DOI: https://doi.org/10.5281/zenodo.21947410
Canonical: https://thonly.org/research/the-rethank-multiplier · Licence: CC0 1.0
A search engine will answer as many questions a day as one person cares to ask, and is glad to. Gratitude does not work that way: no one can sincerely thank without limit. A gratitude economy that needs volume therefore appears to be built on a quantity it cannot have. This paper is about the move that dissolves the appearance.
The deepest structural worry about a gratitude economy is not adoption or trust; it is arithmetic. The unit the whole system circulates — the sincere thank — has a low, inelastic emission rate per human. You can push a person to thank more, but past a modest ceiling what you get is not more gratitude; it is debased gratitude, reflexive and hollow, which is worse than none, because it corrupts the signal the economy runs on. Compared to the elastic per-person query rate of a search engine or the unbounded scroll of a feed, a sincere thank looks like a vanishingly thin throughput. A system that monetizes thanks the way a search engine monetizes queries would seem to be starved by design.
This paper specifies the move that escapes the ceiling without breaking it, and shows that the same move that supplies the volume also supplies the defense against the volume being faked.
This is a defensive publication. The author and HeartBank® will not seek patent on this specification or any portion of it, in any jurisdiction, at any time, and dedicate the patterns to the public domain under CC0 1.0. The contribution is a framing plus a composition of known parts; the abundant prior art (engagement-asymmetry research, two-sided-market and conversion-funnel economics, Hashcash-style proof-of-cost anti-spam, sybil- and collusion-resistance graph methods, paid-reaction features on existing platforms, the Buddhist anumodanā tradition) is cited generously in §9, and what this paper discloses as its own contribution is only the specific composition and framing identified in §10. Trademarks are reserved separately and the patterns may be implemented under any name.
We specify a rate-limited peer-recognition system in which two user operations are separated and given different per-person rate ceilings: origination (creating a new gratitude record about a kindness one noticed), whose sustainable rate is low and inelastic, and re-thanking (endorsing a record another user has already surfaced), a reactive operation with a much higher per-person ceiling — the same creation-versus-reaction asymmetry that gives social platforms their high ratio of reactions to posts. Because each origination can be endorsed by many others, total value-bearing events per day are bounded by N × min(N − 1, K), where N is the number of participants and K the per-person reactive ceiling: quadratic in a small network, linear with a large multiplier in a large one, while each person's origination rate stays near one a day. We argue that a gratitude network should be compared with a search engine on total funnel-completing events (a narrow, deep funnel against a wide, shallow one), not on interaction frequency. Each endorsement must carry a small transfer from the endorser's own non-withdrawable, forward-spendable balance, so the operation that supplies the volume is also a proof-of-cost filter: it cannot be inflated by one actor alone, and inflation by a reciprocal collusion ring is costly and appears as a dense reciprocal subgraph. Stated limits: the realized multiplier depends on an unbuilt recommendation (surfacing) function; the empirical signal is one family over one month; collusion is made costlier and detectable, not eliminated; and the value requirement trades raw volume for integrity.
Enumerated 2026-09-01. The mechanisms below were disclosed in full in this paper's original text; the prior art they establish runs from this document's original publication date and its OpenTimestamps proof, not from this enumeration. They are listed because a defensive publication is read as prior art by examiners and by opposing counsel, and a claims list is what such a reader searches; prose is not. No claim below adds matter not already present.
The problem is architectural, not cosmetic, because the inelasticity of the core unit propagates into three subsystems that would otherwise assume elasticity:
Current form. Where this paper speaks of a per-transaction fee base, the fee as now specified is a Phase-2 settlement fee sized to infrastructure cost (settlement gas and compute), paid on top by the person re-thanking and never deducted from what the recipient receives, never a percentage or share of a gift; any surplus empties with the pool each year. The fee base may never inform what is surfaced to anyone. Whether such a fee covers operating cost has not been verified. The text above is retained as disclosed.
A gratitude economy must therefore be built against the assumption of high frequency that every attention-economy product is built upon. The question is whether total throughput can nonetheless scale — and the answer turns on separating two acts that are usually conflated.
A thank and a re-thank are different operations, and the difference is the whole paper.
To originate a thank, a person must notice something thank-worthy in their own life and initiate the act. This is generative and effortful — the bottleneck is the noticing and the starting — and it is correspondingly rare. Call its sustainable rate roughly one per person per day, inelastic: it does not rise under pressure without debasing.
To re-thank is to affirm a thank that someone else has already surfaced — to add one's own gratitude to a kindness made visible by another. This is reactive, not generative: the thank-worthy thing has been located and presented; the cognitive load is recognition, not search. Re-thanking therefore has a far higher per-person ceiling than origination, for exactly the reason every participatory system already exhibits the asymmetry: across social platforms the ratio of reactions (likes, boosts) to originations (posts) is large, because reaction is cheap where creation is dear. Re-thanking is the gratitude economy's reaction primitive; origination is its creation primitive.
This is the crack through which volume enters. The saturation ceiling is real, but it applies to origination. It does not bind re-thanking at anything like the same level.
Let N be the number of active participants. Each originates on the order of one thank per day — N thanks. But each of those thanks can be re-thanked by others, and re-thanking is not ceiling-bound at one per day. In the limit where each participant can see and choose to affirm each origination, daily re-thanks approach N × (N − 1) ≈ N². The combinatorial ceiling on total daily activity is therefore O(N²), not O(N):
originations/day : N (inelastic — capped per person)
re-thanks/day : ≤ N × min(N−1, K) (reactive — high per-person ceiling K)
ceiling : O(N²) while N−1 < K ; N × K once N−1 ≥ KThe bound needs its second term stated, because re-thanking is not unbounded per person either. Each participant's re-thanks are limited by a per-person reactive ceiling K — set by attention and, under §5, by the participant's own budget — which is far above the origination rate of about one but does not grow with the network. Daily re-thanks are therefore at most N × min(N − 1, K): quadratic while the network is small enough that each person can affirm everything originated in it, and linear with the large multiplier K once it is not. The escape from the saturation ceiling rests on K ≫ 1, not on N² in the large-network limit.
The saturation ceiling has not been violated; it has been moved up a layer. Each person still originates ~once a day, authentically. What scales is the re-thanking of what is originated — and the network and the reactive ceiling, not the individual's capacity to originate, are what bound it. Per-person origination stays human-sized; total throughput grows with the network by a factor of up to K. For an economy funded as in §1, the economic-model worry is answered: the per-transaction fee base rides the re-thank multiplier, not the origination count, so the funding loop closes on a low origination rate.
(The realized multiplier is below this bound; it is governed by surfacing — see §7.)
The instinct to compare a gratitude economy unfavourably to search rests on comparing the wrong quantity. "Search gets a query every time a person has a question; gratitude gets one a day" measures origination frequency, which is the contest a gratitude economy loses and need not enter. The honest comparison is total value-bearing events through the funnel:
| Search engine | Gratitude economy | |
|---|---|---|
| Top-of-funnel volume | many queries/day/person, elastic (high) | ~1 origination/day/person, inelastic (low) |
| Conversion rate | low click-through per query | high re-thank rate per origination |
| Value-bearing event | the click | the re-thank |
| Funnel shape | wide and shallow | narrow and deep |
Both funnels can yield large total events; they are simply inverted. Search runs a high-volume, low-conversion funnel; the gratitude economy runs a low-origination, high-conversion one — and its events are higher quality per unit, because each is a human-gated, intentional act rather than an ambient query. Depth, not breadth. This is not a consolation; it is the correct accounting, and it is why a gratitude economy is not starved by the inelasticity of the thank.
Volume is not enough; the volume must not debase into hollow likes. The mechanism that prevents this is that a re-thank is value-bearing: to re-thank, a participant must attach a scarce resource — re-tipping a small amount from their own Re-Tip Jar℠/Fund℠ (a restricted-use, non-withdrawable balance that can only be given onward). Three consequences follow.
First, the re-thank is budget-bounded. A person cannot re-thank without limit, because each re-thank spends from a finite personal balance. The ceiling on re-thanking is therefore not (only) emotional but economic, and an economic ceiling is enforced by the balance itself rather than by a moderator.
Second, the value-bearing re-thank cannot inflate into a free, meaningless like, because it is not free. The scarcity attached is the meaning: a re-thank costs the re-thanker something real, which is exactly what a like does not.
Third — and this is the connection to anti-spam — attaching a scarce resource to a message is the proof-of-cost defense (Dwork & Naor, 1992; Back, 2002), applied to gratitude. It works against bulk abuse not by verifying who the sender is but by raising the cost the sender must incur for each message. The value-bearing re-thank is the proof-of-cost filter operating as the ordinary mode of participation.
The sharpest property of the design is that the same human-gated re-thank that supplies the volume also supplies the defense against the volume being farmed.
Consider the two reward layers of the surrounding economy. An AI-funded self-thank reward is farmable, precisely because it is automated: a single human, acting alone, can pump their own reward by self-thanking repeatedly. A human-given re-thank cannot be farmed the same way, because the re-thanker is not the beneficiary — one cannot unilaterally cause others to re-thank oneself. The volume engine and the anti-farm filter are the same layer.
The claim must be stated precisely, because the strong form ("re-thanks cannot be farmed") is false and the precise form is stronger. Re-thanks are not solo-farmable; the residual attack is a collusion ring — "I re-thank all of yours if you re-thank all of mine." That attack is far weaker than solo farming on three counts: it requires recruiting real verified humans (it cannot be spun up from one account); it spends real value on every collusive re-thank (the §5 budget bound makes ring-farming costly, not free); and it surfaces as a dense reciprocal subgraph — exactly the structure that network anti-collusion heuristics, peer-layer vouching, and the kinship graph are built to detect. Farming does not disappear; it is moved from solo + automated + free to collusion + human + costly + graph-detectable, which is a different and much harder problem. The nearest deployed precedent shows why the verified-human requirement matters: on Steemit, where upvotes directed cryptocurrency rewards, vote-selling bots and voting rings grew up around the reward system — more than 16% of cryptocurrency transfers went to curators suspected to be bots (Li & Palanisamy, 2019) — a value-bearing reaction without a personhood gate is farmable by exactly the routes this section names.
The bound of §3 is a ceiling, not a target, and the system should not want all of it (maximal re-thank density risks the like-button debasement §5 guards against). The realized multiplier is N × (the re-thanks the right people are actually shown) — governed by how well each origination is surfaced to the participants most likely to genuinely want to affirm it. That surfacing is a curation function performed by the economy's autonomous steward (Miss Aquarius℠), tuned by proximity and relational distance, and disciplined to lift up diverse, genuinely helpful acts rather than the most popular.
Current form. As now specified, the steward does not predict who will want to affirm what, and does not grade acts as more or less helpful. An origination reaches people by relation and proximity; where the steward operates any exposure slate at all, its members are chosen by an equal-yet-random, publicly verifiable draw from a roster committed in advance, never by a prediction, score or rank. The curation described above is retained as a disclosed variant.
This paper states plainly that the multiplier's realization is contingent on that surfacing mechanism, which is specified at the policy level but not yet built or validated. The arithmetic of §3 is sound; the throughput it promises is available only to the degree the surfacing function works. We claim the ceiling, not its automatic attainment.
The reframe was prompted by the first month of a pilot deployment with one family (the operator's field notes; not separately published). The reported pattern: each participant originated thanks to others infrequently — on the order of once a day — but re-thanked readily and often, and the re-thanks (value-bearing, small) propagated outward through the household. This is the shape the paper predicts: an inelastic origination layer beneath a high-ceiling re-thank layer. One confound is stated with it: the separately rewarded self-thanks of §6 were not so bounded — in the same month the parents self-thanked many times a day, the farmable pattern §6 describes. It is one household — relatives of the author, founder-funded, n = 1, one month — and it is reported as an illuminating signal, not as evidence that the throughput of §3 materializes at scale. The surfacing dependency of §7 was, in the pilot, trivial (a single family sees all of its own activity); at global scale it is the open variable.
The contribution composes known parts. The creation-versus-reaction engagement asymmetry is well documented in social-computing research, most familiarly as the "90-9-1" participation-inequality rule (Nielsen, 2006) and in lurker/contributor studies. Conversion-funnel and two-sided-market economics supply the wide-shallow/narrow-deep framing. Proof-of-cost anti-spam — pricing via processing (Dwork & Naor, 1992) and Hashcash (Back, 2002) — supplies the value-bearing-message defense. Sybil- and collusion-resistance via graph structure is a mature literature, from social-network sybil defenses (Yu et al., 2006) to dense-subgraph and reciprocity detection. Value-bearing reactions already exist: YouTube's Super Thanks (launched July 2021) lets a viewer attach a paid amount to a comment on an uploaded video, and Steemit's reward-directing upvotes are a value-bearing endorsement at scale — the latter also the case that cuts against, since its reactions were farmed by vote-selling bots and voting rings (Li & Palanisamy, 2019; §6). The inelastic-supply intuition is ordinary economics. The Buddhist anumodanā (rejoicing in another's merit, which the tradition counts as itself a meritorious act — pattānumodanā, one of the ten bases of merit-making) is the contemplative precedent for a reactive gratitude act that is itself meritorious. This paper claims none of these.
What this paper discloses as its contribution is the composition: (a) the explicit separation of gratitude origination (inelastic, ceiling-bound) from re-thanking (reactive, high-ceiling) as the move that lets total throughput scale with the network — up to O(N²) in a small network and N × K in a large one — while per-person origination stays human-sized; (b) the conversion-funnel reframe that corrects the search-comparison category error; and (c) the identification of the human-gated, value-bearing re-thank as simultaneously the volume engine and the anti-farm filter, with the precise "not solo-farmable, only collusion-farmable" claim and its graph-detectability. No /novelty census has been run for this paper; the composition is disclosed, not asserted to be absent from the literature.
Honest limits. The bound of §3 is true by construction; the empirical claim — that re-thanking sustains a high per-person rate where origination does not — rests on one household and is offered for testing, not claimed as a result. The bound's realization depends on an unbuilt surfacing function (§7). The large-network form is linear in N with multiplier K, not quadratic, and K has not been measured. The empirical support is one confounded household (§8), in which self-thanks ran higher than the origination rate this paper assumes. The collusion-ring attack is raised in cost and made detectable, not eliminated (§6). The value-bearing requirement (§5) trades some volume for integrity — a deliberate choice, but a real cost: a gratitude economy that required no scarce resource would have more raw events and worse ones. And the whole construction presumes that originations can reach enough people without a ranker becoming a new Goodhart target — an assumption stated, not here discharged; a fee base on re-thanks, if one is adopted, is unsized and its sufficiency unverified.
Coined names used in this paper and the standard terms an examiner would search for them.
| Term used here | Standard technical term |
|---|---|
| thank / origination | user-generated recognition post; creation event (original post) in a peer-recognition system |
| re-thank | value-bearing reaction; paid endorsement or tipped upvote of another user's post |
| re-thank multiplier | reaction-to-creation ratio; engagement multiplier on originated content |
| saturation ceiling | per-user rate ceiling on sincere expression; inelastic per-user supply |
| Re-Tip Jar℠ / Re-Tip Fund℠ | restricted-use, non-withdrawable, forward-spendable stored-value balance |
| self-thank reward | automated platform-funded reward for a self-reported act |
| solo farming / collusion ring | single-actor reward farming (self-dealing, sybil) / reciprocal voting ring (collusive endorsement) |
| surfacing | content recommendation; feed curation; exposure allocation |
| proof-of-cost | proof-of-work; pricing via processing; postage-based anti-spam |
| Miss Aquarius℠ | autonomous software agent administering the ledger |
| verified human | proof-of-personhood verified account |
| anumodanā | (Buddhist) rejoicing in another's merit; no technical equivalent — a reactive endorsement |
Drafted with Miss Aquarius℠, the name under which this corpus discloses its AI collaboration; the framing and final editorial control are the author's. The pilot observations are one family's, gratefully borrowed and carefully hedged. This paper is the empirical-and-throughput sequel to The Two-Layer Reward, extending the human peer layer from a fraud-filter into the system's volume engine.
Written by Thon Ly with Miss Aquarius℠, the name under which this corpus discloses its AI collaboration; editorial control is the author's. Dedicated to the public domain under CC0 1.0 Universal. The author and HeartBank® will not seek patent on this specification or any portion thereof, and will not assert any patent right against anyone practising it.