 2025; Jehu-Appiah 2026 arXiv:2603.27626; Ay & Löhr 2015 *Theory in Biosciences* | "'Umwelt' is now a live term in AI — a Bayesian Umwelt world model, umwelt-like 'aspects' for LLM agents, a measure-theoretic Umwelt on a sensorimotor loop — and none uses a density matrix, κ/χ maps, dark states or shadow-first egress." |
| 15 | new | B (DARK until the replay runs) | [Sowinski et al. 2023](https://arxiv.org/abs/2304.03286) *PRX Life* (method template); Virgo et al. ALIFE 2025; Weinberger 2024/2026 arXiv:2403.12324 | "With viability chosen as 'lease kept, PR not reverted and truth run green at horizon τ', observed semantic information of a deployed software agent has not been computed by anyone; the design is Sowinski et al.'s (2023) scrambling threshold, the estimation precedent for text traces is EELMA, and the value is DARK until the ledger replay runs." |

### Deadline and venue corrections

| Item | Correction | URL read |
|---|---|---|
| FDG 2027 | Full papers **2027-01-11 23:59 AoE** (not ~01-04); workshops 2026-11-16; games/demos **2027-03-08**. Secondary (LinkedIn); official sub-pages unreadable | [fdg2027.org](https://fdg2027.org/) |
| ISF 2027 | Abstracts **2027-03-07**; Paphos 06-20→23; no TZ printed | [isf.forecasters.org](https://isf.forecasters.org/) |
| CLeaR 2027 | OpenReview duedate **2026-11-17 11:59 UTC**; 12 PMLR pages; one author pledges two reviews | [OpenReview](https://api2.openreview.net/invitations?id=cclear.cc/CLeaR/2027/Conference/-/Submission) |
| AAAI-27 workshops | Loyal Agents: ≤ 8 pp + refs, single-blind. HAIC: 8/4/2 pp, EasyChair. CALE-AI: 11-20, **no TZ**. All in person; pass $305–$557 | [Loyal Agents](https://loyalagents.github.io/aaai27/), [HAIC](https://human-ai-workshop-7p350s.webnode.page/call-for-papers/), [CALE-AI](https://caleai.github.io/) |
| arXiv endorsement | Since **2026-01-21**: auto-endorsement needs an academic email *and* prior authorship in the domain; staff cannot waive | [arXiv blog](https://blog.arxiv.org/2026/01/21/attention-authors-updated-endorsement-policy) |
| Quantum | **€600** default, €100 reduced, or full waiver "no need to justify" | [payment](https://quantum-journal.org/payment/) |
| SciPost | Snippets say academics only; site walled — **UNVERIFIED after search 2026-11-02** | none readable |
| EPJ QT "Quantum Education" | Collection to 2026-12-31; requires **"at least initial results of an empirical evaluation"**, so a DARK-outcomes paper goes to AJP/EJP | [collection](https://link.springer.com/collections/cjbbdcbabf) |
| Datadog | Count monitors without `default_zero()` default to **"Evaluate as zero"**; Grafana and Dynatrace default to a dash | [Datadog docs](https://docs.datadoghq.com/monitors/configuration/) |
| APCs | IJF hybrid **USD 3,500** (subscription route free); ACM Open "waivers are rare"; PRX Life $3,105; J. Phys. Complexity £2,110; Entropy CHF 2,600 with waiver | [Elsevier](https://legacyfileshare.elsevier.com/els_com_pricing/article-publishing-charge.xlsx), [APS](https://journals.aps.org/authors/open-access-physical-review-physical-review-letters), [IOP](https://publishingsupport.iopscience.iop.org/?p=3528), [DOAJ](https://doaj.org/api/search/journals/issn:1099-4300) |
| Journals | Minds & Machines Editorial Manager says "Do not use for live manuscript submission"; FoP single-blind, 19 days; JPM title page needs an institutional address (write "Independent researcher") | [EM mind](https://www.editorialmanager.com/mind/), [JPM](https://www.ubplj.org/index.php/jpm/about/submissions) |
| ALIFE 2027 | Prague 2027-07-19→23; no CFP yet; 2026 pattern 3–8 pp, single-blind, no fee, extension mid-April AoE | [announcement](https://alife.vscht.cz/news/260821announcement) |
| Chaos / ICLR | Only the Bollt topic open (2027-01-31, no TZ); ICLR 2027 workshop list due 11-29 | [Chaos](https://publishing.aip.org/publications/journals/special-topics/chaos/), [ICLR](https://iclr.cc/Conferences/2027/CallForWorkshops) |
| Kalshi fees | 2022 CFTC schedule read as primary (taker-only); Help page 2026-04-19 says **maker fees are now charged**; July 2026 PDF 429, its 0.0175 formula **UNVERIFIED** | [CFTC PDF](https://cftc.gov/filings/orgrules/rule091222kexdcm003.pdf), [Fee Rounding](https://docs.kalshi.com/getting_started/fee_rounding) |

### Findings that change a scaffold's wording

- **P1.** Add Warmuth & Kuzmin 2006 as "the missing classical-side anchor for 'belief = ρ updated by evidence'"; "'Quantum digital twin' is a term collision … and must be disambiguated in print."
- **P2.** "Re-state the claim as 'no peer-reviewed negative result for a quantum filter against a classical blend/persistence on real streams'"; Belavkin 1987 is a CISM chapter.
- **P3.** The watermark "carries no 'dark' distinct from 'not yet', and stops at the engine boundary"; Eaton 2005: "zero correct answers on missing-data questions under the 0-encoded display."
- **P4.** "Shisher & Sun 2022 remains the nearest published negative"; still blocked by the 79/103 re-run.
- **P5.** Retire the "60 of 346" caveat; "Nobody declares the surface and uses it to gate learning"; CALE-AI: "treat as a day earlier than AoE."
- **P6.** State it as "'absence certificate for a streaming estimator, with bit-exact CPU replay' … positioned against Akgul 2025, Atlas 2025, Srivastava 2024 and the SPICE draft."
- **P7.** "A paper must now cite Liardi 2026 as the nearest prior or a referee will."
- **P8.** "This item must be re-searched before a paper claims the gap."
- **P9.** Print "no game evolves a density matrix under open-system dynamics."
- **P10.** "A paper may no longer say 'no game uses entropy as a currency'"; cite Tribus 1961 "for the *word*, not for the measure"; Saha 2021's power is **10³ k_BT/s (1,066)**, not "103".
- **P11.** Retire "fees appear in no abstract"; re-derive the fee code against Kalshi's changed model; cite Fan–Sun–Zhao ("a high price is therefore a sign that an event is likely rather than imminent").
- **P12.** "Cite Bara as the nearest statement and position the contract as the provenance-chain instance"; Duma 2026 (71.58% of review comments agent-authored) and Chowdhury 2026 (45.20% vs 68.37%) are the first population numbers on agents reviewing agents.
- **P13.** Claim "first for a software agent under a governance viability"; rename two of the three η's.
- **Loyal Agents draft.** Single-blind; it cannot share text with any double-blind draft.
- **CLeaR draft.** 12 pages, reviewer pledge, clock 11-17 11:59 UTC; cite Wang 2026, Freeman 2026, Seitzer 2021, CAIAC 2024.

### Searches still not run, for a third pass

- **P12:** Sheridan & Verplank 1978; SWE-bench Verified percentages; `"trust laundering" OR "provenance semiring"`; maintainers' perspective on AI PRs 2026; SPRT-style autonomy promotion; Admiralty code wording; ICD 203; Fetna 2026 (index only).
- **P5:** CASAS recognition with automation-triggered events; `"performative"` × cyber-physical/twin/building; IoT actuator→sensor graphs used for learning; `"self-caused" observations … "learning rate" gate`; Goodhart 1975; Ortega 2021 citer refresh; Yang & Newman 2013 full text.
- **P6 / P8:** holonomy as intrinsic clock; Berry phase in RL decisions; solid angle on the Bloch sphere as a trigger; event-driven intrinsic time in RL; geometric phase in time-frequency processing; Kuramoto lag as a gauge field; Altafini's "gauge transformation" wording; lakeFS internals; ICLR venues of GIN, Li–Arora, Nanda, Git Re-Basin, Entezari, Lim.
- **P7:** `"partial information decomposition" OR "O-information"` × recommender/closed-loop/robotics; the explicit published-null query; venue probes on iopscience, pubs.aip.org, journals.aps.org; Rosas "entropic conjugation" 2025 (UNVERIFIED).
- **P3:** `"data age" perception`; Esri BUG-000177544; Codd 1990 / Date 2005; USENIX/SREcon; QTrail-DB and the watermark paper for the "min of inputs" rule; HCIL-2003-45.
- **P2:** `"quantum particle filter"` tracking; `"quantum-inspired" "state estimation"` 2025; Lanza–Solinas–Mastrogiovanni 2022–2024; Bernoulli/continuous-time HMM presence filters; innovation-based per-sensor reliability; Belavkin in finance/cognition; BITS Pilani item (UNVERIFIED).
- **P1 / P9:** `"density matrix" "world model" agent belief`; quantum Bayesian networks in IoT/twins; quantum-like recommenders; `"open quantum system" … belief decay Lindblad`; `Umwelt robotics von Uexküll`; IEEE IoT-J and *Sensors* contents; QUADRI paper; *Quantum Reports* review model; *Adaptive Behavior* / *Artificial Life* as venues.
- **P9 / P10:** mixed-state Bloch interactives; quantum-games taxonomy 2024–25; Quantum Game Jam 2023; SSH/lattice-gauge games; "scientific accuracy" rubric; CHI PLAY and FDG proceedings; `Landauer game "bit erasure"`; `"information gain" … player pays to observe`; AJP Szilard/Landauer full text; `"Bayesian surprise" game reward`; Hungry Harvest, The Qubit Game, SpinDrops, Quantum Odyssey noise model; Semantic Scholar search (429 throughout).
- **P11:** exact-title LWZ 2007; Ng et al. full text; `"no-trade region"`; `"information share" OR "Hasbrouck"` Kalshi Polymarket; two arXiv aggregation queries; Kalshi July 2026 PDF; JPM ISSN.
- **P13:** "Value of information: a framework for human–agent communication" (title only); Sowinski 2023 full text; CLD ALIFE 2025 abstract; `"information bottleneck" "agent memory"` 2025; S2 graph refresh; *Artificial Life* APC/review; Interface Focus 2027 S2O.
- **Venues:** arXiv treatment of "quantum-like" primaries; SciPost; PCI RR list; CHI PLAY/VIS/ISIT 2027; FDG official sub-pages; MDPI and IOP collection pages; ACM journal APC; IJF Guide for Authors; Frontiers topics on agent trust; negative-results tracks 2027.

### Does the order change?

**No.** P12 → P5 → P6 → P7 → P2 → P3 → P8 → P1 → P9 → P10 → P4 → P11 survives because every finding that moved a judgment adds a citation, not a measurement. P12 keeps first place with four November doors and a theorem to cite (Bara) instead of prove. P5 is stronger: the 575-citer sweep turns its A/B into an A for the CPS instance and CLeaR's clock is exact. P6 and P7 each gained the prior a referee would have produced; P2 gained the peer-reviewed negatives it lacked. The one conditional swap is P8 ↔ P1: P8's gap is "unchanged on evidence, not unchanged after search", so if its five clock queries are not run by the time P7 ships, P1 — whose owed search ran and came back empty — goes first and P8 waits for Chaos on 2027-01-31. P11 stays last with a second blocker: Kalshi's fee model changed under it and the July 2026 schedule is unread. **P13 slots ninth, between P1 and P9**: its only measurement is a replay of events the P12 ledger already holds, its earliest door (ALIFE 2027) follows a mid-April pattern with no CFP yet, and it reads best after P1 has stated the record/belief split it rests on; if the ALIFE CFP posts a March deadline it moves ahead of P1.

### Prophecy, revised where the sweep moved a judgment

Four windows are shorter than the report said. The "record as truth, state as projection" architecture arrived in the agent literature inside 2026 — Nakajima's "The Log is the Agent", ESAA, Rovai, McCann, ChronoMem, StateLedger at NeurIPS 2026 — none peer-reviewed but the last and none carrying an uncertainty object, so P1's early bet is now only the ρ-as-derived-view half; cite the architecture half as settled. The garbling corollary the report expected "by 2028" was written in September 2026 as Bara's no-minting proposition; P12's claim shrinks to the provenance-chain instance and its window to one cycle. P6's "early by roughly two years" holds only for the absence quadrant, since Akgul et al. already commit every training step to a Merkle chain in the presence quadrant. P7's thesis that nobody has written the synergy-loss mechanism is half false after Liardi et al. 2026: the offline, model-fitting analogue is in print, leaving the runtime null and the sign-blindness theorem as the unoccupied ground. Two judgments soften rather than shorten. The `quantum-mind` library (PyPI, October 2026) shows hobbyists building density-matrix belief for robots and self-reporting losses to a Bayes filter, and Fellner et al. 2026 is a peer-reviewed negative for quantum forecasters; the program's "measured negatives" edge is now an edge in *filters on real streams*, not in quantum-inspired ML at large. Quantum cognition stays a foil: its 2022–2026 open-system papers fit no data. Last, the 2026 claims-ledger cluster (Chen et al.'s 18/5/2/1 scorecard, claimkeeper, claim-ledger, Fetna if it verifies) means a DENIED row is no longer "the one thing a reviewer cannot find elsewhere"; what remains unfound is a DENIED row that a referee rather than the author writes, bound to a speaker-class tier and a replayable record. That triple is the sentence every paper in the slate should now lead with.