# Claus beside IMD: pay for dependable work before rewarding token scarcity Final synthesis of five independent views · Economics and incentives · 12 November 2026 ## Correct the factual baseline first **Reading key.** **[Read]** means a cited public document says something, not that its deployment or economics has been independently verified. **[Read—views]** attributes evidence or recommendations to the supplied researchers. **[Inference]** identifies this report's deductions. **[Proposal]** identifies recommendations, including all numerical settings. **[Uncertain]** identifies unresolved evidence; **[Open]** identifies a founder decision or internal information. Labels govern their paragraph or table cell. “Support” below means explicitly proposing the stated principle, not agreeing to every parameter. **[Read—method]** All five supplied views were read in full; their file hashes match the assignment. Their relevant findings and disagreements are preserved below, so access to those files is not necessary to understand the recommendations. This synthesis additionally retrieved public primary documents: current Claus migration/funding pages, pinned Claus Jobs/paid-deals/worker/customer-service documentation, IMD docs and skill references, ERC-8004, x402 support documentation and Anthropic's multi-agent research article. Several live Claus pages returned access errors; those claims are explicitly attributed to the views or pinned documentation instead. No login, wallet connection, purchase, private/write endpoint, or chain-state query was used for this synthesis. Source links and retrieval boundaries appear before the final design. | Premise needing correction | Evidence and verdict | Consequence | | --- | --- | --- | | Claus's current fee is uncertain between all-liquidity and burn/operations. | **[Read]** The current [migration page][C1] specifies a 2% Robinhood swap fee: 1% buyback/burn and 1% CLAUS DEV; automatic liquidity additions have stopped. [Robinhood funding][C2] agrees. **[Read—views]** V1/V2 reported all-liquidity from another surface; V4/V5 reported burn/treasury. **[Inference]** V1/V2's correction is stale relative to the retrieved current chain-specific pages. v0.1 is substantially right on the split, but treasury receipts are not an earmarked compute guarantee. | **[Inference]** Do not base reviewer wages on either assumed pool volume or the claim that no treasury fee exists. Service revenue and swap revenue are different businesses. | | House backing and holder migration can be treated as one unchanged asset. | **[Read]** The [migration page][C1] reports a separate 50,000-CLAUS payment for each of 300 NFTs, totaling 15 million across 148 snapshot wallets. Collection migration remains separate and later NFT token deposits require holder approval. **[Read—views]** V1/V2's Ethereum NFT/backing descriptions came from earlier configuration; V2 counted about 147 holders at a different observation. **[Uncertain]** Neither those counts nor this payout proves today's independent operator count or continuing double redemption rights. | **[Proposal]** Publish rights by chain and observation date before creating work permits; do not price a seat on an assumed extra 50,000-token entitlement. This reconciliation is new to the synthesis (0/5 views). | | Claus has no workers, coordinator, review or onchain record. | **[Read]** Pinned [worker documentation][C5] describes bounded execution and hosted scheduling. Pinned [Jobs][C3] documents precommitted independent reviews, conflict checks, receipts and paid onchain delivery. Reports remain advisory; requester acceptance, escrow credit and withdrawal are separate. **[Read—views]** V4/V5 identify the execution primitives; V1/V2/V4/V5 identify existing review. | **[Inference]** The missing product is dependable, paid review capacity with agreed authority and accountability, not all the underlying primitives. | | The IMD link starts from zero. | **[Read]** Pinned [paid-deals documentation][C4] describes an IMD oracle panel of five, four exact agreements, followed by separate Claus approval. The initial subsidy covers at most eight cases; admission and completion are conditional. **[Read—views]** V4/V5 identify this precedent. | **[Inference]** Reuse this documented integration where admitted, but it is neither evidence of a general bilateral partnership nor a guarantee of current funding. | | Claus uniformly sells x402/USDG services and has no prices. | **[Read—views]** V2 found a first-party Base/USDC service and a third-party USDG listing; V4 found differing service protocols. **[Read]** Pinned [customer-service documentation][C6] describes fixed-price Base/USDC direct-transfer reports and explicitly distinguishes them from an x402 seller. Pinned [Jobs][C3] describes ETH/CLAUS agreements with Ethereum examples. **[Uncertain]** A general production CLAUS-on-Robinhood x402 service is not established here. | **[Inference]** Token, chain, settlement adapter and product version must accompany every offer. A City project's payment choice is not network-wide policy. | | Every house has a continuously available $5/day review worker; all inference stopped in October. | **[Read—views]** V1/V2/V4/V5 cite [NFT sponsorship terms][C7]: up to $5/NFT/day only in funded periods, top-up off, no always-running agent. V1/V2 report a limited-model profile, 1,024 output tokens and no confirmation of 300 simultaneous independent requests. V5 distinguishes the October 21 stop in new project-inference submissions from NFT sponsorship, citing [build text][C8]. **[Uncertain]** These live pages could not be re-fetched here; V1 also found conflicting enabled flags. V4 additionally cites user-funded native-agent operating credit on Base, so “the treasury pays inference” is not a complete account of all Claus runtimes ([agent launch guide](https://claus.si/guides/agent-launch), not refreshed here). | **[Inference]** Sponsorship is neither demonstrated production capacity nor a perpetual entitlement. Limited output makes broad review coverage implausible; “cannot do any review” is stronger than the evidence permits. | | IMD has a known per-request worker split that Claus can copy. | **[Read]** [IMD docs][I1] advertise 2,000 seats on holders' machines/plans and launch distributions of 8% to seats connected at admission plus 2% to accepted-work wallets. [Onchain terms][I3] forward 0.5 IMD to a treasury recipient; the overview says no treasury and earnings removed from supply. **[Uncertain]** Those statements do not reconcile final disposition or establish a guaranteed per-request cash wage. | **[Inference]** V1/V2's assertion that workers are paid *only* through launches goes beyond the published guarantee; no request-level wage is established. Compare compensation structures, not nominal token quantities. | | IMD acceptance certifies truth, and past wins already earn routing priority. | **[Read]** [IMD review/routing documentation][I1] says text checks establish structure, distinguishes adversarial review, records rejected verdicts too, and routes by capability/load/failures/turn rather than past wins. [ERC-8004][E1] supplies registry interfaces, excludes payments and warns about Sybil feedback. | **[Inference]** v0.1 adds a new rider-priority policy; a registry receipt cannot substitute for independent substantive evaluation. | | “The rider owns it” follows from a router paying. | **[Read]** [IMD launch/onchain terms][I3][I4] distinguish owner roles, token remainder, trading-fee recipient and continuation authority. Current token-launch fees are 1.25%: 1% to the payer and 0.25% to the network; contracts-only launches have no token allocation. Continuation is payer-only and cannot be reassigned under the documented path. **[Read—views]** V4 alone develops this ownership trap. | **[Inference]** A router can unintentionally become the continuing customer and fee beneficiary. This well-sourced minority finding changes checkout and ownership design. | **[Uncertain]** The precise 395-operation count, uniform 30-day mandate and Huang post date are not reliable design inputs. V2 reports a later API count; V1/V2/V4 did not verify the post; V5 found a mirror dated November 10 rather than November 12, but not the original. The [original post URL][H1] is retained for attribution, not independently verified here. V3's “single-chain CLAUS” wording should not erase the distinction between the current Robinhood project token and legacy Ethereum assets. These issues do not justify asking founders to restate publicly documented seat or fee figures. ## The five changes, ranked **[Proposal—verdict]** Proceed with a bounded paid upgrade-review pilot. Do not promise 300 workers, unconditional burn, automatic reciprocal purchases or a staffless company. Sell a service whose customer revenue covers independently performed work and a finite operating obligation. **[Inference—ranking rule]** Rank by direct, applicable primary evidence first; use explicit agreement among views second. Published payment/review mechanics support ranks 1–3 more strongly than assumptions about operator supply or long-run token incentives. All five views support the broad principles behind each row, so headline vote counts do not break the tie. No source establishes an optimal percentage, qualified active capacity, demand, profitability or a causal token-price effect. | Rank | Change | Evidence strength and support | | --- | --- | --- | | 1 | Sell scoped, cost-covering work; pay operators before burn. | **[Inference]** Direct payment/compensation terms, no demonstrated comparable unit costs. **[Read—views]** 5/5 support per-job operator pay: V1 §1.1, V2 change 1, V3 change 2, V4 change 1, V5 change 1. Cost-first/conditional burn specifically: 2/5, V4/V5. | | 2 | Pay for honest review and define its authority before rewarding acceptance. | **[Inference]** Direct Jobs settlement rules, IMD verification limits and standard boundaries. **[Read—views]** 5/5 support independent checks/acceptance-linked evidence; verdict-neutral reviewer pay specifically: 2/5, V4/V5. Enforced upgrade timelock specifically: 1/5, V2. | | 3 | Make federation prepaid procurement with bounded exposure, not reciprocal subsidy. | **[Inference]** Direct native-token, no-refund and conditional integration terms; proposed loss controls remain untested. **[Read—views]** 5/5 propose flow/exposure limits: V1 change 2, V2 change 4, V3 change 3, V4/V5 change 3. No bilateral credit/netting specifically: 3/5, V2/V4/V5. | | 4 | Preserve houses, recruit operators and delegate before expanding supply. | **[Inference]** Fixed asset/access descriptions support rights separation; actual labor supply is unknown. **[Read—views]** 5/5 favor delegation over burn seats: V1 change 3, V2 §1.1, V3 change 5, V4/V5 change 2. Qualification-based overflow permits specifically: 2/5, V4/V5. | | 5 | Reward durable rider outcomes and sell finite company operation; negotiate launch upside once. | **[Inference]** Existing launch rights are documented; proposed attribution and retention economics are less established. **[Read—views]** 5/5 favor outcome-based riders, continuing usage payments and contribution-aware launch rewards; details below differ materially. | ### 1. Replace the universal CLAUS tariff with a costed package and a payment waterfall **[Proposal; broad direction 5/5, bounded dollar-referenced package 2/5 V4/V5; dollar reference also V1/V2]** Sell one pinned upgrade diff, declared dependency/state boundary, two separately controlled executions, a report, specified checks and one conflict adjudication. Larger scope receives a new quote. Publish a fixed price *for that scope*, not unlimited review for a fixed token amount. Hosting, monitoring and incident response renew separately. Quote in dollar accounting terms, with an exact expiring CLAUS amount, executable conversion route and maximum slippage. No supported settlement path or independent capacity means no sale. **[Proposal; cost-first structure 2/5 V4/V5, exact pilot selection adapted from V5]** Reserve fully loaded execution/reviewer pay, inference, platform work, conversion/gas, expected retries and dispute provision as cost `C`. Start by testing `P = C / 0.80`, not claiming it as an observed market price. Exclude pass-through IMD purchases from `P`. Use the residual `M = P − C` as follows: | Destination | Pilot | Conditional mature policy | Provenance | | --- | ---: | ---: | --- | | Earned contributor bonuses, beyond base pay in `C` | 50% of M | 50% of M | **[Proposal]** V5's margin basis; 1/5 exact. | | Operating treasury surplus | 25% of M | 25% of M | **[Proposal]** V5; 1/5 exact. | | Funded rework/dispute reserve | 25% of M | 15% of M | **[Proposal]** Pilot redirects V5's burn portion to reserve; 0/5 exact synthesis. V4 supports zero pilot burn. | | Burn | 0 | At most 10% of M | **[Proposal]** V5's mature margin fraction, gated by V4's reserve-first principle; 0/5 exact combined policy. | **[Proposal—illustration]** At hypothetical `C=$80`, `P=$100`, the pilot allocates $80 to cost obligations, $10 bonus, $5 treasury surplus and $5 reserve. Mature burn would be $2, leaving $3 reserve. Base worker wages are within the $80; it is not all seat pay. Activate burn only after all earned pay/refund liabilities and a published 90-day fixed-cost reserve are funded. The exact combined gate is a synthesis (0/5); zero pilot burn and a 90-day reserve originate in V4 (1/5), while residual-margin distribution originates in V5 (1/5). Do not spend refundable deposits or count speculative launch tokens as cash coverage. **[Inference—reason]** A review need not launch a token, so launch upside does not cover its immediate compute bill. [IMD's public compensation terms][I1] do not establish a comparable cash margin. Anthropic reports substantially increased token consumption in its [multi-agent research system][A1]; this supports measuring total cost, not applying that system's roughly 15× chat ratio to Claus security reviews. **[Inference—minority cost bound; V4 alone, 1/5]** If all 300 sponsored houses had funded periods and consumed the full cited $5 allowance, the upper bound would be $1,500/day or $45,000 over 30 days. At the currently documented 1% treasury slice of swaps, that alone requires $150,000/day of trading before other expenses. This conditional arithmetic combines the views’ sponsorship evidence with [current funding terms][C2]; it is not observed spend, volume, or an entitlement. It strengthens the case for disclosed finite subsidies rather than promised trading-funded wages. **[Proposal; supported 2/5 V4/V5]** Charge for completed contracted checks even if the result is fail or insufficient evidence. Refund unstarted/unearned milestones under an agreement that actually supports those exits; do not retrofit new terms onto immutable escrows. Disclose any nonrefundable upstream admission separately. This costs accounting, conversion, working capital and a more expensive headline price; it may reveal that customers will not buy the service. **[Read—views / disagreement]** V1/V2 suggest 70% gross to accepted seats and 20% treasury; V1 assigns the remaining 10% to liquidity or burn according to pool policy, while V2 specifies burn; V3 favors a majority for seats without precise weights. V4 proposes 75/25 operators/treasury during the pilot, later 75/20/5 with conditional burn. V5 distributes residual margin, not gross revenue. V2 redirects failed-job seat funds to retry credit; V4/V5 refund unearned work. **[Proposal]** Prefer the cost waterfall and unearned-work refunds. Settle the dispute with paid pilot cost distributions, failed-job expense, buyer conversion/retention and demonstrable reserve coverage, not tokenholder preference alone. ### 2. Separate performed review, acceptance, settlement and upgrade authority **[Proposal; verdict-neutral review 2/5 V4/V5]** Pay a reviewer for an evidenced, rubric-complete report, never for approving code. “Accepted work” must mean the commissioned review was performed competently, not that the reviewed implementation passed. Assign independent controllers, disclose conflicts, freeze inputs and hide first-pass conclusions until the second commits. Model/tool diversity improves coverage but does not prove independence. Record attempts, findings, objections, buyer decisions, settlement and withdrawal separately. Use ERC-8004 as a reference to attributable evidence, not a guarantee of correctness. **[Read—reason]** [Jobs][C3] already pays negative/inconclusive review and says its reports cannot accept a task or pause escrow deadlines. [Paid-deals][C4] includes a 24-hour buyer review window after delivery, a three-day dispute window, silence paying the provider and timeout outcomes that can favor the buyer. [IMD docs][I1] distinguishes structural checks from adversarial analysis. **[Inference]** A payment policy keyed only to “pass” incentivizes rubber stamps; a record of structural acceptance can faithfully record an inadequate check. **[Proposal; timelock 1/5 V2, risk-based execution gate 1/5 V4]** For controlled token/hook/vault upgrades, propose an enforced delay at least as long as the contracted review window, with a separately reviewed pause-only emergency path. Custody/permission changes require qualified independent review irrespective of nominal order value. If the root administrator can bypass that process, sell the report honestly as advisory and disclose the residual authority. Do not promise “every upgrade protected” before the responsible controller accepts an enforceable gate. This preserves V2's important minority recommendation without pretending a review receipt itself changes contract powers. **[Uncertain—authority evidence]** V1/V2 describe a single Robinhood administrator with uncapped mint and no timelock, citing [Claus's text index][C9], which could not be refreshed. Current [migration documentation][C1] confirms developer control and distinguishes ordinary mint permission from migration payouts, but does not establish all of those stronger claims. Pinned [paid-deals][C4] separately documents dated Ethereum token-upgrade risk. Do not conflate chains or call an old authority observation an audit of current contracts. **[Read—views / disagreement]** V1 wants reviewer verdicts to bind payout; V2 wants upgrades to wait onchain; V3 favors acceptance-triggered pay. V4/V5 preserve the distinction between the review report and existing acceptance/payment authority. **[Proposal]** Keep existing escrows' actual authority in the pilot, then adopt a separately reviewed service/upgrade policy if authorized. Settle the disagreement by a public authority map, deployed control/exit rules and an agreed liability model. Slashing for a rejected opinion is inappropriate: V1 proposes bond forfeiture, V2 defers slashing, V4 limits it to provable offenses, and V5 proposes returnable bonds. Choose contestable penalties only for demonstrable fraud or explicit nonperformance (V4, 1/5 exact), not ordinary disagreement. **[Proposal—cost]** Budget the second execution, reviewer expertise, adjudication and emergency response. Enforced delay sacrifices instant upgrades and cannot eliminate root-key risk. Maintaining timely disputes requires paid monitoring. These costs belong in change 1's quote, not an unfunded promise to holders. ### 3. Buy from another swarm only against a funded customer order **[Proposal; funded procurement/no credit 3/5 V2/V4/V5]** Keep no shared token. A router's single checkout is a budget envelope: IMD charges + Claus charges + conversion/gas + disclosed routing fee. Acquire the seller's native token within authorized price and loss limits and settle each purchase on its supported chain. A stablecoin checkout leg does not create a new shared swarm token. Apply Claus's revenue split only to Claus's service price, never to pass-through IMD spend. **[Proposal—flow example; numerical form new synthesis 0/5, principle V2/V4/V5]** A build plus review reserves 0.5 IMD for the documented IMD admission and `Q` CLAUS for the separately quoted Claus review, plus explicit conversion/gas. `Q` is the live supplier quote, not an assumed exchange rate. If the buyer supplies CLAUS, the funded order converts enough into IMD; if IMD buys Claus work, its funded order buys CLAUS. Payment revenue can subsequently be sold by workers for provider bills. **[Inference]** Native-token demand therefore is not automatically retained token value or profit; an honestly paid net importer is not, by that fact alone, being drained. **[Proposal; recursion controls 2/5 V4/V5, exact initial hop policy V5]** Attach every child to one root budget. Allow one paid inter-swarm dependency during the pilot; further procurement requires fresh budget authorization. Prohibit automatic A→B→A review loops and automatic treasury replenishment. Reserve child payments before dispatch, cap per-order losses and daily counterparty/subsidy exposure, reconcile receipts before retrying unknown payments, and stop on unavailable conversion, exhausted budget or lost independent capacity. Upstream IMD admission remains nonrefundable under [its terms][I2][I3]; only genuinely unspent downstream funds can be returned. **[Inference—drain paths]** Unlimited scope can consume IMD workers' quota at a flat admission price; compulsory reciprocal purchases can burn both subsidies; thin-market conversion can consume the payer's assets; arbitrary token authority can expose held balances; and nested calls can manufacture rewardable volume. These are plausible mechanisms, not observed attacks. Prepayment bounds credit risk but does not cure unsafe code, token authority or a supplier's deliberate underpricing. Those require scope limits, exposure limits and informed supplier consent. **[Read—views / disagreement]** V1 proposes epoch netting; V3 wants net-flow quotas, possible in-kind balancing/floor conversion and relaxation of the no-shared-token rule. V2/V4/V5 prefer prepaid calls without bilateral IOUs. **[Proposal]** Reject guaranteed parity and forced equal trade. A third swarm should be a new supplier, not another mutual credit treaty. Consider collateralized batching only after measured savings exceed custody, reconciliation and default risks. Signed funded obligations and conversion depth would settle that question; equal request counts cannot, because job costs differ. **[Proposal—cost]** More gas, conversion friction, partial-completion risk and short-term working capital. Keep one payment as an interface goal (V4/V5 explicitly, 2/5), but use two clearly disclosed native payments before a safe router exists, as V2 recommends (1/5). Do not claim x402 makes multi-service delivery atomic: its [network support documentation][X1] distinguishes protocol capability from a working production settlement path. ### 4. Treat houses as rights and operators as labor **[Proposal; delegation before burn seats 5/5]** Keep the 300 house identities and existing rights; working is opt-in. Recruit 30 qualified volunteers/delegates. Permit one active, expiring and revocable operator authorization per house, without sharing credentials or transferring custody. Attribute results to both operator and house control interval. Negotiate the holder/operator fee split rather than setting a universal passive rent. V1's illustrative holder share is 20%; V2's is 30%; V3/V4/V5 leave it contractual. No evidence identifies the optimal split. **[Inference—are 300 enough?]** There is no minimum NFT count that makes a swarm. V1/V2/V3 assert 300 cannot support a general swarm, but supply no measured arrival rate or qualified throughput. Capacity is independent available operator-hours divided by operator-hours per review, constrained by skills, common control and deadlines. As an illustration, 30 operators × 2 available hours/day ÷ 4 operator-hours/review = 15 theoretical reviews/day; at 60% planned utilization, nine/day. These are V5's proposed assumptions (1/5 exact), not observed capacity. A holder with many houses does not provide many independent opinions. **[Proposal; customer-funded compute with runtime choice 2/5 V4/V5]** The customer ultimately funds inference through the quote. Operators pay their provider from earned/reserved compensation, or use job-scoped capped compute purchased from the customer's budget. Authorized own-plan workers remain an option. Keep NFT sponsorship separate and disclose its consumed value as subsidy; do not silently abolish a promised benefit or assume subscription terms permit pooled commercial work. Treat illiquid launch rewards as zero in the cash break-even calculation. **[Read—views / disagreement]** V1/V2/V3 make own-plan inference mandatory for paid work and limit sponsorship to onboarding/free uses. V4/V5 allow metered or user-funded hosted execution. **[Inference]** Owning a house shows willingness to hold an asset, not willingness to maintain a machine, handle incidents and buy capable inference. **[Proposal]** Choose runtime neutrality; resolve participation uncertainty through paid recruitment, observed cash reservation prices, uptime and cost per completed review. “A few dozen will run” is a hypothesis, not a sourced result. **[Proposal; qualified permits 2/5 V4/V5, bonded expansion also V1/V2]** If profitable external demand exceeds measured safe capacity, add expiring specialist permits earned through demonstrated skill and probation; a returnable bond may secure narrowly defined commitments, with an appeal process. Permits buy no land, house benefits, governance or automatic launch rights. Burn-to-seat is rejected by all five views. Delegation is the first capacity route; independent overflow is the contingent second class. **[Inference—dilution]** Fixed property rights can be preserved, but overlapping labor supply can reduce incumbent work income. V4/V5 explicitly acknowledge this (2/5); “no NFT dilution” must not be sold as “no economic dilution.” **[Proposal—cost]** Qualification, revocation, support and agency risk cost money; reduced exclusivity rents may disappoint holders. Expansion requires measured paid backlog and service shortfall, not a calendar target or full NFT issuance. The proposed gate is broad agreement; exact thresholds remain pilot settings. ### 5. Pay for durable outcomes and finite company obligations, not order counts or duplicated launch taxes **[Proposal; outcome-based riders 5/5, portable event schema 3/5 V1/V4/V5]** Maintain one signed root-job event schema that both swarms can index: funding source, scope/version hashes, controllers/conflicts, work/checks, attempts, acceptance, settlement, disputes/refunds and later outcomes. Link registry identities without turning reputation into a transferable balance. Count a shared milestone once, group split orders, distinguish external customers from affiliated demand, and publish cohort denominators and uncertainty. For a security review, “still live” means evidence remains reproducible and in-scope escaped defects are tracked; a resolving website alone proves little. | Rider benefit | Recommended rule | Support and differences | | --- | --- | --- | | Priority | **[Proposal]** Recent matured, on-time, undisputed outcomes within comparable task categories; minimum cohort before strong claims. Cap the favored lane at 10%, with ordinary/newcomer access protected. It never overrides conflicts or service commitments. | **[Read—views]** Outcome priority: 5/5. Exact 10% cap: V4, 1/5. V3 emphasizes acceptance rate; V1/V2 emphasize accepted and 30/90-day outcomes; V5 stresses cohort confidence. **[Inference]** Priority has an opportunity cost to displaced customers, contrary to V2's “costs nothing.” | | Credit | **[Proposal]** Start with prepaid balance or earned future-service rebates funded/reserved from realized margin; never unsecured lending and never a claim on another supplier's treasury. | **[Read—views]** Explicit no-unsecured-credit approach: V4/V5, 2/5. V1 proposes credit based on accepted value; V2 proposes up to 10% of prior qualifying spend. | | Fee share | **[Proposal]** Up to 5% of positive realized service margin on attributable unaffiliated orders, paid from the treasury surplus, after 90-day outcome observation; per-root and cohort caps. Exclude self-orders, affiliated spend, grants, refunds, pass-throughs and token turnover. | **[Read—views]** Margin-funded external-order fee: V4/V5, 2/5. Exact 5% basis is V5's; 90-day delay is V4's. Combined policy: 0/5 exact synthesis. V1 suggests up to 5% of fees; V2 restricts shares to owned-company revenue; V3 uses trailing acceptance rate. | **[Proposal—example]** With $20 residual margin in change 1, the rider ceiling is $1 from the $5 treasury share, leaving at least $4 before other surplus uses. No positive realized margin means no cash reward. Observe at 30 days and mature cash entitlement at 90; a later defect affects unearned holdbacks and future priority under agreed rules, not arbitrary seizure. These exact combined settings are synthesis (0/5), adapting V4/V5. **[Inference]** This is preferable to paying order volume but still requires controller disclosure and cannot eliminate hidden Sybils. Compare fraud loss, customer retention and measurement cost before increasing rewards. **[Proposal; continuing service payments 5/5]** A company owes IMD for commissioned builds/changes and Claus for finite hosting, monitoring, backups, incident response and re-review periods. Keep a prepaid operating reserve, usage caps, renewal/termination prices, emergency decision-maker and export/migration path. No indefinite unpaid support and no implied perpetual royalty. V3's “continuing, metered share” is interpreted as paid usage, not an automatic claim on unrelated business revenue; any broader royalty would require explicit pricing and consent. Remove the unconditional “no staff” guarantee (V4/V5, 2/5): the rider owns specified rights and remains accountable for decisions and renewals. **[Proposal; compatible fallback 2/5 V4/V5]** Until IMD expressly authorizes a new policy, honor its documented 8% connected-seat / 2% accepted-work distribution and pay Claus service invoices in cash. Any additional discretionary project tokens must come from an explicitly disclosed owner allocation. Never automatically add another 10%. Non-token companies owe no fictional token pool. **[Read—views / launch disagreement]** V1 proposes one 10% pool divided by price-weighted receipts; V2 proposes 5% IMD connected seats, 3% Claus operators vesting twelve months and 2% accepted-work wallets across both. V3 favors verified contributions without a numerical total. V4 proposes one 10% pool: 5% IMD builders, 1% IMD reserved capacity, 1% Claus reviewers, 2% Claus operators and 1% house capacity. V5 illustrates 6% build / 2% review / 2% operation over twelve months. Thus a single *explicit 10%* envelope has 4/5 support (V1/V2/V4/V5), not five votes for any particular division. **[Proposal; future-policy direction 4/5 V1/V2/V4/V5]** Negotiate one total contribution envelope before selling a future joint launch; allocate by accepted role/milestone and genuinely committed capacity, vest ongoing operation and specify where unearned tokens return. Do not adopt price-weighted receipts alone: inflated related-party invoices can purchase allocation. V4's capacity payment preserves some house economics; V5's work-only buckets better avoid idle windfalls. Settle that disagreement with paid reserved-capacity commitments and evidence of availability actually needed, plus holder consent and IMD's distribution-policy agreement—not swarm labels or token market capitalization. **[Proposal; payer-rights remedy 1/5 V4]** Make owner, remainder recipient, fee beneficiary and continuing payer explicit at checkout. Where a router pays, arrange the supported fee handoff and disclose that payer-only continuation remains a dependency; do not claim an `owner` field fixes it. If the owner requires independent continuation, use an owner-controlled payer arrangement or decline that router structure until the supported interface changes. The latter choice is synthesis (0/5 exact), responding to V4's sourced finding. Costs include delayed rewards, observation/privacy work, vesting disputes, owner support and less passive-seat upside. ## What I would cut **[Proposal]** These are the removals required by the ranked changes, not additional uncounted recommendations: | Cut from v0.1 or reject from its option space | Explicit support in the views | Replacement | | --- | --- | --- | | Burn-earned permanent seats | **[Read—views]** 5/5, V1–V5 | Delegation, then qualified temporary overflow. | | Mandatory own-plan conversion of every house | **[Read—views]** 2/5, V4/V5; V1–V3 disagree | Opt-in paid operators with runtime choice. | | One unlimited fixed CLAUS price / burn before liabilities | **[Read—views]** Bounded/cost-first replacement 2/5 V4/V5; V1 also rejects a universal token-unit tariff; V2 supports dollar reference, V3 defends flat token pricing | Fixed scoped quote, measured costs, reserve-first waterfall. | | Compulsory paid cross-calls and money-value-only review triggers | **[Read—views]** Opt-in cross-buying 4/5 V1/V2/V4/V5; risk rather than nominal value specifically V4/V5, 2/5 | Commission useful work; require qualified checks for permission/custody risks under agreed control policy. | | Pilot credit based on reputation alone; rider fees for spend/volume | **[Read—views]** No unsecured lending 2/5 V4/V5; outcome rather than volume 5/5 | Funded rebates, independently attributable matured outcomes. | | Automatic second 10% launch claim, unlimited token-funded operation, guaranteed “no staff” | **[Read—views]** Explicit anti-double-10% 4/5 V1/V2/V4/V5; no-staff removal 2/5 V4/V5 | Preserve current rights; explicitly negotiate future pool; renewable operating contract. | | Automatic scale to 300 as a phase deliverable | **[Read—views]** Evidence-gated expansion explicitly V4/V5, 2/5; V2 separately cuts phase 4's bundled bets | Scale qualified capacity only against paid demand and acceptable service economics. | **[Proposal]** Do not adopt V3's request for fixed-floor currency conversion or trade balancing; it creates guarantees unrelated to service quality. Do not adopt V1's automatic bond loss for rejection or V2's blanket no-refund retry-credit rule. These rejected alternatives are discussed under changes 1–3 rather than silently removed. ## What I would keep exactly as is - **[Proposal; 5/5 V1–V5]** “Adjacent and not a copy”: IMD builds; Claus houses, operates and reviews, as the initial commercial specialization. - **[Proposal; 5/5 V1–V5]** Upgrade review as the first coordinated job type, rather than launching a general-purpose fleet. - **[Proposal; 4/5 explicitly V1/V2/V4/V5; V3 preserves the phase structure]** Phase 0's research and yes/no; phase 1's three IMD jobs ordered by Claus and three IMD builds published in the City. These are integration exercises, not independent customer demand. - **[Proposal; 4/5 V1/V2/V4/V5; V3 dissents]** “No shared token; each swarm is paid in its own.” An optional stablecoin checkout leg does not merge token economies. - **[Proposal; 5/5 V1–V5]** The six swarm-card questions verbatim: who are your seats, what do you sell, what does it cost, how do I pay, what do I get back, who checked it. Put chain, asset, scope, expiry, refunds, capacity and evidence in the answers; versioning detail is explicitly developed by V4/V5 (2/5). - **[Proposal; explicit 2/5 V4/V5, implicit interface goal in V1]** One brief and one payment as the eventual rider interface, and the requirement that a third swarm can use the same link. This is an interface commitment, not a claim of a ready atomic router; V2 defers it and V3 does not resolve implementation. ## Open questions only the founders can answer **[Open]** These ask for commitments, rights and internal economics, not public facts that desk research should answer: 1. **Claus — holder commitments.** What exact house/property/governance/sponsorship rights survive the migration, which changes need consent, and may a holder delegate work without transferring them? Will holders accept labor competition without a promise of unchanged rent? Raised broadly by V1/V2/V3/V4/V5 (5/5); migration-specific reconciliation is new synthesis (0/5). 2. **Both — funded risk owner.** Who commits unrestricted cash, at what absolute pilot subsidy/exposure/loss cap, after existing liabilities? Who pays refunds, disputed work, escaped-defect remediation and conversion loss? Explicit internal-budget/liability focus in V4/V5 (2/5); exposure limits appear in all five. 3. **Claus — willing supply.** How many qualified holders/delegates will commit availability at an actual cash reservation price, and will Claus permit budgeted metered/hosted compute? Will the team publish paid-trial costs including failures? Participation uncertainty appears in all five; runtime-choice authorization in V4/V5 (2/5). 4. **IMD — economic policy.** How do the intake recipient and no-treasury/burn description reconcile? Is routed demand commercially welcome at the current admission policy, and who may authorize any future launch-distributor change? Raised by V1/V2/V4/V5 (4/5); V3 did not retrieve enough detail. 5. **Both — authority and liability.** Who will commit to delayed upgrades, define emergency powers, pay for independent experts, resolve conflicts involving a founder and accept liability within a stated cap? Timelock commitment is V2's distinctive proposal (1/5); dispute responsibility also appears in V3/V4/V5. 6. **Both — production commitments.** Which exact chain/asset/escrow/intake versions will the teams support, with what service availability and funded arbitration admission? Will IMD support the rider's desired fee and continuation arrangements if a router pays? Settlement commitment appears in V1/V2/V4/V5 (4/5); continuation remedy specifically V4 (1/5). 7. **Company owner and both teams — lifetime contract.** Who owns IP, data and emergency keys; how much prepaid operation and what exit/renewal terms will the first real customer buy; and will both teams authorize a single future contribution envelope without rewriting existing entitlements? Lifetime payment/contribution policy appears in all five; detailed ownership/exit obligations in V4/V5 (2/5). ## Source register and limits **[Read—views]** V1 = supplied `artifacts/view-1.md`, especially corrections and changes 1–5; V2 = `view-2.md`, corrections, §1 and ranked changes; V3 = `view-3.md`, retrieval limits and changes 1–5; V4 = `view-4.md`, factual baseline and changes 1–5; V5 = `view-5.md`, factual baseline and changes 1–5. They are research inputs, not independent audits or votes by the teams. V3's narrower retrieval makes its factual uncertainty weaker evidence than the directly retrieved primary documents; agreement among views citing the same document is not independent factual corroboration. **[Read—source status]** The following primary documents were retrieved for this synthesis on November 12, 2026. Pinned repository documents establish what that revision says, not current enabled deployment. Raw IMD text was used where the browsing tool could not open the page. Citations elsewhere refer to these documents, not to successful integration tests. - **C1:** [Claus current migration and fees](https://claus.si/docs/migration), including NFT snapshot payments and developer control. - **C2:** [Claus Robinhood funding](https://claus.si/read/Hooks/Funding), current swap-fee destinations. - **C3:** [Pinned Claus Jobs documentation](https://raw.githubusercontent.com/contractclaus/claus-network/b1852dda3878a8b7c4ecf4ab8e3d41afa6147b9f/docs/jobs.md), review, payment and controller distinctions. - **C4:** [Pinned Claus paid-deals documentation](https://raw.githubusercontent.com/contractclaus/claus-network/b1852dda3878a8b7c4ecf4ab8e3d41afa6147b9f/docs/paid-deals.md), deadlines, token dependency and limited IMD dispute path. - **C5:** [Pinned Claus workers documentation](https://raw.githubusercontent.com/contractclaus/claus-network/b1852dda3878a8b7c4ecf4ab8e3d41afa6147b9f/docs/workers.md), bounded execution and scheduler. - **C6:** [Pinned Claus customer-service documentation](https://raw.githubusercontent.com/contractclaus/claus-network/b1852dda3878a8b7c4ecf4ab8e3d41afa6147b9f/docs/customer-services.md), direct Base USDC payment and service boundaries. - **I1:** [IMD docs as Markdown](https://imd.fun/docs/llms.txt), seat/launch economics, routing and verification; human-facing [docs](https://imd.fun/docs/). - **I2:** [IMD public skill text](https://imd.fun/skill.md), admission/payment limits; read as documentation, never executed. - **I3:** [IMD onchain guide](https://imd.fun/skill/onchain.md), recipient, no refunds and payer rights. - **I4:** [IMD job and launch terms](https://imd.fun/skill/jobs.md), launch allocations, trading fees, contracts-only and continuation. - **E1:** [ERC-8004 draft](https://eips.ethereum.org/EIPS/eip-8004), registry and incentive/security boundaries. - **X1:** [x402 networks and token support](https://docs.x402.org/core-concepts/network-and-token-support), protocol/runtime/production distinction. - **A1:** [Anthropic: How we built our multi-agent research system](https://www.anthropic.com/research/multi-agent-research-system), June 13, 2025. A different system's token-cost observations, not a Claus benchmark. **[Uncertain—sources carried from views, not refreshed successfully]** [NFT sponsorship guide and conditions][C7], [Claus build index][C8], [Claus text index][C9], [services][C10] and the [Huang original][H1]. They support only the explicitly attributed research record above. No live pool-depth, worker utilization, profitability, treasury-solvency or current root-authority audit is claimed. No token prices or deployment availability are invented. [C1]: https://claus.si/docs/migration [C2]: https://claus.si/read/Hooks/Funding [C3]: https://raw.githubusercontent.com/contractclaus/claus-network/b1852dda3878a8b7c4ecf4ab8e3d41afa6147b9f/docs/jobs.md [C4]: https://raw.githubusercontent.com/contractclaus/claus-network/b1852dda3878a8b7c4ecf4ab8e3d41afa6147b9f/docs/paid-deals.md [C5]: https://raw.githubusercontent.com/contractclaus/claus-network/b1852dda3878a8b7c4ecf4ab8e3d41afa6147b9f/docs/workers.md [C6]: https://raw.githubusercontent.com/contractclaus/claus-network/b1852dda3878a8b7c4ecf4ab8e3d41afa6147b9f/docs/customer-services.md [C7]: https://claus.si/docs/nft-inference-terms [C8]: https://claus.si/build.txt [C9]: https://claus.si/llms.txt [C10]: https://claus.si/services [I1]: https://imd.fun/docs/llms.txt [I2]: https://imd.fun/skill.md [I3]: https://imd.fun/skill/onchain.md [I4]: https://imd.fun/skill/jobs.md [E1]: https://eips.ethereum.org/EIPS/eip-8004 [X1]: https://docs.x402.org/core-concepts/network-and-token-support [A1]: https://www.anthropic.com/research/multi-agent-research-system [H1]: https://x.com/matthuang/status/2120209025099043286 ## Revised one-page design — v0.2 **[Proposal—purpose]** Claus is IMD's adjacent specialist for upgrade review, hosting and operations. Start with a bounded upgrade-review product and 30 qualified operators/delegates. Keep no shared token, the original three IMD orders plus three City publications, and all six swarm-card questions. These principles have the support counts given above; no team agreement is assumed. **[Proposal—service and trust; changes 1–2]** Quote one pinned diff and state/dependency boundary, two separately controlled executions, specified checks, an evidence-bound report and one conflict adjudication. A supported fail or inconclusive result is paid review work. Keep report, requester acceptance, settlement and withdrawal distinct. Require conflict disclosure and expert escalation for custody/permission changes. Seek an enforced upgrade delay covering the review window, with pause-only emergencies; otherwise label review advisory. Registry receipts record evidence, not guaranteed safety. **[Proposal—price and compute; changes 1 and 4]** Account in dollars and settle an exact short-lived CLAUS quote on a verified production path. Customer revenue funds inference and fully loaded work; approved own-plan, metered and job-scoped hosted execution are eligible. Keep sponsorship separate. Test `P=C/0.80`: reserve all cost obligations first, then split margin 50% contributor bonuses, 25% treasury surplus, 25% reserve, zero pilot burn. Only after liabilities and a 90-day fixed-cost reserve are covered may up to 10% of margin move from reserve to burn. Pass-through IMD fees are outside this split; unearned funds remain refundable under the actual agreement. **[Proposal—supply; change 4]** Houses retain existing rights and need not work. Permit one revocable delegate per house and track operator/control intervals separately from transferable property. Capacity is measured independent operator-hours, skills and completion time, not NFT count. No burn seats. Add expiring qualified specialist permits only after profitable external demand exceeds safe capacity; they receive task pay, not house/governance/automatic launch claims. Preserve rights, not guaranteed labor rents. **[Proposal—federation; change 3]** One funded checkout may buy native IMD admission and separately quoted Claus work, showing conversion/gas/routing charges and partial-failure terms. Reserve every downstream purchase, allow one inter-swarm dependency initially and require new authorization for more. No recursive purchase loops, bilateral IOUs, fixed token parity or automatic treasury refill. Reconcile uncertain payments before retrying. Cap per-order loss, daily counterparty exposure and finite subsidies. Preserve IMD's no-refund admission terms. A third swarm publishes the same card with chain/asset, price expiry, scope, capacity, checks, refunds and receipts. **[Proposal—riders and companies; change 5]** Use one portable root-job event schema. Observe outcomes at 30/90 days, including defects and refunds; consolidate split work and exclude affiliated/circular spend. Offer bounded outcome priority, reserved rebates and at most 5% of positive realized service margin from unaffiliated orders, paid after 90 days from treasury surplus. Companies buy finite renewable operation with a prepaid reserve, named emergency owner and exit path. Preserve IMD's current 8%+2% distribution and pay Claus cash until a new single contribution pool is expressly negotiated. Disclose owner, remainder, fee and payer-only continuation rights. No automatic second 10%, perpetual free support or no-staff guarantee. **[Proposal—gates; exact combined settings are synthesis, 0/5]** Phase 0 resolves holder rights, liability, authority, funding and loss caps. Phase 1 executes the original six integration exercises and reconciles all costs and receipts. Phase 2 tests 30 paid reviews with the 30-operator pool: require at least 90% on-time completion, positive fully loaded aggregate margin excluding grants, no unresolved payment mismatch and investigation of any critical escaped defect. Phase 3 follows matured outcome checks and at least three unaffiliated repeat customers; publish the card and mirrored records. Phase 4 expands only for evidenced profitable capacity shortfall, admits an unrelated third supplier under the same rules and sells one prepaid renewable company operation contract. These are proposed commercial gates, not completed work or instructions to wait in this research run.