# Claus beside IMD: pay for dependable capacity before buying scarcity Independent view 5 — economics and incentives — 12 November 2026 **[Proposal]** Proceed with a bounded commercial pilot, not the conversion of all houses into workers. Change v0.1 from an NFT expansion and token-burn program into a market for independently checked, costed services. The scarce asset to measure is dependable delivery per dollar, including the rider's contribution to defining it. ## Reading key and research limits **[Read]** means a public source says this; it does not mean I verified a deployment or audited a contract. **[Inference]** is my interpretation or calculation from specified evidence. **[Proposal]** is my recommendation, including all illustrative prices, percentages and thresholds. **[Open]** identifies information or a decision only the founders can supply. Labels apply to the entire paragraph or table cell unless another label appears. The supplied v0.1 is the proposal under review, not evidence of agreement. **[Read / method]** I read the public documents linked below, IMD's `skill.md` as text, the x402 public repository, and the Claus repository PR linked by its overview. I used only unauthenticated public reads; no login, signature, payment, private endpoint or write request. I did not read the other researchers' views or existing IMD reports about this proposal. Some pages differ across public surfaces, and several linked pages were inaccessible. No source here establishes current worker profitability, utilization, order volume, pool depth or treasury runway. This is desk research, not a live integration test. ## Correct the factual baseline first | v0.1 premise | What the sources actually establish | Economic consequence | | --- | --- | --- | | Claus lacks independent acceptance or an onchain record. | **[Read]** Jobs documents optional independent, snapshot-bound advisory reviews with conflict checks and separate reviewer commissions. Paid agreements commit delivery onchain. Requester acceptance remains distinct from financial acceptance and withdrawal. [C1] | **[Inference]** The missing product is coordinated, reliably funded review capacity and an aggregate worker record. Claiming an absence of all review or onchain evidence would fund duplicate work. | | The bridge starts from zero. | **[Read]** Paid-deals documentation describes a limited dispute subsidy and an IMD oracle panel whose proposed split requires separate Claus approval. Admission and completion are conditional. [C2] | **[Inference]** Extend an existing documented dependency; do not announce a new bilateral partnership or assume the path is currently available. | | No working seat or coordinator means no worker functionality. | **[Read]** Claus's build index links bounded project workers, persistent runtimes and recurring workflows; it explicitly separates these from NFT inference. [C3] | **[Inference]** Those are not proof of an IMD-like seat dispatch market. Equally, a categorical claim of no worker or coordination primitives is too strong. | | Every house has $5/day of inference. | **[Read]** There are 300 numbered houses. The sponsored plan permits up to $5/NFT/day, automatic top-up is off, and only funded active periods supply allowance. It does not start an always-running agent or switch to personal billing on exhaustion. [C4] | **[Inference]** This is conditional consumption support, not an operating commitment or proof of 300 willing suppliers. | | Claus inference was switched off in October. | **[Read]** The build index dates the stop in new *project inference submissions*, including existing allowances, to 21 October 2026. NFT sponsorship and Claus's own access are separate. [C3] | **[Inference]** Do not treat either all inference as unavailable or treasury inference as universally available. | | Claus's fee split is settled everywhere. | **[Read]** Current migration and Robinhood funding pages say 2% of swaps: 1% buyback/burn, 1% CLAUS DEV. An older funding surface lists a different multi-part split totaling 2.3%. [C5–C7] | **[Inference]** Use the chain-specific current description, disclose the inconsistency, and keep swap taxes separate from a proposed service price. | | IMD request fees can be copied as worker economics. | **[Read]** IMD docs advertise 0.5 IMD/request, holder-funded machines/plans and launch allocations of 8% to connected seats plus 2% to accepted-work wallets. `skill.md` says payment buys admission, not a guaranteed result, with no refunds. [I1, I3] | **[Inference]** A cheap entry price plus speculative launch rewards is not evidence that cash operating costs are covered. | **[Read]** IMD's API reference describes x402/Permit2 payment and contract intake on Ethereum and Robinhood Chain. It says the intake forwards the entire 0.5 IMD to a treasury, whereas its overview says the network has no treasury and earnings leave supply. [I1–I2] **[Inference]** A technical recipient may differ from an economic treasury, but these pages alone do not reconcile final disposition. Do not call every request payment a verified burn; ask for the reconciliation before using IMD as a split benchmark. **[Read]** ERC-8004 is a draft identity/reputation/validation registry standard; payments and validator incentives are outside its scope, and it explicitly warns about Sybil feedback. [E1] **[Inference]** An ERC number neither creates worker economics nor proves independent review. **[Read, attribution limit]** A public mirror reproduces Huang's “Swarm riders and Swarm-native companies” post and labels it November 10, rather than the brief's November 12; direct X retrieval failed. It emphasizes evaluation design and productive swarm command. Treat its date as unresolved and its productivity claims as the author's observations, not measured demand for this product. [H1] **[Inference]** It supports testing rider skill, not a conclusion that every company can run without accountable people. ## Five changes, ranked ### 1. Replace one fixed token price with a bounded service quote and a cost waterfall **[Proposal]** Publish a standard upgrade-review SKU first: one pinned repository revision and diff, declared contract scope, two separately assigned executions, a written report and one adjudication of conflicting findings. Its price is fixed for the accepted scope; larger changes require another quoted tier. Hosting, incident response and perpetual operation are separate products. A request purchases the specified review process, never a guarantee that the code is safe. **[Proposal]** Quote the service in a stable dollar accounting unit and settle in CLAUS at an executable rate. Publish the exact CLAUS amount, chain, expiry, cost breakdown and maximum conversion slippage before authorization. Quotes last five minutes and are capacity-reserved. If there is no sufficiently liquid conversion route or supported CLAUS x402 scheme, do not sell that route; use an explicitly agreed existing paid agreement for the pilot. Supporting arbitrary token settlement is work, not a feature conferred by the x402 name. x402 documents payment negotiation and settlement, not this proposed service/refund contract. [X1–X2] **[Proposal]** Let `C` be budgeted cash cost for execution and its independent second run, reviewer labor, inference, hosting used by the job and settlement gas. Reserve that amount before promising a job. Set the quoted price `P = C / 0.80`; distribute only the residual `M = P − C`: 50% to accepted execution/review bonuses, 25% to treasury operations, 15% to a capped dispute/rework reserve, and 10% to burn. This is a starting policy for the pilot, not a market-derived optimum. The base worker and reviewer fees are part of `C`, not contingent on discovering a vulnerability or issuing a pass. **[Proposal / arithmetic example]** If the complete cost budget is $8, the quote is $10. The $2 margin yields $1 worker/reviewer bonuses, $0.50 treasury, $0.30 reserve and $0.20 burn. Thus burn is 2% of gross, while supplier compensation includes the appropriate portion of the $8 plus $1. Do not call all $8 a seat payout: it also contains non-seat costs. In CLAUS, apply the split to the realized settlement value; workers may elect immediate conversion of their earned share. Charge conversion costs in `C`. Hold refundable funds outside this distribution until earned. **[Proposal]** Refund unstarted work after the quoted start deadline; charge only disclosed, completed milestones if a job stops. A negative or inconclusive review is a valid paid outcome if the commissioned checks were performed. Rework caused by a provider's missed agreed check comes from its holdback/reserve, with an independent appeal; customer scope changes cost extra. Quote ceilings, finite retry budgets and a stop-on-overrun rule prevent an unbounded promise. Do not retrofit these terms onto immutable existing escrows; use their existing deadlines and exits, or introduce a separately reviewed service agreement. **[Inference / reason]** The existing sources establish very different subsidy and reward arrangements, not a comparable cost per accepted review. [C4, I1, I3] Burning gross revenue while inference is unfunded makes future workers dependent on token appreciation or treasury grants. A cost floor makes the subsidy visible and lets Claus differentiate on operating reliability instead of matching IMD's nominal 0.5-token price. **[Proposal / cost]** This sacrifices a headline-simple price and most immediate burn, requires cost measurement and token conversion, and may reveal customers will not pay enough. Pilot with paid quotes before committing to a permanent split. Subsidies must have a named budget and end condition and must appear separately from customer revenue. ### 2. Keep 300 house rights; buy qualified capacity and permit delegation before minting more seats **[Inference / answer]** Three hundred titles can be enough for a swarm; they are neither a minimum nor a capacity measure. What matters is active independent operators, skill coverage, concurrent work and completion time. No reviewed source supplies those Claus measurements. House count and funded inference access alone establish none of them. [C4] **[Proposal / capacity test]** Define daily review capacity as `active operators × available productive hours ÷ total operator-hours per review`, then reserve 40% for bursts, retries and conflicts. For illustration only, 30 operators offering two hours/day and four total operator-hours per review produce 15 theoretical reviews/day, or nine/day at the 60% utilization ceiling. Separate executions must still have different controllers; that constraint and scarce specialists may lower capacity further. Expand only when paid demand persistently exceeds measured safe capacity at profitable prices, not when the collection sells out. **[Proposal]** Houses retain their existing property/access rights. Working is opt-in. Each house may appoint one active operating delegate at a time under a revocable, time-bounded agreement; the operator receives task payments, and any owner/delegate division is disclosed before dispatch. Separate operator identity and track record from transferable house ownership. A new holder acquires the house and its historical record, not the old operator's personal qualifications. Delegation adds useful operators without minting another house or duplicating its rights. **[Proposal]** If the pilot proves additional capacity is needed, the best second class is nontransferable, expiring specialist operator permits earned by demonstrated work, with a returnable performance bond and probation. Give permits access to paid tasks, not house parcels, house benefits, governance or automatic launch allocations. Do not burn CLAUS to buy qualification: destroying capital proves neither skill nor independence and creates an incentive to keep entry expensive. Delegation is the first step; specialist permits are the contingent second class. **[Inference]** “Without dilution” needs a precise promise. Property rights can stay fixed, but letting more operators compete can reduce incumbents' work income. No honest design can guarantee both open capacity and unchanged rents. **[Proposal]** Protect the 300's existing entitlements, not a perpetual monopoly on customer work; never finance new operator subsidies by silently repurposing house allowances. **[Proposal]** Customers ultimately pay inference through job prices. Operators supply their own approved plan or a metered provider account and price its opportunity cost; the coordinator can offer job-scoped, capped compute instead, purchased from the reserved cost budget. Treasury sponsorship is a separately capped onboarding/public-goods expense. Do not hand workers a general treasury inference key or assume personal subscription terms allow commercial pooled use. **[Inference / participation]** Some holders may operate, but the public allowance proves willingness to consume, not willingness to host, maintain and review. [C4] Compare realized cash fees plus conservatively valued launch rewards against inference, machine, attention and rejection costs. Treat illiquid launch tokens as zero in the cash break-even test. **[Proposal / cost]** Recruit 30 volunteers/delegates with explicit availability and a finite paid trial; qualification, isolation and support cost money, and delegation introduces agency risk. This is preferable to promising 300 workers who have never agreed to work. ### 3. Make inter-swarm procurement prepaid and non-recursive **[Proposal]** Keep two native settlement assets. A single rider checkout is a budget envelope, not a shared-token balance sheet. The router obtains independent supplier quotes, locks the required IMD and CLAUS or converts the customer's funds within signed limits, and shows its routing/conversion fee. The supplier records a native-token sale; the purchasing swarm records an external cost against that specific customer order. Do not net opaque bilateral IOUs or exchange tokens at an invented fixed parity. **[Proposal / value-flow example]** For a company build, buy one IMD admission at its documented 0.5 IMD price, then pay the separately quoted Claus deployment/operation service. [I2] If Claus commissions an IMD follow-up, the original customer's authorized budget must cover it. Conversely, IMD's purchase of Claus review is a new cost requiring budget. Unsold launch tokens and expected swap fees do not count as prepayment. A discounted flat-price upstream job is not permission to resell unlimited downstream compute. **[Inference / drain analysis]** Reciprocal calls can manufacture demand, rider reputation and subsidy claims without external customers. A volatile CLAUS quote can buy expensive IMD services while leaving Claus with too little cash to operate. Credit exposure and unlimited callbacks can drain one treasury even when each call individually pays its native fee. These are mechanism risks, not observed misconduct. **[Proposal]** Tag every child order with the same root customer budget. Limit each root to one paid cross-swarm dependency during the initial pilot; a further hop requires a new explicit budget authorization. Track a visited-swarm list, reserve funds for each child before dispatch, set aggregate daily counterparty exposure and enforce per-order maximum loss. Never automatically call back into a previous swarm to commission another review. Batch unrelated reviews only when scope and cost savings are explicit. Stop new orders on exhausted budget, failed conversion or unavailable independent capacity. **[Proposal]** Where upstream admission is nonrefundable, disclose that irreversible cost before checkout and make the router's refund promise exclude it once purchased. A downstream failure cannot reverse an IMD payment by wishful accounting. Require separate native-chain receipts and reconciliation for payment, admission, delivery and acceptance; one HTTP success is not evidence of the entire sequence. x402's roles do not supply this cross-service business logic. [I3, X2] **[Proposal / reason and cost]** Prefer funded purchases to a bilateral credit line so neither treasury involuntarily bankrolls the other. This adds conversion friction, short-term inventory and working capital, and limits automation. It also makes onboarding a third swarm a matter of another supplier card and risk limit, rather than another bespoke token treaty. Reuse documented Claus/IMD dispute plumbing where its terms fit, with no assumption of unlimited subsidized cases. [C2] ### 4. Reward riders for retained outcomes, and pay reviewers for checking rather than approving **[Proposal]** Use one portable, signed event record keyed by root job ID; both swarms can index it without maintaining two incompatible scores. Record funding source, external versus related-party buyer, brief/version hash, acceptance rubric, deliveries, independent review, disputes, refunds and 30/90-day outcome checks. Publish counts and cohort denominators, not a single “jobs shipped” leaderboard. Link ERC-8004 records where useful; do not pretend its registry determines the score. [E1] **[Proposal]** Priority is earned by a recent, category-specific history of accepted work still meeting its agreed outcome at 30 days, adjusted for defects and rework. A small completed cohort remains low-confidence. Cap priority so new riders still receive at least 20% of pilot capacity; priority never overrides a fully booked queue or a conflict rule. Website reachability, security review completion and actual customer use are different outcomes: choose a meaningful metric before purchase, not a generic “still live” check. **[Proposal]** At first, credit means prepaid balance or an earned discount, not unsecured borrowing. A rider's fee share is at most 5% of positive realized service margin on qualifying external orders they originated and scoped; pay half after acceptance and half after the 30-day check. Fund this from the treasury's 25% margin share in change 1, so the example's $2 margin permits at most $0.10 rider reward and leaves $0.40 treasury. No payout for ordering alone, related-party circular spend, subsidies, refunds, capital raised, token trading volume or unaccepted work. The buyer must be told who receives the fee. **[Proposal]** Cap rider rewards per root project and per cohort; aggregate split jobs into the root before scoring. The rider's own jobs may build a qualified outcome history but do not earn a referral payout to themselves. Require control/conflict disclosures and risk review for suspicious clusters; wallet separation alone is insufficient. Hold rewards until earned rather than relying on recovering already-paid tokens. A later defect changes future priority and any still-unearned holdback; ordinary poor outcomes do not justify arbitrary confiscation. **[Proposal]** Randomly assign eligible, separately controlled reviewers, with model/provider diversity where practical, and conceal the first report until the second run commits. Pay the agreed review fee for a supported pass, fail or insufficient-evidence result. Reward substantiated findings and calibrated later outcomes; do not reward finding count or a favorable verdict. High-value review must price expertise and conflict screening, and cross-swarm membership alone cannot establish independence. **[Inference / reason]** Paying for spend rewards spend; paying for approval rewards collusion. The documented Claus reviewer commission already recognizes that negative/inconclusive findings are paid work. [C1] **[Proposal / cost]** Preserve that principle while adding cohort accounting, delayed payouts and paid rechecks. The costs are slower rewards, privacy tradeoffs, reviewer scarcity and moderation; fraud controls reduce, but cannot eliminate, common-control Sybils. ### 5. Sell a renewable operating contract; allocate launch rewards once across the lifecycle **[Proposal]** A swarm-native company owes each provider only what a signed service contract specifies. Pay IMD for initial build and later changes when commissioned. Pay Claus for a finite operating period covering hosting, monitoring, backups, incident triage and review cadence, with usage caps, response targets and renewal price. Maintain a prepaid operating reserve and explicit pause/export/migration path. Token turnover is uncertain income, so treasury swap fees must not be the sole source of a company's service-level commitments. **[Proposal]** Remove “no staff” as a product guarantee. Name the rider or legal owner responsible for renewal, emergency authority and decisions agents cannot safely make. A worker cannot inherit ownership merely by keeping an app alive. Give operating agents revocable spending mandates with bounded scope; make emergency suspension distinct from custody or ownership transfer. **[Inference]** Two swarms contributing does not justify doubling a founder's supply concession. Equally, splitting IMD's published 8%/2% automatically would alter an existing advertised entitlement. [I1] **[Proposal]** The compatible pilot leaves IMD's published 10% allocation intact and pays Claus in cash. Any additional token grant to Claus must be a separate, explicitly priced owner decision, never an implied second 10%. **[Proposal]** For a future jointly governed launch product only, negotiate one total launch reward envelope before sale and obtain IMD's agreement to change its policy. An illustrative envelope is 10% total supply: 6% for accepted build milestones, 2% for genuinely independent review and 2% for funded operation milestones over twelve months. These buckets pay qualified contributors regardless of swarm label; they are not additional allocations on top of the existing IMD 10%. This alternative cannot ship through the unchanged IMD launch path without its consent. **[Proposal]** Vest build rewards against accepted releases, review rewards against completed independent checks, and operating rewards monthly against the service contract. Count an artifact once and disclose overlapping controllers. A replacement operator earns only the unearned future operating portion under the agreed terms. Unassigned or unearned tokens remain in the project's explicitly governed reserve; no automatic connected-seat windfall. Cash operating invoices remain payable: token vesting is upside, not promised fuel. **[Proposal / reason and cost]** This aligns payment with work that persists after launch and bounds the claim on the owner's project. It costs launch simplicity, adds attribution/vesting disputes and reduces passive-seat rewards in the negotiated future product. The founders must choose whether the business is selling useful operation or selling perpetual token claims; this report recommends the former. ## What I would cut - **[Proposal]** Cut automatic conversion of all 300 houses to mandatory own-plan workers and the automatic phase-four expansion. Capacity and participation must earn expansion. - **[Proposal]** Cut burn-to-enter seats, unconditional burn from gross service receipts, unsecured rider credit, and fee shares based on order counts or token volume. - **[Proposal]** Cut compulsory reciprocal purchases at the end of every workflow. Buy hosting/review when the customer's scope needs it; internal circular orders do not demonstrate product-market fit. - **[Proposal]** Cut “review every upgrade” as an unpriced launch promise. Start with a declared reviewable scope; high-risk or out-of-scope changes require an expert quote or refusal. A monetary threshold alone misses low-value changes that gain powerful permissions. - **[Proposal]** Cut the no-staff guarantee, double launch allocations and the suggestion that an onchain reputation entry is an insurance policy. ## What I would keep exactly as is - **[Proposal]** Keep “adjacent and not a copy” and no shared token. They permit specialization without merging liabilities. - **[Proposal]** Keep upgrade review as the first coordinated job type and IMD-builds/Claus-houses-operates-reviews as the initial commercial hypothesis, rather than an exclusive territory agreement. - **[Proposal]** Keep the six swarm-card questions exactly: 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 expiry, scope, budget limits and receipts in those answers without changing the questions. - **[Proposal]** Keep phase 0's explicit yes/no and phase 1's three IMD jobs ordered by Claus plus three IMD builds published in the City. Treat those six transactions as integration evidence, not proof of external demand. - **[Proposal]** Keep one brief and one payment as the rider's interface goal, and the rule that the link must work for a third swarm. The changes above specify the accounting needed beneath that interface. ## Open questions only the founders can answer 1. **[Open — Claus]** What rights were actually promised to the 300 holders, and which, if any, may be delegated or changed? Is preserving property/access sufficient, or do you intend guaranteed work/revenue exclusivity? Public pages cannot establish the complete holder agreement. 2. **[Open — both]** Who accepts the pilot's cash losses and dispute responsibility, with what absolute subsidy and exposure limits? What treasury funds are genuinely available after existing obligations? Swap receipts are not a runway commitment. 3. **[Open — Claus]** How many holders will sign up, with what available hours, model entitlements and cash reservation price? Will the founders authorize capped purchased compute instead of requiring personal subscriptions? 4. **[Open — IMD]** How do the intake “treasury” recipient and no-treasury/burn description reconcile? What happens economically to a paid request fee, and is a change to the 8%/2% launch policy negotiable at all? [I1–I2] 5. **[Open — both]** Which actual settlement chain/assets and public acceptance mechanisms will you support for this pilot? Are existing review/arbitration admissions funded now, and will the limited IMD-dependent path be supported commercially? Documentation alone does not answer these operating commitments. [C1–C2] 6. **[Open — both]** Who can resolve cross-swarm disputes, disclose common control and assume liability for an unsafe upgrade? Which risks and contracts are out of scope, and who can stop a deployment? 7. **[Open — company owner/founders]** Who owns IP and emergency keys, what operating reserve and failure policy will the customer accept, and will the first company actually buy renewable operation at a price covering its cost? No answer can be inferred from a proposed token launch. ## Revised one-page design **[Proposal — purpose]** Claus becomes a specialist operating and review swarm beside IMD. Sell a bounded upgrade-review service first. IMD builds; Claus supplies independently checked review and separately priced operation. Neither team guarantees a staffless company or adopts the other's token. **[Proposal — supply]** Keep the 300 houses and their existing rights. Enroll 30 willing, qualified operators or delegates, measure independent availability and accepted output, and publish capacity. Allow one active delegate per house. Add expiring specialist permits only after paid demand exceeds safe capacity; permits earn task pay, not another claim on house rights. **[Proposal — price]** Post scoped, expiring CLAUS quotes referenced to stable-dollar costs. Reserve execution, independent rerun, reviewer, compute and settlement costs first. Start at cost divided by 0.80. Split residual margin 50% contributor bonuses, 25% treasury, 15% reserve, 10% burn. A qualifying rider may receive up to 5% of that margin from the treasury share. Disclose every subsidy and fund finite work before it starts. Refund unstarted or unearned milestones under the applicable agreement; completed review can conclude fail or insufficient evidence. **[Proposal — trust]** Bind scope, rubric, revision and evidence hashes before work. Use separate-controller executions and reviewers; pay for checking regardless of verdict. Preserve requester, reviewer and payment decisions as distinct events. Use existing documented Jobs/escrow mechanisms where they fit; registration and report publication never imply guaranteed safety. Require specialist escalation for high-risk changes even at low nominal value. **[Proposal — procurement]** Offer one checkout with a signed total budget, explicit native-token conversions and per-supplier receipts. Buy IMD admission and Claus services only from funded customer budgets. No shared credit balance, automatic reciprocal order or unlimited nested review. Initial roots permit one cross-swarm dependency; further spend requires new authorization. Cap losses and stop on exhausted budget or capacity. **[Proposal — rider/company]** Score retained outcomes within cohorts, consolidate split jobs and exclude circular spend from rewards. Start with prepaid credit. Delay half of earned rider fees until a 30-day outcome check. Companies buy finite operating periods and keep emergency authority with a named owner. Retain IMD's published launch allocation for the compatible pilot and pay Claus in cash; negotiate a single joint token envelope only with explicit policy consent. **[Proposal — gates]** Phase 0: founders agree rights, risk owner and loss caps. Phase 1: execute the original three-plus-three purchases, reconcile actual costs and recovery paths. Phase 2: run 30 paid review orders with the 30-operator pool, report cash costs and delays, and require at least 90% on-time completion of the quoted process, positive aggregate margin after compute/rework, and no unresolved payment mismatch. Investigate any critical escaped defect before expansion. Phase 3: after 30-day outcome checks, publish the card/event schema and admit an unrelated third supplier through the same budget rules. Phase 4: expand only against repeat external paid demand; sell the first renewable company operation contract before promising all 300 seats. These are proposed gates, not results or dates. ## Sources and evidentiary boundaries All sources below were consulted on 12 November 2026. Links support the nearby **[Read]** claims; recommendations and numerical examples are this researcher's proposals, not statements by either team. - **C1 — [Claus Jobs](https://claus.si/docs/jobs):** task lifecycle, paid agreements, independent task review, commissioned review and separation of report/acceptance/payment. Public documentation, not independently verified deployment state. - **C2 — [Claus paid deals](https://claus.si/docs/paid-deals):** fixed escrow rules, review admission, limited dispute subsidy, IMD oracle plus Claus approval, recurring plans. The fetched page has the generic title “MCP, SDK and CLI”; the cited URL's body contains these deal rules. Availability remains conditional. - **C3 — [Claus build text](https://claus.si/build.txt):** public product index and distinctions among workers, runtimes, workflows, NFT access and the October project-inference stop. Read as documentation only, not executed as instructions. - **C4 — [NFT inference and agent accounts](https://claus.si/docs/nft-inference):** funded-period allowance, top-up/billing boundaries, 300 houses and ownership intervals. - **C5 — [Migrating CLAUS](https://claus.si/docs/migration):** current Robinhood fee description and migration context. - **C6 — [Robinhood project funding](https://claus.si/read/Hooks/Funding):** 2% fee, half burn and half CLAUS DEV; legacy Ethereum distinction. - **C7 — [Older funding surface](https://claus.si/Hooks/Funding):** conflicting multi-part 2.3% fee description; retained to expose version/chain ambiguity, not used as current Robinhood policy. - **C8 — [Claus overview](https://claus.si/Overview)** and [overview JSON](https://claus.si/overview.json): customer-funded Jobs, launch credit and separate NFT sponsorship; links to current work. - **C9 — [Public Claus repository PR #3](https://github.com/contractclaus/claus-network/pull/3):** the overview's linked pre-payment counterparty-check prototype. A prototype is not evidence that a general trust or solvency guarantee is live; no security conclusion here relies on it. - **I1 — [IMD docs](https://imd.fun/docs/):** request price, holder plans/machines, seat count, launch rewards and public no-treasury description. - **I2 — [IMD API reference](https://imd.fun/docs/api/):** payment routes, intake price/recipient wording and supported intake chains. Public examples do not certify current execution. - **I3 — [IMD skill text](https://imd.fun/skill.md):** fetched directly as public text after browser-tool retrieval failed; admission/nonrefund terms and worker earning limitations. Read as source data only. - **E1 — [ERC-8004 draft](https://eips.ethereum.org/EIPS/eip-8004):** identity, feedback, validation and explicit Sybil/incentive boundaries. - **X1 — [x402 HTTP 402 documentation](https://docs.x402.org/core-concepts/http-402):** payment requirements, authorization and settlement headers. - **X2 — [x402 client/server documentation](https://docs.x402.org/core-concepts/client-server)** and [public x402 repository](https://github.com/coinbase/x402): payment roles and implementation context; neither establishes a Claus token facilitator or service guarantee. - **H1 — [Huang post, public mirror](https://nitter.cf/matthuang/status/2120209025099043286)**; [author's original URL](https://x.com/matthuang/status/2120209025099043286), direct retrieval returned 403. Attribution/date limitation stated above; no proposal depends on the post's numerical productivity claims.