# Maxwell's demon with a farm: surprisal-priced measurement as a game economy

*Working title:* **The player is the demon: a farming game that pays −kT·log p per harvest, charges the look, and grades its own thermodynamics exact / faithful / suggestive**

Author: [withheld in the repository copy]. Contact: [withheld].

Status: **DRAFT — not submitted. Luke submits by hand, and only if he chooses to.**
Revision: 2026-11-02 second sweep (related work, venues).

## Venues

| Venue | Why | Fee / barrier | Review | Length | Deadline | URL |
|---|---|---|---|---|---|---|
| Entropy (MDPI) | Thermodynamics-of-information section. The only special issue surfaced on 2026-11-02, "Explaining Economic and Social Science Phenomena Through Physical Models, Third Edition" (2027-02-28, no TZ; quantum-like models and decision making), is an econophysics call, not a thermodynamics-of-information one; known only from a search snippet. | APC up to CHF 2,600 (DOAJ record); has_waiver = true, policy unread; mdpi.com returned 403 on 2026-11-01 and again on 2026-11-02, so scope, length and special issues are **UNVERIFIED after search 2026-11-02** | Single anonymous | Not read | Rolling | https://doaj.org/api/search/journals/issn:1099-4300 (registry record; journal page needs a human browser) |
| European Journal of Physics | Pedagogical Lindblad and Zeno already live here. | Hybrid; green OA free; APC £2655 / €3050 / $3665 if OA | Single-anonymous | ≤ 4000 words | Rolling | https://publishingsupport.iopscience.iop.org/journals/european-journal-of-physics/about-european-journal-physics/ |
| American Journal of Physics | The derivation plus design note. Keep it on bookkeeping: AJP is "not an appropriate venue for controversial subjects" including foundations of QM. | No fee under the exclusive licence; OA $3,800 | Double-anonymous | Normally ≤ 6 pages | Rolling | https://www.aapt.org/Publications/AJP/About/editorial_policy.cfm |
| Physical Review E (Statistical physics) | Only if the per-cycle work and efficiency of the closed biome yield a bound or a solvable cycle that is new. | Subscription route: no charge listed; OA $2,910; no waiver for unaffiliated high-income authors; ORCID required | Single-anonymized | No limit | Rolling | https://journals.aps.org/pre/authors |
| Frontiers in Physics RT 87035 "Stochastic Dynamics and Nonequilibrium Thermodynamics in Nonlinear Complex Systems" | Open topical call that fits. | APC with fee support that does not affect acceptance | Single-anonymized; reviewer names published | Original Research 12,000 words | 2027-05-02 (no time zone stated) | https://www.frontiersin.org/research-topics/87035 |
| Quantum Science and Technology, "Focus on Thermodynamics in Quantum Coherent Platforms" | Long shot: the focus issue is ongoing and invited (24 such issues listed, no deadlines; page read 2026-11-02), so the route is a proposal email to the journal. Only if the per-cycle calculation yields a result. | IOP hybrid; APC not read | Not read | Not read | Ongoing, invited | https://iopscience.iop.org/journal/2058-9565/page/Focus-issues |
| arXiv | Primary `physics.ed-ph` (own domain); cross-list `cond-mat.stat-mech` (`cond-mat` domain) and/or `quant-ph`. One endorsement per domain. Since 2026-01-21 automatic endorsement needs an academic email *and* prior arXiv authorship in the domain (policy post read 2026-11-02), so a first-time unaffiliated author takes the personal-endorsement path; staff cannot waive it. | "Independent" accepted | — | — | Daily 14:00 US Eastern | https://info.arxiv.org/help/endorsement.html |

## Abstract (draft)

In a closed quantum system the only irreversible act available to an agent is measurement. A farming game built on an exact density-matrix engine makes that sentence a rule of play. Planting is a unitary Rabi pulse, which cannot raise purity, so a fog cannot be planted. Harvesting is a Born-rule sample, and it pays surprisal: E = −kT·log p, with kT read from the live von Neumann entropy of the field. This is the per-outcome Szilard work for finding a particle in a region of probability p, and its average is the Shannon term in the Sagawa–Ueda bound. Reaping a dephased field pays kT·ΔS, Landauer's price; the three canonical channels are sold as merchant contracts with no reverse trade, which is the second law as a market rule. Opening the world flips the verb: a Zeno-pinned biome is frozen by frequent looks, and in that regime the outcome is nearly certain, so the surprisal paid and the erasure bill both go to zero. Looking spends; watching keeps, and keeping is nearly free. The player's own identity is a 12-qubit density matrix that decays toward maximally mixed with τ = 300 s unless renewed by learning. We grade each mapping exact, faithful or suggestive with the test behind it, defend kT-from-entropy as a Jaynes multiplier rather than a bath temperature, and report that no published game prices the act of looking in bits. Learning outcomes are DARK.

## Contributions

1. **A playable Landauer economy, graded.** Harvest pays −kT·log p (faithful: the exchange rate to credits is a game constant); Landauer reap pays kT·ΔS; purity and entropy are shown on every inspect surface (exact). Each row names its assay.
2. **Measurement as the only irreversible act, as a rule.** The fog-cap theorem p_north ≤ (1 + |r|)/2 enforces that planting never purifies; projective collapse is deterministic-seeded so a saved universe replays; weak measurement decays coherence by √(1 − η). The closed biome is a public toy of the single-temperature measurement engines (Yi, Talkner and Kim 2017; Elouard and Jordan 2018).
3. **kT from the entropy of belief, defended and bounded.** The physics literature holds T as the bath's. We read kT from the field's live entropy and defend it as the Lagrange multiplier of a maximum-entropy inference (Jaynes 1957). We grade it faithful, not exact, and say what would make it exact.
4. **The dial: looking spends, watching keeps.** One rate axis from rare looks (full unitary evolution, high surprisal per look) through anti-Zeno acceleration to Zeno freezing, with the cost axis attached. Elouard et al. (2017, PRL) supply the paradox: in the Zeno regime the erasure cost becomes negligible. Proposition, to be tested with the game's assays: frequent harvest of a frozen biome pays ≈ 0 surprisal, a rare harvest of an evolved biome pays a lot, and the ledger balances by construction. The test is owed.
5. **Identity as an open system.** The faction density matrix decays unless renewed by learning; we read this against Still et al. (2012) (non-predictive memory is dissipated work) and Kolchinsky–Wolpert (2018) (semantic information is what keeps a system viable), and mark it as narrative with a faithful-grade mechanic beneath it.
6. **A negative finding with its caveat.** No published game or interactive simulation prices the look or pays surprisal; every "demon game" found lets the player sort for free. Stated narrowly: no game computes the entropy or information it charges (Board&Dice's *Entropy*, 2026, spends entropy as a resource it never computes; Maxwell's puzzling demon, Steam 2024, has no bits). The finding rests on the first pass's four searches plus the Seskir et al. (2022) review and, since 2026-11-02, ten further searches, ISSN-filtered Crossref checks of AJP, PRPER, Physics Education, EJP, The Physics Teacher, J. Chem. Educ. and Simulation & Gaming, the PhET catalogue, ComPADRE OSP, itch.io and Steam; nine queries were refused by the search budget (listed under Threats), so CHI PLAY and FDG proceedings are covered only through Crossref metadata.

## Outline (EJP/Entropy version ≈ 4500 words)

1. **The demon, the farm, and the claim** (400 words). Szilard's one-molecule engine and Landauer's erasure cost in two paragraphs, plain English first. Then the game's sentence: the player is the demon, the field is the gas, and the price list is the second law. State the three grades and that learning outcomes are unmeasured.
2. **Closed world: measurement is the only irreversible act** (700 words). Exact unitary evolution; purity conserved to machine precision; the fog-cap theorem as a one-line proof that a Rabi pulse cannot raise the north-pole probability above (1 + |r|)/2. Born sampling with deterministic seeds. Why this is a toy of the measurement-fuelled engines (Yi–Talkner–Kim; Elouard–Jordan; Buffoni et al.; Bresque et al.): work comes from the measurement channel, there is no hot bath, and the closed world has no attractors (spectrum conservation; "identity, not destiny").
3. **The price of a look** (700 words). E = −kT·log p as the asymmetric-Szilard payoff k_BT·ln(1/p); its average as the Shannon term in Sagawa–Ueda (2008, 2010), equal to the mutual information only for an error-free measurement. Weak measurement at efficiency η as the imperfect case (Naghiloo et al. 2018: a look can lose information). The exchange rate to credits is a game constant: faithful. Jacobs (2012) and Guryanova, Friis and Huber (2020) as the two sentences the game teaches: looking is never free at finite temperature, and a perfectly sharp look costs infinitely.
4. **kT from live entropy** (500 words). The design choice stated honestly: the field has no bath, so the engine reads kT off the von Neumann entropy of the live state. Jaynes (1957) as the licence. Esposito and Van den Broeck (2011): a confident-but-wrong state carries a relative-entropy work penalty, the thermodynamic reading of "honest uncertainty costs bits". Reeb and Wolf (2014): Landauer is an equality with finite-reservoir corrections, so a finite "bath" of qubits needs its correction stated. Grade: faithful.
5. **Open world: dephasing contracts and the Landauer reap** (600 words). T₂ dephasing, T₁ decay, bistability with two steady states above threshold, EIT dark states. The three canonical channels as merchant contracts with no reverse trade: you can buy dephasing, you cannot sell it back. Landauer reap pays kT·ΔS. Sagawa's (2019) warning that thermodynamic reversibility, logical reversibility and Landauer heat are three distinct things, and which one each contract touches.
6. **Looking spends, watching keeps** (600 words). The rate–cost diagram assembled from Misra–Sudarshan (1977), Itano et al. (1990), Facchi–Pascazio (2008), Kofman–Kurizki (2000), Erez et al. (2008), Jacobs (2012), Latune–Elouard (2025) and Elouard et al. (2017, PRL). The Zeno-pinned biome as the freezing corner; `zeno_assay.py` as the test. The balance proposition of Contribution 4, marked owed.
7. **Identity as an open system** (400 words). The 12-qubit faction density matrix with τ = 300 s. Still et al. (2012) on retained non-predictive bits; Kolchinsky–Wolpert (2018) viability; Sowinski et al. (2023) semantic threshold. Narrative plus a faithful mechanic, not a measured result about players.
8. **Negative space, threats, what this is not** (600 words).

## Related work and the negative space

**Classics and syntheses.** Szilard (1929), *Z. Phys.* 53, 840. Landauer (1961), *IBM J. Res. Dev.* 5, 183, DOI 10.1147/rd.53.0183 (resolved 2026-11-02; the first pass could not resolve it on Semantic Scholar). Bennett (1982), *Int. J. Theor. Phys.* 21, 905; Bennett (2003), *Stud. Hist. Philos. Mod. Phys.* 34, 501. Leff and Rex (1990), *Am. J. Phys.* 58, 201, Resource Letter MD-1, the AJP hub to cite beside Leff and Rex (2003), *Maxwell's Demon 2*. Earman and Norton (1998, 1999), *Stud. Hist. Philos. Mod. Phys.* 29, 435 and 30, 1, the critics (resolved 2026-11-02). Maruyama, Nori and Vedral (2009), *Rev. Mod. Phys.* 81, 1 (Semantic Scholar's "2007" is the arXiv year). Parrondo, Horowitz and Sagawa (2015), *Nature Physics* 11, 131, DOI 10.1038/nphys3230 (resolved 2026-11-02). Seifert (2012), *Rep. Prog. Phys.* 75, 126001. Esposito and Van den Broeck (2011), *EPL* 95, 40004. Goold et al. (2016), *J. Phys. A* 49, 143001. Reeb and Wolf (2014), *New J. Phys.* 16, 103011. Sagawa and Ueda (2008), *PRL* 100, 080403; (2009), *PRL* 102, 250602; (2010), *PRL* 104, 090602; (2012), *PRL* 109, 180602, fluctuation theorem with information exchange (resolved 2026-11-02). Sagawa (2019), tutorial in *Energy Limits in Computation* (Springer), arXiv:1712.06858.

**Experiments.** Toyabe et al. (2010), *Nature Physics* 6, 988. Bérut et al. (2012), *Nature* 483, 187. Koski et al. (2014), *PNAS* 111, 13786. Ribezzi-Crivellari and Ritort (2019), *Nature Physics* 15, 660, the continuous demon that looks at fixed intervals and acts only on change: the closest published object to a watching harvester. Saha et al. (2021), *PNAS* 118, e2023356118 (power figure resolved 2026-11-02 from the PNAS text: "maximum power of 10³ k_BT/s" in the Significance statement, 1,066 k_BT/s as the best measured value in the Discussion; the earlier "103" was a lost superscript, and the ≈10⁴ figure there is an extrapolation, not a measurement); Saha et al. (2022), *PRL* 129, 130601, Bayesian feedback removes the noise-induced dead zone; Saha et al. (2023), *PRL* 131, 057101. Naghiloo et al. (2018), *PRL* 121, 030604.

**Cost of measurement.** Jacobs (2012), *Phys. Rev. E* 86, 040106(R). Guryanova, Friis and Huber (2020), *Quantum* 4, 222. Latune and Elouard (2025), *Quantum* 9, 1614. Elouard, Herrera-Martí, Clusel and Auffèves (2017), *npj Quantum Information* 3, 9. Elouard, Herrera-Martí, Huard and Auffèves (2017), *PRL* 118, 260603.

**Zeno.** Misra and Sudarshan (1977), *J. Math. Phys.* 18, 756. Itano et al. (1990), *Phys. Rev. A* 41, 2295. Facchi and Pascazio (2008), *J. Phys. A* 41, 493001. Kofman and Kurizki (2000), *Nature* 405, 546. Erez et al. (2008), *Nature* 452, 724. Fischer, Gutiérrez-Medina and Raizen (2001), *PRL* 87, 040402, Zeno and anti-Zeno in one unstable system; Streed et al. (2006), *PRL* 97, 260402, continuous versus pulsed Zeno (both resolved 2026-11-02).

**Measurement-fuelled engines.** Yi, Talkner and Kim (2017), *Phys. Rev. E* 96, 022108. Elouard and Jordan (2018), *PRL* 120, 260601. Buffoni et al. (2019), *PRL* 122, 070603. Bresque et al. (2021), *PRL* 126, 120605.

**Prediction, semantics, inference, reward.** Still, Sivak, Bell and Crooks (2012), *PRL* 109, 120604. Kolchinsky and Wolpert (2018), *Interface Focus* 8, 20180041. Sowinski et al. (2023), *PRX Life* 1, 023003. Jaynes (1957), *Phys. Rev.* 106, 620. Achiam and Sastry (2017), arXiv:1703.01732 (NIPS 2016 Deep RL workshop); Pathak et al. (2017), ICML: surprisal as intrinsic reward, never tied to a temperature. Gravina, Liapis and Yannakakis (2016), GECCO, 677, DOI 10.1145/2908812.2908817: surprise as a search fitness for generating content, the nearest game-research use of surprise as a reward signal, and still unpriced in temperature. Tribus (1961), *Thermostatics and Thermodynamics* (Van Nostrand), pp. 64–66: credited with the word "surprisal" only; the log-probability measure is Samson's (1951), so the paper does not write "coined" without that caveat (book not read; exact wording UNVERIFIED after search 2026-11-02).

**Games.** Seskir et al. (2022), *Optical Engineering* 61(8), 081809. Jensen et al. (2021), *Phys. Rev. Research* 3, 013057. du Buisson, Sivak and Bechhoefer (2024), *Adv. Phys. X* 9(1), the information-engine review. Added 2026-11-02 as neighbours (none prices a look): Timberlake (2010), *The Physics Teacher* 48, 516, DOI 10.1119/1.3502501, with the companion OSP model "Ideal Gas with Maxwell's Demon" (ComPADRE 10096); Criado and López (2004), *J. Chem. Educ.* 81, 1679, a pedagogical demon simulation; Davies, Thomas and Zahariade (2021), *Am. J. Phys.* 89, 1123, the harmonic quantum Szilárd engine for students; Chattopadhyay and Paul (2025), *Quantum Inf. Process.* 24, DOI 10.1007/s11128-025-04918-z, the current pedagogical review of k_BT ln 2; Aste (2026), *Phys. Educ.* 61, 045051, a school-level bridge from information-theoretic to thermodynamic entropy; Maxwell's puzzling demon (muratsubo Games, Steam, 2024-08-29), a temperature block-puzzle with no bits; *Entropy* (Board&Dice; Battista, Luciani, Mangone; reviewed 2026-06-19), a board game in which entropy is gained and spent as a resource.

**Negative space.** Nearest neighbours and what they do not do. The SFU information engine (du Buisson et al. 2024) is a priced measure-and-act loop, but an apparatus, not a thing a learner can turn. Quantum Moves 2 scores control fidelity, not an information cost. The itch.io demon games and IN2UB's "Saved by the Paradox" let the player sort for free with no bit accounting; the ComPADRE "demon" is Creutz's thermometer, not Maxwell's. The RL curiosity line pays −log p but has no temperature and no erasure bill. In physics, kT is the bath's; nobody reads it from the entropy of a belief. The Zeno regime, the anti-Zeno regime, the cost per look and the vanishing erasure bill are all published, but separately; no paper puts information per look and cost per look on one axis for a teaching audience. The caveat is narrower after the second sweep: Board&Dice's *Entropy* (2026) spends entropy as a resource and Maxwell's puzzling demon (Steam 2024) is a temperature puzzle, so the sentence to write is "no game computes the entropy or information it charges", never "no game uses entropy as a currency"; nine queries remain unrun.

**Second sweep (2026-11-02).** The three queries the first pass could not run ("information engine" game; entropy as currency board game; Maxwell's demon serious game) were run, with seven more (Landauer interactive teaching; "kT ln 2" classroom activity; surprisal as in-game currency; PRPER Maxwell's demon simulation; semantic information for software agents; Tribus 1961), plus Crossref bibliographic queries filtered by ISSN for AJP, PRPER, Physics Education, EJP, The Physics Teacher, J. Chem. Educ. and Simulation & Gaming, the PhET metadata catalogue (no simulation mentions entropy, Maxwell's demon, Landauer or Szilard), ComPADRE OSP (17 demon items, none with an information cost), itch.io tag pages and Steam. Verdict unchanged in substance, narrowed in wording: *Entropy* (Board&Dice, 2026) makes entropy a resource gained from creation tokens and spent on actions, with no information, measurement or probability mechanic; Maxwell's puzzling demon (muratsubo Games, Steam, 2024-08-29) is a block-pushing puzzle where blocks have temperature and no bits; so the defensible claim is "no game computes the entropy or information it charges". PRPER returns nothing for "Maxwell demon" (ISSN-filtered Crossref), confirming the first pass. New neighbours, all verified at Crossref unless flagged: Timberlake (2010), *TPT* 48, 516, with the OSP demon model (demon on/off, velocity reversal, two-chamber temperatures; no information, Landauer or memory cost); Criado and López (2004), *J. Chem. Educ.* 81, 1679; Davies, Thomas and Zahariade (2021), *AJP* 89, 1123, an exactly solvable harmonic quantum Szilárd engine written for students — the teaching-level physics behind a measurement-fuelled engine, still not an artefact; Chattopadhyay and Paul (2025), *QIP* 24; Aste (2026), *Phys. Educ.* 61, 045051; Leff and Rex (1990), *AJP* 58, 201, Resource Letter MD-1; Gravina, Liapis and Yannakakis (2016), GECCO, surprise as a search fitness rather than a currency paid to a player. Citations resolved: Saha et al. (2021) reads "maximum power of 10³ k_BT/s" in its Significance statement and gives 1,066 k_BT/s as the best measured value in the Discussion (the "103" was a lost superscript in a plain-text rendering; the ≈10⁴ k_BT/s there is an extrapolation to a smaller bead, not a measurement); Parrondo, Horowitz and Sagawa (2015), DOI 10.1038/nphys3230; Landauer (1961), DOI 10.1147/rd.53.0183; Sagawa and Ueda (2012), *PRL* 109, 180602; Earman and Norton (1998, 1999), *Stud. Hist. Philos. Mod. Phys.* 29, 435 and 30, 1; Maruyama, Nori and Vedral (2009), *RMP* 81, 1; Streed et al. (2006), *PRL* 97, 260402, and Fischer, Gutiérrez-Medina and Raizen (2001), *PRL* 87, 040402, the two experimental anchors for Contribution 4; Bruineberg et al. (2022), *Behav. Brain Sci.* 45, e183; by Crossref issue date Baldi and Itti is 2010, Moore and Schroeder 1997, Misra and Sudarshan 1977. "Surprisal": Tribus (1961), *Thermostatics and Thermodynamics*, pp. 64–66, is credited with the word; Wikipedia's "Information content" entry credits the log-probability measure itself to Samson (1951); so cite Tribus for the word and not for the measure, and qualify "coined" (book not read; exact wording UNVERIFIED after search 2026-11-02). Venues: the QST focus issue "Thermodynamics in Quantum Coherent Platforms" is ongoing and invited (proposal by email; a long shot); Quantum's €600 fee is waivable without justification but its "very significant" bar stands; the only Entropy special issue surfaced, "Explaining Economic and Social Science Phenomena Through Physical Models, Third Edition" (2027-02-28, no TZ), is an econophysics call known from a search snippet (mdpi.com 403; UNVERIFIED after search 2026-11-02); Interface Focus theme issues remain unchecked (journal page 403). Nine queries were refused by the session's search budget and are NOT RUN: `Landauer game itch.io OR steam "bit erasure"`; `"information gain" game mechanic player pays to observe hidden state bits currency`; `"Physics Education" journal "Maxwell's demon" app OR game OR simulation`; `"American Journal of Physics" Szilard engine OR Landauer erasure teaching`; `PhET simulation entropy OR "Maxwell's demon"`; `CHI PLAY OR FDG OR "Simulation & Gaming" game "information theory" entropy mechanic`; `"Bayesian surprise" OR "surprise search" game design reward`; `"Interface Focus" theme issue 2026`; `quantum thermodynamics game "Szilard engine" interactive demon play bits`. Crossref metadata is the only coverage of CHI PLAY and FDG, and Semantic Scholar search was rate-limited on every attempt, so the negative finding is wider than before but not closed.

## Evidence (from the dossier; every row carries its tier)

| Claim | Pointer | Tier |
|---|---|---|
| Each biome is a density matrix evolved exactly, U = exp(−iH·dt), eigendecomposition in native C++; purity conserved to machine precision | SpaceWheat `Core/QuantumSubstrate/`, `native/src/`, `QuantumComputer.evolve()` | exact (owner's grade) |
| Born rule + projective collapse, deterministically seeded; weak measurement with coherence decay √(1 − η), η = 0 and 1 tested | `Core/Actions/`, `tests/test_drain_qubit.gd` | exact |
| Purity and von Neumann entropy on every inspect surface; kT computed from live entropy; mutual-information edges | inspect overlays, M-Graph | exact |
| Surprisal economy: harvest pays E = −kT·log p; kT from live entropy; the credit exchange rate is a game constant | `EnergyPricing` | faithful |
| Player identity: ρ over a 12-qubit faction concept space decaying toward maximally mixed with τ = 300 s unless renewed by learning | `FactionDensityMatrix` → native `QuantumMythosEngine` | faithful |
| Spectrum conservation; dominant eigenstate labelled identity not destiny (closed systems have no attractors); absence of chaos taught as a theorem | Chapter 2 and 4 arcs | exact |
| T₂ dephasing, T₁ decay, Zeno pinning, bistability (two steady states above threshold verified headlessly), EIT dark states, Landauer reap paying kT·ΔS, fog-cap theorem p_north ≤ (1 + |r|)/2, three canonical channels as merchant contracts (no reverse trade) | `tools/channel_assay.py`, `transition_assay.py`, `plant_assay.py`, `zeno_assay.py`, `eit_assay.py` | exact / faithful per ledger |
| GKSL dissipators present; counts drift between README (164 / 66) and FOR_PHYSICISTS (64 of 162) — **re-derive with `tools/atlas_plates.py` before quoting** | `LindbladBuilder`, `biomes.json` | exact within the model; count unresolved |
| Honesty ledger: three grades, conservative, nothing suggestive presented as exact | `docs/FOR_PHYSICISTS.md` | the artifact |
| Verification: headless smoke suite; `tests/` 221 entries; engine-free Python assays exit nonzero on failure; atlas re-derives the thermodynamic map | repo | PINNED (repo CI) |
| Design history: role separation H vs L; dephasing first; openness is a place | `docs/FARMING_A_DENSITY_MATRIX.md` | narrative |
| Frequent harvest of a Zeno-frozen biome pays ≈ 0 surprisal; rare harvest of an evolved biome pays much; the ledger balances | — | **owed** (proposition; assay to be written) |
| Per-cycle work and efficiency of the closed biome in the game's units | — | **owed** |
| Learning outcomes, playtest counts, engagement | — | **DARK** |

## DARK (do not type a number)

- Learning outcomes; whether any player's model of entropy changed.
- Measured conversion ratios or efficiencies of the game's cycles; the comparison with Toyabe (2010) or Koski (2014) is a plan, not a result.
- The balance proposition of Contribution 4 until the assay exists.
- Any claim that kT-from-entropy reproduces a bath result; it is a design choice with a licence, graded faithful.
- Biome and channel counts until re-derived.

## Threats to validity

- **A game constant hides in the economy.** The exchange rate from surprisal to credits is a constant; "literally Landauer bookkeeping" holds up to that constant, and the abstract must say so.
- **kT is not a bath temperature.** Jaynes licenses a multiplier, not a thermometer. A referee may call the reading suggestive rather than faithful; show the licence and its limits (Reeb–Wolf finite-reservoir corrections).
- **Self-graded ledger.** The author wrote the grades and the tests. Engine-free assays and a QuTiP cross-check are the mitigations; independent re-runs are the remedy.
- **The negative finding is wider but not closed.** The first pass's four searches plus one review; "information engine game", "entropy as currency board game" and "thermodynamics game Physics Education" were run on 2026-11-02 with seven more and the ISSN-filtered Crossref, PhET, ComPADRE, itch.io and Steam checks. Nine queries were refused by the search budget and are NOT RUN (listed in the second sweep), CHI PLAY and FDG are covered only through Crossref metadata, and Semantic Scholar search never answered; a board game or a short Physics Education note could still exist unindexed.
- **AJP's scope flag.** Keep the paper on bookkeeping and mechanics, with no interpretational claim about what ρ "is".
- **One artifact, one author.** No player data, no comparison game.

## What this paper is not

- Not a learning-outcomes study. Outcomes are DARK.
- Not the engine-and-ledger paper (honesty-graded-physics-game); that scaffold owns the rubric, the engine verification and the pre-registered study. This one owns the thermodynamics.
- Not a new bound or engine cycle; if one appears during the per-cycle calculation, PRE becomes an option and the paper changes shape.
- Not an interpretation-of-quantum-mechanics paper. The density matrix is the state of a simulated system; nothing is claimed about minds.
- Not a claim that any player is a thermodynamic demon; the mapping is to the mechanics.

## Luke's steps, if he wants to submit

1. Have a seat write the balance assay: for a Zeno-pinned biome and an evolved biome, log surprisal paid per harvest and the kT·ΔS reap across harvest rates; commit it under `tools/` with the other engine-free assays so it exits nonzero on failure. Only then does Contribution 4 leave "owed".
2. Have a seat compute per-cycle work and efficiency in the game's units for one closed biome, exchange constant shown, and decide whether the result is a design note (EJP/AJP) or a result (PRE).
3. Run `tools/atlas_plates.py` and fix the biome and channel counts.
4. Register an ORCID. Register at arXiv as "Independent". Since 2026-01-21 automatic endorsement needs an academic email and prior arXiv authorship in the domain, so the personal-endorsement path is the one that applies, and arXiv staff cannot waive it. Start a submission in `physics.ed-ph` to trigger the endorsement email; for `quant-ph` find an endorser via "Which authors of this paper are endorsers?" on arXiv:1702.01917 (Elouard et al.); for `cond-mat` on arXiv:2402.17860 (du Buisson et al.). One endorsement per domain.
5. Choose the journal. EJP: ≤ 4000 words, single-anonymous, green OA, no fee. AJP: anonymize fully, ≤ 6 pages, propose to the Computational Physics editors first. Entropy: open mdpi.com in a browser, confirm the APC, the waiver and whether a thermodynamics-of-information special issue is open.
6. If targeting Frontiers in Physics RT 87035, submit before 2027-05-02 (no time zone stated) and request fee support at submission.
7. Done 2026-11-02: Saha et al. (2021) read from the PNAS text; quote "10³ k_BT/s" (Significance) or 1,066 k_BT/s (best measured, Discussion), never "103". Remaining: if the QST focus issue is wanted, email the journal a proposal; it is invited-only.
8. Log the submission in the authority record when there is a receipt.