The VIBE Ledger: Immutable Sensemaking and Regenerative Capacity

Version 5.0 Status: Current owner-directed economic architecture

V7 implementation boundary (2026-10-05): factual QUALIA and evaluative VIBE now use separate native positions, budgets and wallets. See unified claims v2 and V7 product and forecasting boundary. The six-component Sepolia generation is finalized. Earned settlement remains paused. Forecast confidence, proper scoring and event resolution are an adjacent, unqualified extension, not a new multiplier in this whitepaper's reward formula.

For the deployed V7 development generation, a mixed review has one immutable position per funded capacity, with lane-specific subject rank and correction lineage. VIBE endorsement/challenge costs1/2; QUALIA assertion/denial costs1/1. The actual test implementation uses hyperbolic wallet and stake decay, not the experimental exponential family discussed below. Its exact versioned manifest, integer rounding and historical reads govern execution. The architectural paper does not replace those release bytes or qualify earned rewards. Decision date: 2026-08-13 Production target: Base L2

This paper supersedes the active economic interpretation of V3, V4, and the V4.1 draft. Those files remain historical design records. They are not deleted or silently rewritten.

Status vocabulary

Every mechanism in this paper has one of four statuses:

• Owner-settled invariant: structural protocol direction. Simulator work may test it but may not replace it with a contradictory mechanism.
• Working mathematical function: the current family used to make the structure precise and testable. It is not a permanently approved implementation or coefficient set.
• Simulator parameter: an explicit experimental input that must appear in the run manifest.
• Open protocol decision: implementation or calibration still requiring evidence and approval.

Passing a simulation does not promote a working function, parameter, or open decision to protocol law.

Abstract

Knomee gives a verified participant scarce, perishable VIBE capacity for making consequential judgments about any stable, canonically identified entity. Accepted judgments become permanent, signed, replayable history. A VIBE committed by one participant has the same direct protocol weight as a VIBE committed by any other participant. Reputation does not multiply a stake. It changes the participant's future opportunity to receive and spend VIBE.

Each settled stake creates an immutable historical position, or jar, with a fixed owner, subject, semantic commitment, original amount, time, ID, provenance, and queue rank. The live VIBE inside that jar may decay and may be restored with fresh VIBE, but it can never exceed the jar's original size. Reaffirmation preserves one position; it is not a new independent judgment.

Later independent semantic flow may confirm or contradict earlier positions. Confirmation can improve an early sensemaker's reputation, which can increase future VIBE replenishment. This "dividend" is new future expressive capacity, not a transfer from the later staker and not a conserved payout pool. The system is designed to rank useful sensemaking, not to conserve a fixed token pot.

1. Product and protocol boundary

The following are owner-settled invariants:

1. Base L2 is the production target. Sepolia is development and rehearsal only.
2. Daily Knomee replenishment is part of the ordinary product loop.
3. A participant spends scarce VIBE on judgments as they arise.
4. Accepted judgments are immutable, signed, replayable history.
5. Judgments may concern any stable, canonically identified entity.
6. Reputation is derived from immutable history and changes future VIBE capacity or replenishment.
7. Oracles replay canonical history through an exact cutoff block under a frozen deterministic projector. They do not adjudicate individual claims.
8. Changed positions append new claims and stakes. Historical commitments are never withdrawn, erased, refunded, overwritten, or silently netted away.

The human phrase Daily Knomee describes a human-scale replenishment flow. It does not require a midnight reset, a discrete faucet, or a fixed 24-hour implementation interval.

2. One VIBE has one direct weight

Owner-settled invariant. For current live stake, one VIBE contributes one unit of direct protocol weight regardless of the staker's reputation.

For two positions with equal live amounts:

[ L_i(t)=L_j(t) \quad\Longrightarrow\quad \operatorname{DirectWeight}_i(t)=\operatorname{DirectWeight}_j(t) ]

The protocol must not calculate:

[ \operatorname{DirectWeight}_i=L_i\times\operatorname{ReputationMultiplier}(u_i) ]

Reputation instead affects future flow:

[ \text{useful sensemaking}\rightarrow\text{higher derived reputation} \rightarrow\text{more future VIBE opportunity} ]

This avoids the recursive accelerator in which reputation grants more VIBE and also makes every granted VIBE weigh more.

3. Immutable positions are permanent jars

For stake position (i), define:

[ P_i=(u_i,s_i,v_i,W_i,r_i,t_i) ]

where:

• (u_i) is the claimant;
• (s_i) is the stable subject identifier;
• (v_i) is the signed semantic commitment;
• (W_i) is the immutable original VIBE amount;
• (r_i) is the immutable queue rank; and
• (t_i) is the creation time or block.

Claim ID, stake ID, signature, content commitments, chain provenance, and relation metadata are also immutable source facts.

The jar persists forever in historical state. Its original size and queue rank never change. A derived live amount, eligibility flag, reputation contribution, moderation tombstone, or current display view must not overwrite those source facts.

4. Live stake decays inside the jar

Let (L_i(t)) be the current live or effective amount of position (i):

[ 0\le L_i(t)\le W_i ]

Between reaffirmations, live stake decays smoothly with block or time progression. The current working family is exponential decay:

[ \frac{dL_i}{dt}=-\lambda_sL_i ]

or:

[ L_i(t)=L_i(t_0)e^{-\lambda_s(t-t_0)} ]

The family is a working mathematical function. (\lambda_s), the implied half-life, time unit, fixed-point approximation, and rounding behavior are simulator parameters or open protocol decisions. They are not frozen by this paper.

Decay reduces current influence and dividend productivity. It does not erase the historical jar, its original commitment, its liability, or its queue rank.

5. Reaffirmation restores but never enlarges

Owner-settled invariant. A participant may spend fresh scarce VIBE to reaffirm the same historical judgment. The top-up is a new append-only event referencing the original position. It is not a mutation of the source record and not a second independent vote.

The working transition is:

[ L_i(t^+)=\min(W_i,L_i(t^-)+q) ]

Production validation should reject overfill before accepting it:

[ q\le W_i-L_i(t^-) ]

A reaffirmation:

• costs fresh VIBE;
• preserves the original claim, semantic commitment, ID, amount, and rank;
• never makes (L_i>W_i);
• creates no additional queue position;
• creates no independent semantic confirmation; and
• cannot generate a self-dividend.

6. Independent semantic confirmation

Implementation refinement, 2026-09-14: the participant-confirmed aspect candidate v4 replaces cosine as economic agreement authority with source-bound assertions and anchored comparison reports. Topical similarity, assertion support/contradiction and positive/negative evaluation are separate. Two complaints about slow service can confirm each other; opposite recommendations do not invert an agreed description. This is a local implementation candidate, not a public contract upgrade or qualified language model. Numerical reward curves below remain unchanged.

Each approved multi-aspect review retains ONE subject-level jar and rank. For funded claim weights w, positive cost is w, negative cost is 2w, and total cost must fit the explicit budget without cancellation. Hierarchical summaries do not create rewards. New same-subject reviews still forfeit earlier upside and preserve downside, including across aspects/visits.

For the preserved earlier vector experiments, the working measure was:

[ a_{ij}=v_i\cdot v_j=\cos(v_i,v_j) ]

Positive alignment supports, near-zero alignment has little or no effect, and negative alignment contradicts. A configurable noise floor may suppress incidental similarity.

The semantic model, extractor, vector dimensions, noise floor, quantization, and version are not permanent protocol law until separately frozen. Any production projector must bind them as immutable, reproducible inputs.

For aggregated signed independent confirmation mass (x_i), the current working bounded family is:

[ F(x_i)=\frac{x_i}{K+|x_i|},\qquad K>0 ]

It is monotonic, symmetric, bounded in ((-1,1)), and has diminishing marginal effect. (K) is roughly the confirmation mass required for half-scale effect. The function family is working; (K), the aggregation rule that produces (x_i), numeric encoding, and any later asymmetry are simulator parameters or open decisions.

7. Self-confirmation is zero

Owner-settled invariant. A person cannot pay themselves by duplicating, restaking, or topping up their own position. For positions (i) and (j):

[ u_i=u_j\quad\Longrightarrow\quad \operatorname{PositiveConfirmation}(j\rightarrow i)=0 ]

Reaffirmation maintains the current influence of one historical judgment. It does not add a new independent source. Repetitive same-owner claims cannot manufacture independence merely by using new IDs, smaller stakes, new wording, or alternating top-ups.

Identity-splitting resistance beyond the same canonical claimant remains a personhood and Sybil model problem that must be tested separately.

8. A new same-subject judgment forfeits old upside

Let (E_i^+(t)\in{0,1}) represent eligibility for future positive dividends.

Owner-settled invariant:

[ E_i^+(t_i)=1 ]

If claimant (u_i) later creates any new judgment about the same subject (s_i), then every earlier position by that claimant on that subject permanently transitions to:

[ E_i^+\rightarrow0 ]

The transition is irreversible. It does not require fuzzy contradiction detection or a judgment that the semantic dimension changed. The new judgment receives its own claim ID, stake ID, amount, and later queue rank.

A top-up of the same position is not a new judgment and therefore does not trigger forfeiture.

9. Lost upside does not erase downside

The forfeiture affects only future positive-dividend eligibility. The earlier position remains historical sensemaking evidence and remains exposed to later independent contradiction.

A new same-subject claim:

• does not erase the old claim, amount, rank, provenance, or live-state history;
• does not claw back benefits already realized;
• does not restore old priority if the claimant changes again later; and
• does not prevent future negative reputation consequences.

The intended asymmetry is:

[ \text{new same-subject claim} \rightarrow\text{surrender old future upside, retain old historical liability} ]

10. Dividends are issuance, not transfers

Owner-settled invariant. A later independent stake is semantic and economic validation evidence. It is not a conserved dividend pool transferred to earlier stakers.

Validation may improve an earlier aligned staker's derived reputation. Reputation then influences a bounded future replenishment policy:

[ A_u(t)=G(R_u(t)) ]

where (A_u) is future VIBE issuance, replenishment rate, or capacity policy. The exact form, coefficients, floor, ceiling, effective epoch, and state transition of (G) remain open.

The protocol does not require:

[ \text{future issuance benefit}\le\text{later validating stake} ]

An excellent early judgment may eventually earn its author much more future VIBE than the raw amount of confirming stake. This inflation of expressive capacity is intentional, bounded by the approved reputation and replenishment curves and by VIBE perishability.

No document or implementation may describe this dividend as a direct token transfer from later stakers unless it is explicitly labeling a superseded historical experiment.

11. Queue priority remains economically meaningful

Owner clarification, 2026-09-08: a later independent person's costly backing can be valuable even when it repeats an earlier judgment. Discovery, attention, consent and adoption are legitimate contributions; new information is not required for every reward opportunity. Earlier supporters retain greater potential, conditional on live stake and later validation. A zero-reward rule for all repeated meaning is not this economic objective. See the ranked-backing reassessment for experimental curves, failure boundaries and unselected repair candidates. No exact curve is frozen by this clarification.

Owner-settled invariant. Early queue rank is more than metadata. An early position that remains eligible and is periodically reaffirmed can retain an economic advantage when later independent semantic flow validates it.

Top-ups do not create ranks. A later new judgment receives a later rank and permanently forfeits the same claimant's earlier positive eligibility on that subject.

For the fresh Base replay generation, rank is one-based, unique, contiguous, and scoped to the canonical subject across charges. Historical V3 zero-based side-queue positions require an explicit generation adapter and are not silently canonical v1 ranks.

The exact queue-advantage function, slope, floor, cap, and interaction with (F) remain simulator parameters. Candidate functions must be named in experiment manifests and compared for early sensemaking, consensus chasing, late copying, split-stake farming, and concentration risk.

12. Wallet perishability and live-stake decay are distinct

Knomee has two separate decay systems:

1. Unspent wallet or capacity perishability. This creates use-it-or-lose-it engagement while smooth replenishment approaches a carrying cap or asymptote.
2. Live stake-position decay. This makes an old judgment require costly reaffirmation to retain current influence and dividend productivity.

They may have substantially different time scales and functions. Neither exact half-life is frozen. Daily Knomee should not be implemented as an artificial midnight cliff unless separately approved.

Claim transaction mechanics, carrying cap, gas sponsorship, automated agent execution, and the exact realization of replenishment remain open.

13. Deterministic oracle projection

Oracles do not resolve claims. They replay complete canonical history through an exact cutoff block under one frozen projector and canonical serialization. Honest oracles must converge on the same byte-identical state.

The replay input must include new-judgment events, immutable stake records, top-up events, semantic artifacts, cutoff binding, and every parameter manifest required to compute live amount, eligibility, alignment, reputation, and future capacity. Derived state includes at least:

• immutable source position references;
• live amount as of the cutoff;
• permanent positive-dividend eligibility;
• signed independent confirmation mass;
• bounded confirmation output;
• derived reputation state; and
• future replenishment state and its effective epoch.

Missing history, unavailable semantic bytes, parameter or projector mismatch, invalid cutoff, duplicate source events, nondeterministic numeric behavior, or divergent roots must fail closed.

14. Economic objective and adversarial acceptance

The design optimizes for useful independent judgment and an economically meaningful ranking of sensemakers. It does not optimize for conserving a fixed pool of VIBE.

Calibration must test at least:

• early correct sensemaking against late consensus chasing;
• independent confirmation against self-confirmation;
• repeated-review and split-stake farming;
• same-subject flip-flopping and irreversible upside forfeiture;
• retained downside after correction;
• top-up overfill and queue duplication;
• identity splitting and coordinated copying;
• long-run issuance, capacity concentration, and perishability; and
• honest participation versus abstention at matched effort and timing.

A passing invariant test proves structure, not economic safety. A parameter candidate must also survive sensitivity analysis, adversarial comparison, deterministic replay, and explicit owner approval.

15. Open calibration and implementation register

The following remain explicitly open:

• wallet decay form and half-life;
• wallet cap, carrying asymptote, and recharge rate;
• live stake decay coefficient or half-life;
• queue-advantage function and parameters;
• semantic noise floor and production semantic model or extractor;
• confirmation scale (K) and signed-mass aggregation;
• reputation aggregation coefficients and time behavior;
• replenishment function (G), floor, ceiling, and effective epoch;
• fixed-point or integer numeric rules and canonical rounding;
• gas, sponsorship, batching, and agent-execution mechanics; and
• exact Base event, state-root, checkpoint, and governance implementation.

The authoritative live list is IMMUTABLE_COMMITMENT_OPEN_ECONOMICS.md. No simulator default or historical contract parameter closes an item in that register.

16. Version and migration history

Migration preserves every accepted historical claim, stake, amount, rank, signature, and provenance record. Older active projections must be rebuilt as versioned derived state. A top-up becomes a new event referencing an original position. A later same-subject new judgment sets earlier positive eligibility to zero from that point forward without erasing prior benefits or future downside.

No Base contract is deployed, upgraded, or broadcast by publication of this paper.