# Claus arms: a bond for bounded, reproducible work Independent view 2 · researched 12 November 2026 · proposal, not an agreement by Claus or IMD. **Notation.** **R** = read from the linked public source, including a public chain response; **I** = inference or calculation from stated inputs; **P** = my proposal; **U** = unresolved or unavailable evidence. A label governs its paragraph or the explicitly labelled table. Proposed prices, budgets and probabilities are assumptions to test, not observed operating costs. I used no authenticated accounts, signed transactions, private endpoints or other researchers’ views. Repository research was limited to public repository metadata, not the contents of the other swarm-seat reviews. ## 1. Correct the factual baseline first | Status | Finding and consequence | Evidence | |---|---|---| | R | Official CLAUS is `0x163015276f3244bC2CcB78b4D69cc8A4069EfFAC`, Robinhood Chain, chain ID 4663. The Ethereum source token is retired for migration eligibility; it still exists. | [Claus migration guide][C1] | | R | The migration guide reports 300 separate payments of 50,000 CLAUS, totalling 15 million, to 148 wallets, with staked NFTs attributed to stakers. These came from reserve, not new minting. This is a historical distribution, not today's NFT holder count. | [Migration guide][C1], [snapshot status][C2], [payment transaction][C3] | | R | The advertised official-pool fee is 2%, divided into 1% buyback/burn and 1% operations. This is a swap fee, not a proposed checking-service fee. | [Hooks][C4] | | R | The live operations status additionally exposes `buybackPips=10000`, `founderPips=9000`, `advisorPips=1000`, and advisor `0x1C31B012a8565567A72F7c661D7f0d9f92182853`. Thus “half treasury” is an aggregate description: the disclosed operational half has a founder/advisor subdivision. | [Read-only migration operations status][C5], fetched 11:37:20 UTC | | R | At Robinhood block **112,498,036**, the token's `owner()` is **`0x979d5810d32292358639f25294472f01c95bdc46`**; that administrator's `owner()` is **`0xbb145ca83272c3806d4ddc75ca1d5514789cf1c5`**. The token implementation slot points to **`0xd68727ddd50e8ba07e4be20d37a5ecf3dd814bff`**, not the implementation in the November 6 deployment manifest. | [Public RPC and methods][RH], [current implementation source][TS], [deployment manifest][C6]; reproduction below | | R / I | Current verified token source has unrestricted-amount owner-only `mint`. Its `migrationPaused` flag gates migration receipts, **not ordinary ERC-20 transfers**. I found no ordinary-transfer pause or blacklist in that implementation and its inherited ERC-20 code. Upgrade authority can nevertheless replace those semantics. The administrator source lets owner or operator call managed targets; filtering ownership selectors does not itself prevent upgrades. | [Matched token source][TS], [matched administrator source][AS] | | R / I | The context's single-wallet ultimate-control concern is justified, but calling that the only privileged actor is incomplete: there is delegated operator authority. Conversely, the unqualified statement in `build.txt` that core upgrades are unavailable must not be applied to Robinhood CLAUS. Its legacy `hook-health.json` identifies **chain 1**, not 4663. | [Administrator source][AS], [build guide][C7], [legacy hook health][C8], [chain-control disclosure][C9] | | R | Current CLAUS deal terms identify immutable escrow **`0xc2292af8b51109a27a4b4e28e9cf9f3aede4190c`**, Ethereum, asset **old CLAUS** `0x1b54E762aa34CF6E28E9C082F2848e28E45DA6b8`. The arbiter is `0xB104063f7a25a712a5899861941eb5cEB24EE105`. Native ETH has a different escrow. | [Public CLAUS deal terms][D1], [ETH terms][D2] | | R | The public arbitration capability is a **limited pilot**, not universal automatic adjudication: five panel members, quorum four, public committed deliverables, case and principal limits, and a genuine Claus decision. The disclosure expressly gives no guaranteed answer or payout. | [Public capabilities, `deals.arbitration`][D3] | | R | IMD documents 2,000 NFT seats and token-launch allocations: 8% to connected seats and 2% for accepted launch work. ERC-8004 itself uses an ERC-721 identity handle, but leaves trust models pluggable and payments outside its scope. An identity NFT is not necessarily a scarce economic seat. | [IMD docs][M1], [ERC-8004][E1] | | R / I | Existing Claus NFT benefits are not zero: the NFT guide describes houses and a sponsored inference plan of up to $5/NFT/day, conditional on funding. “Gain nothing new” should preserve existing documented rights and accrued claims, not falsely say these never existed. | [NFT account and house guide][C10], [Hooks][C4] | **U.** I did not independently verify every migration payment recipient against the payment receipt, or execute the advertised arbitration path. The 148-wallet payout is a project-reported historical fact. The more recent 147-beneficiary NFT measurement below does not contradict it. Public web responses also differ in freshness: the cached web rendering of operations status was older than the direct public JSON fetch. The earlier proposal site returned an access error; nothing here is attributed to its inaccessible contents. [Snapshot][C2], [NFT snapshot][N1], [earlier proposal][OLD] **P — decision.** Keep open, returnable CLAUS bonding and skill trials, but make an arm a provider of **limited, block-pinned factual attestations**. Start with contract control maps. Do not sell “safe contracts,” unlimited loss coverage, token-weighted truth, or independence supposedly proved by two wallet addresses. Launch a paid pilot only after collateral administration and dispute funding are credible. ## 2. Which seat, and what real networks teach us — A | P: comparison and recommendation | Useful property | Reason I would accept or reject it | |---|---|---| | Returnable CLAUS bond + trial — adopt conditionally | Open entry; skill is tested; economically accountable work; exit remains possible. | A bond cannot prove skill, distinct control or adequate compensation. Reserve exposure per job and pay for work. | | Scarce NFT seats, as documented by IMD | Persistent identity and a bounded scheduling population. | Buying the identifier does not test control-check competence. Scarcity can charge newcomers for access unrelated to work; IMD's launch allocations also create conflicts for reviewers. [M1] | | The 300 Claus houses as seats — reject | Existing community and recognizable profiles. | Historical ownership is not checking skill. It creates a hereditary bottleneck and revives the income entitlement being removed. Preserve houses separately. [C10] | | Open permits without bond — allow only for training, publishing and challenges | Lowest capital barrier; useful discovery of new reviewers. | Reputation and service-fee refunds may suffice for casual information; they do not give a fresh pseudonym much to lose. Do not present these as bonded attestations. | | Delegated staking — defer | Holders can finance a capable but undercapitalized operator. | It separates the decision-maker from the loss-bearer, introduces delegation monitoring and passive fee claims, and may route assignments toward accumulated capital. No stake-weighted scheduling. | **R.** The Graph documents a 100,000 GRT indexer minimum, delegated capacity up to 16 times self-stake, query fees and issuance-funded indexing rewards. Its own September 2026 explanation says **15.2% of 2025 indexing rewards went to indexers not serving queries or indexing data consumers used**, motivating service-based reward eligibility. This is observed reward misalignment, not proof that every indexer was unprofitable. [Graph tokenomics][G1], [service eligibility explanation][G2] **I.** Borrow the distinction between self-stake and delegated support, but not issuance for mere presence. Removing emissions does not make operating costs disappear: it makes customer fees and demand the binding constraint. The Graph example cannot establish that a no-emission Claus service will cover its costs. [G1], [G2] **R.** Kleros's Doges on Trial pilot reported 49 depositing addresses and 1,035 submissions, 135 disputed, by session 22. Its largest depositing address had a median **32%** of deposited stake across sessions. Its subsequent write-up documents explicit bribery attacks and an interface problem causing no-shows; its final analysis discusses lazy voting and ambiguous cat-plus-dog images. [Pilot observations][K1], [attack analysis][K2], [lazy-voting analysis][K3] **I.** This is evidence that challenge games can produce real review activity and also concentrated voting, participation failures and coordination on interpretations. Two matching answers are not two independently established facts; common models and shared source errors correlate failures. [K1], [K2], [K3] **R.** UMA's optimistic oracle documents refundable proposer and dispute bonds, a challenge period, and escalation to staked DVM voters. Its stated 99.8% unchallenged-resolution figure describes usage, not independently verified accuracy. [UMA mechanism][U1] **I.** An unchallenged falsehood is cheap if nobody is paid to look. Kleros explicitly describes the challenger's dilemma: honest submissions leave hunters with search costs but few winning cases. It also explains how very large deposits deter submissions. Its account of V1 identifies mainnet gas costs as an obstacle to lower-value claims. These are reasons to budget recurring test work and disputes, not assume a volunteer challenger will always appear. [Curate parameters][K4], [Kleros V1 limitations][K5] **P.** Keep self-funded bonds for the first 100 paid checks. Later test non-transferable, job-specific sponsorship only if the operator still supplies at least **50%** of each job's required collateral and neither sponsor nor stake size buys votes or assignment priority. The 50% is a proposed alignment floor, not an empirical optimum. Do not introduce a tradable stake receipt or a yield promise. ## 3. The actual design I recommend — B, C and product boundaries ### Admission, collateral and timing **P.** Each arm runs on its operator's own machine and AI plan. Use a **100,000 CLAUS minimum registration bond**, with **dollar-valued job exposure**, rather than pretending one fixed token amount guarantees a dollar remedy. The token floor is deliberately a participation parameter: at 1,000 minimum-bond arms it uses approximately 10% of the measured supply. It is **not** a security estimate. Financial security comes from the additional capacity rule below; an operator may need substantially more CLAUS to receive even one job. **P.** Define `B` as actually deposited CLAUS and `P_ref` as an executable USD-per-CLAUS reference conversion. Count only **50%** of `B × P_ref` as eligible coverage. Admit a job only after atomically encumbering its full tranche. A dollar label without a usable conversion, reserve and liquidation policy is not dollar coverage. | P: parameter | Recommended number or rule | Reason | |---|---|---| | Initial trial | 12 randomized cases; at least 10 correct, **zero missed critical control paths** | Allows two noncritical errors while preventing an owner/proxy/mint miss from passing. Include “unknown,” deceptive getters and inherited roles. | | Repeat competence check | 2 hidden calibration cases every 30 days; new implementation families require a fresh trial | A one-time leaked question bank is not continuing competence. | | IMD shortcut | Waive the initial trial only for a verifiable record of **10 accepted comparable control checks in the previous 30 days**; otherwise normal trial | Owning an IMD NFT alone is not evidence of this skill. Everyone posts the same bond and passes ongoing calibration. | | Control-check scope | One chain; one named target/proxy and its control graph, capped at **10 relevant contracts** | Bounds inference, evidence size and price; return a new quote for a larger graph. | | Standard job encumbrance | **$125 of haircut-adjusted collateral per arm**, locked for that job | Two arms supply $250 adjusted collateral for a remedy capped at $100 plus challenge costs. This is modest accountability, not exploit insurance. | | Maximum simultaneous exposures | `min(4, floor(0.5 × P_ref × B / 125))`, counting pending work **and every still-challengeable verdict** | No reuse of collateral across a hidden backlog. Four is a concentration/operations limit, not a claim that four jobs are always funded. | | Top-up for all four slots | Total bond at least `max(100000, ceil(1000 / P_ref))` CLAUS | $500 adjusted collateral requires $1,000 unhaircut collateral. No current conversion or future token value is assumed here. | | Collateral reference | Lower of a **30-minute TWAP** and a size-specific executable liquidation quote, after fees; quote age at most **60 seconds** | A spot trade in the same token must not cheaply manufacture capacity. An unavailable or clearly manipulable reference admits zero new exposure. | | Valuation stress rule | 50% haircut; re-evaluate before every assignment and hourly while exposed | Absorbs an ordinary adverse move, not a guarantee against a crash. A shortfall stops new work; allow **48 hours** to top up. Do not slash an operator merely because CLAUS moved. | | Assignment | Two different declared controllers; random selection among qualified available operators; at most one arm of a controller on the job | Capital and multiple wallets must not buy both sides of a check. Scheduling is publicly auditable, not stake-weighted. | | Sealing | Each arm commits a salted hash covering chain, block hash, target, schema version, answer and evidence digest within **24 hours**; reveal within **2 hours** after both commitments | Prevents copying an already visible answer and makes the exact evaluated proposition recoverable. Publish both reports after reveal. | | Disagreement | Pay both diligent reports; fund one additional independent check; after **24 more hours**, unresolved fields publish “unknown/disputed” | Do not force matching answers or punish an honest disagreement. Third-arm agreement is evidence, not automatic grounds to slash the dissenter. | | Standard challenge period | **7 days** after complete evidence publication | Gives part-time reviewers a weekend; the clock starts only when evidence can be fetched. | | First dispute decision | **7 days** after an admissible challenge; one **48-hour** appeal-filing period; appeal decision within **7 days** | Fits ordinary exits while paying for a real hearing. | | Absolute case limit | **30 days** after the challenged verdict | If adjudication fails, withdraw certification, refund the fee from reserve, publish “unresolved,” and release unslashed tranches. Do not give an attacker an indefinite bond freeze. | | Unbonding | **14 days** after exit request, and never before the last job's challenge/hearing process ends; only the disputed tranche remains locked | Covers the normal seven-day challenge plus seven-day hearing. A timely appeal can extend that tranche, subject to the 30-day case limit. | | Idle bond | **0** issuance, **0** idle yield, no automatic work entitlement | Work fees must pay for work and its financing cost. | **P.** Reference formation must use independently sourced chain/USD inputs and test actual exit depth, not merely a reported pool ratio. Founders must name the reference implementation and publish manipulation tests before admitting dollar-accounted jobs. If that cannot be made credible, run an explicitly informational, fee-refund-only pilot with no advertised dollar collateral coverage. This is a deployment condition, not an invented oracle address. **I.** Operators bearing token volatility may rationally decline a long lock, require higher fees, or top up at inconvenient times. Without emissions, the legitimate offsets are shorter residual exposure, predictable work fees, optional job sponsorship and a cash reserve that pays service remedies promptly. A stablecoin collateral option would improve economic safety but weaken the “CLAUS is the bond” thesis; I would prefer it to dishonest dollar guarantees if the token cannot support reliable collateral valuation. ### Exactly what can be slashed **P.** Slash **a false signed field proved against its committed state**, not an outcome, sentiment, missing prediction or mere disagreement. At launch whitelist only these forms: | P: slashable example | Evidence required | Excluded claim | |---|---|---| | “At chain/block/hash X, `owner()` returns address A.” | Reproducible archive call with selector and decoded result. | “A is the only human who controls it.” | | “The EIP-1967 implementation/admin slot equals A at X.” | Exact slot, pinned storage value, proxy family and code hashes. | “This proxy can never change.” | | “This verified implementation exposes this mint/pause/fee-setting path to role R.” | Matched bytecode/source, complete stated scope, role getter/storage and a reproducible call path. | “No backdoor exists anywhere” or “this token is safe.” | | “This named LP position is held by A, with withdrawal route R.” | Position identity, ownership/custody and applicable manager permissions at X. | Treating all pool-manager inventory as one holder's property. | | “A listed watched field changed between X and Y.” | Both snapshots and the stated observation/notification deadline. | “Nothing important happened” outside the subscribed fields or during an explicitly recorded outage. | **P.** An omission is slashable only if the pre-agreed finite checklist explicitly required that field/path and its reproducible evidence was available. “Unknown: source unavailable” is allowed with failed-read receipts and a reason. Repeatedly returning unknown can fail a competence trial or make an operator unsuitable for a class; it never converts uncertainty into a false statement. Timeouts lose the unpaid work fee, not collateral; **three unexplained timeouts in 30 days** require requalification. **I.** Making every source-code interpretation slashable would punish ambiguity and create incentives to abstain or follow the expected majority. Kleros's ambiguous-image and lazy-strategy observations support starting with much narrower predicates. [K3] ### Challenger economics and adjudication **P.** Anyone can submit a correction or evidence packet without an account fee. To freeze collateral and demand an adjudication, post a **$25 CLAUS-denominated challenge deposit**, valued at filing, plus a **$5 non-refundable processing fee**. A funded public-interest challenge sponsor may post that deposit for someone else; it gets no vote. Challenges must identify one committed field, one contradiction and replay instructions. Merge duplicates by `(job, field, alleged counterfact)`; reward the first complete committed proof, not the fastest copied transaction. Permit a **24-hour commit/reveal** for a hunter to protect a proof from front-running: the commitment must arrive before the seven-day deadline, and a timely reveal preserves that filing time. **P.** The $25 deposit makes indiscriminate freezes costly but is too small to cover a $150 panel alone. Therefore a losing admissible challenge loses its deposit toward the panel bill; the service reserve absorbs the remainder. Cap intake by **prefunding every accepted hearing**. Free corrections stay possible even when paid adjudication capacity is exhausted. A frivolous or duplicate packet cannot restart a window. This limits, but does not eliminate, funded griefing. **P.** For an independent non-IMD-built target, buy the existing **five-member/four-agree IMD process** only after confirming its quote, eligibility and decision authority. For an IMD-built target, or a target built by Claus, use an **outside five-person technical roster, three agreeing**, with disclosed affiliations and **$30 each: $150 per first hearing**. Pay for analysis, not for siding with the customer or agreeing with a majority. A **seven-person outside appeal, four agreeing, $210 total**, must exclude the initial judges. Require a **$250 appeal deposit**; refund it for a successful appeal, charge actual appeal cost on failure. Founders and the buyer must accept the roster and enforceable signed-decision mechanism before the sale; I am not asserting that such a service is already contracted. **R / I.** IMD's API allows configurable panel sizes and quorum; “one per seat” is its documented selection unit, not “one proven independent controller.” It advertises an oracle request at 0.5 IMD. Neither statement guarantees five independent experts, recusal controls or a dollar cost below $150. External case budgets here are proposals, not conversions of that public token fee. [IMD oracle API][M2] **P.** Use a priority waterfall, not a fixed percentage that underpays an actual victim: 1. **Customer first:** refund the checking fee and documented recheck expense, **$100 total maximum per standard job**, once only across both arms. State prominently that this does not insure protocol deposits, trading losses or foregone opportunities. 2. **Challenger:** return the successful challenger's deposit, then pay **20% of gross slash proceeds, capped at $50 per standard job, subordinated to the customer remedy**. The reserve tops up the proven-error bounty to $50 if proceeds are insufficient; no extra bounty for multiple false fields caused by the same root error. 3. **Reserve reimbursement:** recover actual prepaid hearing/appeal expense, never a profit share. Do not take a panel fee out of the customer's promised refund. 4. **Burn:** destroy **100% of the remainder** using the token's actual burn operation, with the transaction and supply delta published. **P.** On a fully proven erroneous joint verdict, slash each culpable arm's **job-encumbered tranche**, not its collateral committed to unrelated correct work. If one arm's disputed field was correct, do not slash it. No more than the agreed encumbrance can be extracted for one job. Repeated wrong jobs can exhaust a bond and require retraining. Publish who lost what, the counterproof and the signed decision. **P — numerical example.** If realized slash proceeds are $250 and customer remedy is $100, the challenger receives $50; $100 remains for a $150 hearing bill, so the reserve absorbs $50 and the burn is zero. With $500 proceeds under the same remedy and $150 hearing cost, the burn is $200. The normalized splits are therefore **40/20/40/0** and **20/10/30/40** percent for customer/challenger/panel reimbursement/burn. Promising a fixed burn percentage before these obligations are met would be misleading. **I / P.** A controller operating both arms, buying its own check and challenging its own planted error must not make money: refunding its own fee is not a gain, the bounty is smaller than forfeited collateral, real panel costs are charged, and the rest burns. Do not pay a separate treasury-funded bounty merely for an accusation or majority disagreement. Hidden affiliation and external bribes remain possible. For external gain `G`, detectable fraction `d`, and realizable slash `S`, deterrence requires roughly `d × S > G`; neither matching answers nor a $125 tranche secures a contract with enormous exploitable value. Sell a check with a limited remedy, not that security claim. ### Products and conflicts **P.** Product (a) ships first: a public, versioned control form with per-field evidence, two sealed reports, and an explicit unknown state. Use different source/RPC paths where feasible and record model/provider families without pretending that different names imply independent cognition. **P.** Product (b) is a later **$299 starting quote**, two arms at **$60 each**, at least **$500 adjusted collateral per arm**, one open upgrade exposure per controller, and customer remedy capped at **$500**. It covers a single immutable implementation hash and calldata package, not an evolving branch. Require **14 days of project timelock**, a review within **48 hours**, a **7-day challenge**, and no execution before finality; a challenge extends the execution gate. If the project's timelock cannot enforce that gate, label the review advisory. Keep broad vulnerability opinions non-slashable. Reprice anything whose real review/panel budget exceeds the quote. **P.** Product (c) is **$29/month** for up to **10 named fields**, after the baseline check: poll every **5 minutes**, notify within **15 minutes**, two independently operated watchers, and daily signed checkpoints. Each arm reserves **$125 adjusted collateral per active monthly subscription**, including its seven-day post-termination challenge tail. The same four-exposure ceiling includes these subscriptions; do not disguise a month's exposure as one instant check. Alerts can publish immediately as “unconfirmed” and acquire two-arm confirmation later. Otherwise a failed second watcher would suppress urgent warnings. Limit refunds to the month’s fee plus recheck, within the same **$100** cap; do not sell incident response or continuous exploit detection. **I / P.** Two conflicts require changing the original wording. An arm necessarily holds CLAUS as collateral, so **Claus arms cannot be independent reviewers of CLAUS itself** under the no-holdings rule; buy outside reviews paid in another asset. IMD seats that automatically receive launch tokens must decline affected reviews unless allocations can be refused and holdings/conflicts cleared. A trial exemption does not cure inventory or builder conflicts. Disclose direct holdings, allocations, employment and declared common control; disclose that indirect hidden exposure cannot be excluded. [IMD allocation rules][M1] ## 4. Fees that can actually pay for the work — D **P.** Keep the proposed order, but contract for all costs before accepting the customer's money. “Arms first” means reserved work compensation, not draining funds needed for a promised refund. Receive CLAUS on Robinhood, fix a **60-second quote**, and reserve or convert the dollar budget promptly with a **1% maximum execution slippage**; a failed quote is not a completed sale. Include swap/project fees in the execution budget. Do not silently charge token holders for service operations through the existing swap treasury. **P.** Offer a tightly scoped control check at **$69 during the pilot**. Test **$49** only after at least **150 paid checks/month** and actual cost evidence support it. These are proposed customer prices, not competitor quotations. Do not market a $5 complete check if two independent operators and adjudication cannot be paid at that price. | P: standard-check budget assumption | Dollars/check | Derivation | |---|---:|---| | Two initial arms | 24 | $12 each: $3 inference/tool use, $8 for 20 minutes of review at $24/hour, $1 financing/risk allowance. A personal AI subscription is not zero marginal economic cost. | | Additional check on disagreement | 2 | Allow a $12 third arm on roughly one in six jobs. | | Planted-defect work | 3 | 10% additional paired calibration work costs $2.40 per paid check; round up for maintenance. Operators know calibration exists, not which assignment it is. | | First-hearing provision | 3 | Assumed 2% hearing incidence × $150; **not an observed dispute rate**. | | Appeal provision | 2 | Assumed 0.5% of jobs × $210 external appeal cost, rounded up; losing appeal deposits reduce realized expense. | | Proven-error bounty | 1 | Assumed 2% proven errors × $50; separate from paying the panel. | | Refund/recheck provision | 4 | Assumed 4% × the full $100 remedy cap, rather than assuming only the service fee needs refunding. | | Settlement/execution | 2 | Gas, swaps and receipt publication; reprice if actual cost is higher. | | Coordination | `3 + 600/J` | $3 variable per job plus $600/month fixed cost spread over `J` paid jobs. Illustrative fixed budget: $200 infrastructure and 10 hours at $40 for dispatch, bookkeeping and support. | | Total | **`44 + 600/J`** | Includes all the rows above, with the refund approximation disclosed. | **I.** At 30 paid checks/month, the model costs **$64/check**, leaving $5 from a $69 quote. At 150/month, cost is **$48**, leaving only $1 at $49; at 100/month, $49 would lose $1/check. A $49 quote breaks even at **120/month** under these assumptions, before any unbudgeted cost. With 10% rather than 2% first hearings, the panel provision rises from $3 to $15 and the $49/150-job model loses $11/check. Work is not sustainably funded just because burns occur. **P.** Hold the expected **$15/check** for additional checking, calibration, hearings, proven-error bounties and refunds in a ring-fenced dollar reserve, rather than treating it all as distributable cash. Account for expenses and release unused provisions only after exposure ends. Have Claus fund **$5,000 initial working reserve from an expressly disclosed existing operating budget**, not emissions or bond principal. If that funding is unavailable, presell a smaller explicitly funded pilot or do not sell remedies. No new NFT payment is created. **P.** Begin service-fee burns only after **100 paid completed checks**, **90 days** of operating history, all due work/refunds paid, and reserve assets above `max($5000, outstanding customer remedy caps + 90 days of budgeted dispute/bounty/calibration/recheck expense)`. The 90-day budget must use the worse of observed incidence or the assumptions above. Burn only realized net surplus after reconciliation, **monthly**, with receipts. Halt burns automatically below the reserve target. This can mean no service-fee burn for a long time. The existing swap buyback program is a different flow and is not changed by this proposal. [Current swap fee description][C4] **P.** Keep coordination at **documented cost**, including real labor and fixed infrastructure, under a quarterly approved budget with monthly receipts. “No treasury cut” must mean no unexplained percentage skim; it cannot mean that assigning, publishing and resolving work are free. Publish separate ledgers for customer fees, reserve assets/liabilities, coordination reimbursements, collateral movements and burns. Do not count token appreciation, deposits or returned bonds as revenue. **P.** For the later $29 monthly watch, budget $12 to two watchers, $7 reserve/exception handling, $5 coordination and $2 infrastructure, leaving $3 before any burn. These are allocation targets to verify at scale, not a basis to subsidize a new low-volume product. A $299 simple upgrade quote starts with $120 operator work and $100 reserve/coordination/execution; $79 remains to absorb scope uncertainty. Do not infer that these later products have demonstrated margins. ## 5. Measurements and their limits — E, F, H ### E. CLAUS locked and the administrative promise **R.** At Robinhood block **112,498,036** (`0x6b49574`), hash **`0xca6bf6dfc47b9e764c4e80f932f1338c6da341921ed05f73425d98b04d58340a`**, timestamp **2026-11-12 11:38:59 UTC**, `totalSupply()` returned **983,872,663.400486075752008619 CLAUS** (18 decimals). Replaying the token's **14,881 Transfer logs** from genesis to that block produces exactly the same sum of positive balances. The first returned token transfer is at block **107,148,941**. This is current Robinhood total supply, not circulation, original issuance, or the sum of Ethereum and Robinhood ledgers. [Public Robinhood read endpoint][RH], [token explorer entry][RT] | I: hypothetical enrolled arms, each posting exactly the recommended 100,000-CLAUS registration minimum | CLAUS bonded | Share of measured Robinhood total supply | |---:|---:|---:| | 10 | 1,000,000 | 0.101639% | | 100 | 10,000,000 | 1.016392% | | 1,000 | 100,000,000 | 10.163917% | **I.** These are exact **registration-floor scenarios**, not estimates that every arm can work four jobs. Under the recommended exposure gate, the all-four-slot total is `N × max(100000, ceil(1000/P_ref))` CLAUS, and its supply percentage is that quantity divided by **983872663.400486075752008619**, times 100. With no approved, size-tested reference in this proposal, I cannot truthfully give one numeric full-capacity token requirement. The base table deliberately does not invent one. Burns and later minting would change the denominator; deposits merely lock existing units temporarily. A burn credits no holder's wallet and creates no new payout entitlement; equal rights per token do not mean equal cash income for every wallet. **I.** A single ultimate owner, an uncapped mint function and replaceable transfer logic weaken both collateral and the meaning of burning. Minting can dilute the relevant denominator; a replacement implementation can prevent payout even if the escrow's bookkeeping is perfect. Moving ownership into an administrator contract is not a decentralization guarantee when one wallet controls that administrator and a delegate can call privileged targets. [TS], [AS], [C9] **P — what Claus must lock before calling this a credible bond.** - Remove discretionary token minting **irreversibly**, including every upgrade route that could restore it. Preserve only fixed, already committed migration entitlements; do not confuse receipt settlement with open-ended minting. - Make ordinary transfers, bonding, refunding and burning immutable against pause, blacklist, confiscation or new transfer tax. Freeze the token's economic implementation and its ProxyAdmin upgrade power after independent verification; keep cosmetic metadata in a separately limited mechanism if needed. Renouncing a token owner while another proxy administrator can replace its code is insufficient. - Put bond custody in an independently reviewed, immutable contract: no admin sweep, arbitrary-recipient withdrawal or retroactive slash rules. Only narrowly specified signed final decisions can slash locked job tranches. An emergency switch may stop **new admission**, never withdrawals already earned under the rules. - For operational hook/vault changes that remain necessary, use **3-of-5 independently controlled signers and a 14-day enforced notice/timelock**, publish target/selector/calldata and a conflict register, and hard-limit fee/withdrawal authority. Protect the fee reserves from those powers. These operational controls do not substitute for freezing the collateral token's economics. - Publish supply, owner/role/implementation/LP custody snapshots, timelock queues and reserve reconciliations. Because arms hold CLAUS, commission outside reviewers for these snapshots. **U.** These are proposed conditions, not completed changes. Neither a one-hour notice promise nor verified source alone makes them true. The live administrator source actually refuses renouncing its own ownership, so founders would need an explicit migration of the relevant powers into the final restricted arrangement, not a claim that one renounce call solves everything. [C9], [AS] ### F. Public launch volume and first customers **R — repository observation.** I fetched public metadata through [GitHub's organization repositories API][GHAPI], sorting by creation descending, 100/page, until reaching before 5 November. I counted non-fork repositories in `identity-md-launches` whose names start with `launch-`, using `created_at`, not `pushed_at`. The seven complete UTC dates give: | UTC day | I: new `launch-` repositories | |---|---:| | 2026-11-05 | 20 | | 2026-11-06 | 40 | | 2026-11-07 | 21 | | 2026-11-08 | 18 | | 2026-11-09 | 2 | | 2026-11-10 | 8 | | 2026-11-11 | 8 | | **Seven-day total / daily mean** | **117 / 16.71** | **I / U.** This measures public launch **repositories**, not successful token deployments, unique products, funded customers or paid checking demand. It includes websites, continuations and repeated attempts. `goal-` repositories are excluded; three such repositories appeared in this seven-day window. The incomplete 12 November day is excluded, including the current research wave. Repository contents and other researchers’ views were not read. An organization repository can be deleted, made private or created later, so this is a retrieval-time observation, not an immutable chain count. [GitHub organization][GH], [API][GHAPI] **U — explorer requirement.** The Robinhood Blockscout token list and CLAUS page produced no readable token list through the web tool; direct public API requests and Robin Etherscan requests returned HTTP 403. Consequently **I could not establish an explorer-derived, chain-wide “new tokens/day” count**. No zero or guessed rate is substituted. [Blockscout token list][RL], [token entry][RT], [Robin Etherscan token list][RE] **R / I — public-chain fallback.** A documented public Pinax RPC did permit chain reads. I therefore additionally measured a carefully defined class for the complete UTC day **11 November 2026**: distinct addresses emitting a three-topic ERC-20-shaped mint Transfer during the day, with no runtime code immediately before the day, and 32-byte `decimals()` and `totalSupply()` responses at day end. The resulting **4,256 token-like addresses/day** are an exact count for this mechanical definition, **not the unavailable explorer total, a full ERC-20 conformance test, or a count of legitimate launches**. There were 381,029 mint events, 4,862 distinct ERC-20-shaped emitting addresses, and 4,256 with absent pre-day code; all 4,256 passed the two day-end metadata reads. The proxy can miss genuine tokens and include nonconforming imitations, so it is not a formal lower bound for all ERC-20 births. The exact coverage appears in the measurement appendix below. [Robinhood documented providers][RH], [Pinax read endpoint][RPC-RH] **P.** First sell to launch teams needing a verifiable admin disclosure, project treasuries checking a dependency before integration, and launch directories/wallet interfaces buying reusable control forms in batches. The buyer should be purchasing a dated control map with evidence, not a badge claiming exploit safety. Approach IMD builders as customers while keeping IMD-built disputes outside IMD. **R / I.** Claus already lists adjacent services: Blockpin describes block-pinned admin/upgrade checks, and FlowExecutor advertises $5 grades and a $250/month lane. These are public offers, not verified invoices or volume. They show competition and plausible product vocabulary; they do not establish willingness to pay $69 for our more expensive two-arm service. [Public project directory][DIR] **P / I.** Test 30 independently purchased control checks before scaling. At the measured 16.71 repositories/day, an explicitly hypothetical 5% conversion is **0.84 checks/day**, around **25/month**, not 1,000 employed arms. At $49 that does not cover the fixed-cost model. Use $69 for the bounded paid pilot, $49 only with measured volume/cost support; consider batches priced from demonstrated reuse, not arbitrary fee discounts. Later test $299 frozen-upgrade work and $29/month narrow watches. None of those price points is validated by token issuance counts. ### H. Concentration: addresses, custody and hidden control **I — common method.** For balances/counts `b_i`, top-k share is `sum(largest k b_i)/sum(b_i)`; HHI is `sum((b_i/sum(b_i))²)`, on a 0–1 scale. Addresses are lowercased and zeros excluded. Do not compare a custody address directly with a known beneficial owner, and do not interpret address HHI as controller HHI. | I: measurement | Supply/count and positive addresses | Largest / top 5 / top 10 | HHI | Pin and evidence | |---|---|---|---:|---| | CLAUS on Robinhood, raw balances | 983,872,663.400486… CLAUS; **3,983 addresses** | **9.8416% / 21.5976% / 29.8839%** | **0.0177659** | Block **112,498,036**, hash above; Transfer replay, total-supply equality and top-ten `balanceOf` checks. [RH], [RT] | | CLAUS, remove only the identified PoolManager | 887,044,002.823777… CLAUS; **3,982 addresses** | **4.5443% / 15.1106% / 23.6705%** | **0.0099405** | Same pin. This is a sensitivity view, not free float or a beneficial-owner census. [RH], [TS] | | Claus NFTs, raw onchain custody | 300 NFTs; **69 addresses** | **61.6667% / 73.0000% / 77.3333%** | **0.3871556** | Project's block-pinned NFT dataset, Ethereum **26,398,605**; corroborated by public explorer holders. [N1], [N2] | | Claus NFTs, staking beneficiary attribution | 300 NFTs; **147 beneficiaries** | **7.3333% / 26.6667% / 38.6667%** | **0.0216889** | Same dataset and pin; project-reported staker mapping, not independently reconstructed vault history. [N1] | | IMD seats, pinned `ownerOf` census | **2,000 NFTs; 789 addresses** | **3.8500% / 12.1000% / 19.2500%** | **0.0069405** | Ethereum **26,398,619**; all IDs 0–1999 queried at one block, no failed calls. [ETHRPC], [M3], [N3] | **R.** Claus collection: **`0x80396c7131159eb92e7e84839e51c9887a7a95d9`**, Ethereum. Its staking vault **`0x4e5916428452891e1deb59d162c20266ed08cad4`** holds **185 NFTs**, not 185 proven seats of one controller. The NFT dataset's pin is hash **`0x46f2646db69276154916df2a9310d10df6b52ca3575a7986c6777c6015d22add`**, observed **11:37:19.610 UTC**; block time is 11:37:11 UTC. IMD collection: **`0x0000ec93127baa929e58e97dd0095a2bfb38ec1d`**, Ethereum. Its census pin is hash **`0x79f92c9ba823feb06709e70f54f3278826f615c6ce23aaa3b2d46ee9d3d0fcb0`**, timestamp **11:39:59 UTC**. [N1], [N2], [M3], [ETHRPC] | R / I: largest CLAUS addresses at the Robinhood pin | CLAUS, rounded | Share | |---|---:|---:| | `0x8366a39cc670b4001a1121b8f6a443a643e40951` — PoolManager identified in published contract source | 96,828,660.577 | 9.8416% | | `0x63b9a33525ebbfc765ac9c81e852c57bff0a2cef` — control/beneficiaries not established | 40,309,609.986 | 4.0970% | | `0x44e4208f7806e009beff792f7c323ff1b359d308` | 27,029,617.278 | 2.7473% | | `0x1c57ac0d42d70665bbcba2f11e58cd9210500fb9` | 24,924,923.424 | 2.5333% | | `0x9d42cee9a2a6bd1b8a185bb9aed6dcb5ce53b7ed` | 23,399,613.116 | 2.3783% | **R / I.** At that same pin the token contract holds **4,157,673.496 CLAUS** and the disclosed ultimate owner wallet holds **5,571,333.695 CLAUS**. Neither number establishes all founder-controlled holdings. Several large addresses return 23 bytes of code; code presence alone must not be used to classify them as pooled custodians or independent organizations. The pool's balance is inventory across positions, not the project owner's personal balance. [RPC-RH], [published source bundle][C11] | R / I: largest Claus NFT beneficiaries in the project snapshot | NFTs | |---|---:| | `0xc4210bfd73206c3208ae1f0854ee146ed8f1271b` | 22 | | `0x0feaf114f65361dfaaf981f2662e3843376dacbe` | 16 | | `0x85b1f16a7cc5da288dc796ae8300099c3ac9c66c` | 16 | | `0xdb585820d3b56dbe452b2d7774e7852dbd0976ea` | 15 | | `0x9bf5c4f08c797a05933fcd3db4cb94b62b184709` | 11 | | R / I: largest IMD seat addresses in the pinned census | NFTs | |---|---:| | `0xb7a832db45f240f876432bf8b222bff30b4be8ff` | 77 | | `0xb64199cd690aa4e78f1a6625660347ce63031d72` | 54 | | `0x80271ce20184e38f4afe90d4ca134304d197aca2` — explorer-labelled IMDStrategy | 47 | | `0xe3de112ae639b45b7d22c4e42742020db679d80c` | 34 | | `0x6ba9ad599e6271252817933831e3148936cf742a` | 30 | **U / I.** The IMD strategy's underlying claims were not unwound; exchange/custody users, delegation, borrowing, shared operators and common controllers behind fresh wallets remain unknown. An initial unpinned explorer holder pagination summed to **2,002** rather than the collection's 2,000; I rejected that as a current census and replaced it with the pinned `ownerOf` measurement. It is a useful example of why a moving index or stale balances cannot support exact ownership claims. None of the tables measures the concentration of *active qualified workers* or actual dispute votes. [N3], [ETHRPC] **P — cheapest workable common-control test.** Require a signed controller declaration binding all known operational and payout wallets; verify both signatures; screen public shared funding, payout and deployment histories; prohibit selecting two arms from the same declared controller; randomize counterparts and publish assignment pairs. Add a **14-day probation** with **five calibration jobs** before regular assignment. Trials and probation are disclosed unpaid synthetic training, never unpaid customer work; candidates fund their own inference. Budget **$10 one-time onboarding review** from a disclosed pilot setup budget, not a recurring NFT fee or idle-bond reward. Use pairwise-answer/evidence similarity to trigger review, never as standalone proof of hidden identity. **I.** This raises the cost of casual duplication and creates an attributable rule. It does **not enforce one human/controller, one voice against a determined pseudonymous adversary**. A common funder can be an exchange; unrelated wallets can share a model; the same controller can fund fresh wallets independently. ERC-8004 explicitly warns about Sybil/spam risks and treats identity/reputation as separate from stronger validation. Larger stakes do not solve identity. [E1] **P.** Advertise “two declared distinct operators, with conflict screening,” not “proved independent minds.” For high-stakes work, obtain an externally verified organizational identity and contract or use named independent professionals; disclose the resulting admission restriction. Do not claim both completely permissionless pseudonyms and perfectly enforced one-controller voting. ## 6. What I would cut, including the honest NFT answer — G **P.** Cut permanent burn-to-enter seats, house-based admission, passive bond yield, launch-token compensation, stake-weighted assignment, and any new NFT share. Cut blanket IMD trial exemptions based on asset ownership. Cut automatic slashing for disagreement, broad safety opinions, whole-bond penalties for one bounded mistake, and “customer made whole” language that suggests uncapped insurance. Cut an unconditional burn promise, cheap-loss-making check prices, and the suggestion that two fresh wallets prove independence. Defer delegated staking, frozen-upgrade enforcement and watches until the first product has paying customers and auditable error/cost data. **I — strongest NFT-holder objection.** “We supported Claus when houses, backing and fee participation were part of the story; you have already removed new rewards and now want our status to mean nothing in the new business.” This is a stronger objection than simply demanding a free seat. The existence of historical NFT rewards, retained accrued claims, houses and inference access is documented. Whether any additional promises were made to particular holders is not established by the sources I read. [C4], [C10], [C9] **P — honest answer.** Acknowledge that history and preserve existing entitlements; do not claim holders consented to a broader settlement merely because the 50,000-CLAUS payments occurred. A house need not buy permanent claims on a new checking service. Holders can qualify and work on the same terms as anyone else, and holders who received CLAUS already have the same token rights per unit as others. New service fees must first pay its workers and obligations. If old commitments conflict with this separation, founders must resolve and disclose that conflict before launch rather than conceal it under new token economics. **P.** Offer a permanent public thank-you and provenance/history page, preserve existing house identity/art, publish a clear guide to old claims and migration receipts, and invite holders to the same free public training as everyone else. A cosmetic “early supporter” label is acceptable if it gives **zero payments, zero job priority and zero extra votes**. Avoid holder-only fee discounts or grandfathered assignment rights: those recreate an economic preference even without a cash dividend. ## 7. Questions only the founders can settle **U / P.** These are authorization, commitment and institutional-design questions; the public-chain measurements are not questions being delegated back to founders. 1. **Claus:** Will you irreversibly remove discretionary minting and freeze collateral-token transfer economics and upgrade backdoors, including delegated operator paths? If not, what weaker, explicitly discretionary service are you willing to describe honestly? 2. **Claus:** Which entity undertakes the $100 customer-remedy obligation, supplies the initial $5,000 reserve, holds its dollar assets, and approves the ongoing cost budget? Which actual Robinhood settlement/reserve asset and reference implementation will it use? 3. **Claus and IMD:** Will IMD formally accept the eligible first-hearing work, provide conflict recusals at controller level, publish its real quote, and permit an outside decision to bind disputes over IMD-built contracts? Existing generic oracle capability is not that agreement. 4. **Both:** Which independently affiliated people or organizations will contract to be outside judges and appeal judges, sign the enforceable decisions, and accept the proposed deadlines and compensation? Who bears the loss when they cannot answer? 5. **Claus:** What precise historical NFT, inference, backing and accrued-reward commitments remain outstanding, and will you explicitly preserve them without adding a new service-income claim? A founder's binding commitment cannot be inferred from a migration payment. 6. **IMD:** Can a seat refuse launch allocations or otherwise avoid disqualifying token exposure, and will the network recognize shared controllers across multiple seats for these jobs? If not, which work will be excluded from the Claus partnership? ## 8. Five sentences for a CLAUS holder — I **P: proposed explanation, exactly five sentences.** Claus should pay people to check facts, with no new income for the 300 houses. Anyone can qualify by passing a practical trial and posting a returnable CLAUS bond, but an idle bond earns nothing. Two declared separate operators must publish evidence for the same facts, and a proved false statement can cost their job collateral after a fair challenge. Customer fees pay the workers, refunds, disputes and real operating costs before any genuine surplus burns. This only deserves trust if Claus locks the powers that could mint more collateral or change the rules for transferring it. ## Measurement appendix: reproduce, distinguish and bound the observations **R / I — access and integrity.** All network calls were public GETs or JSON-RPC read methods (`eth_chainId`, `eth_blockNumber`, `eth_getBlockByNumber`, `eth_getLogs`, `eth_getCode`, `eth_getStorageAt`, `eth_getTransactionReceipt`, `eth_call`). No transaction, login, private registry or authenticated account was used. Public read transport may use HTTP POST for JSON-RPC; that does not write chain state. Explorer failures were not treated as negative chain evidence. [Robinhood connection documentation][RH], [Ethereum provider][EP] **R — exact ownership/control reads.** At Robinhood block `0x6b49574`, `owner()` (`0x8da5cb5b`) on the token and then its returned administrator gave the addresses above. EIP-1967 slots queried were implementation `0x360894a13ba1a3210667c828492db98dca3e2076cc3735a920a3ca505d382bbc` and admin `0xb53127684a568b3173ae13b9f8a6016e243e63b6e8ee1178d6a717850b5d6103`. The latter returned **`0xcbc4bbd337e62c926c995e66a231e5e2b0a722a8`**; its owner is the administrator above. Administrator `operator()` returned **`0x8aca8890d34086d7a30b9c06fae25e0c3cda4cfa`** and `managedTarget(tokenProxyAdmin)` returned true. The ultimate owner wallet returned empty runtime code at the pin. [RPC-RH], [AS] **R.** I also compared RPC-fetched implementation and administrator runtime bytes at that pin with the respective Sourcify responses' `runtimeBytecode.onchainBytecode`; both were identical. Sourcify reports matched verification. This is source attribution and byte identity, not an audit or proof of future immutability. The deployment manifest is useful history but was not substituted for the live implementation slot. [TS], [AS], [C6] **I — CLAUS balance replay recipe.** Query the complete Transfer topic `0xddf252ad1be2c89b69c2b068fc378daa952ba7f163c4a11628f55a4df523b3ef` for the exact token from genesis to the pin; sort by block, transaction index and log index; subtract each amount from the sender and add it to the recipient, treating the all-zero event address only as mint/burn accounting. Remove zero balances, assert no negative holder balances, and compare the sum with `totalSupply()` (`0x18160ddd`). I additionally compared the largest ten balances with pinned `balanceOf(address)` (`0x70a08231`) calls. Split log ranges if a provider imposes limits. Contract code can invalidate naive Transfer accounting in other tokens; the sum and selected balance checks are important limits here. [RPC-RH], [TS] **I — NFT method.** For Claus, fetch all 300 `items` in `nfts.json`, verify unique IDs, then group separately by `owner` and `beneficiary`. The exact fetched JSON had SHA-256 **`116947eec1e839bdef6d4c871e18c21571a505ff0fc16a0ae85fb9dd606282b5`**. Its beneficiary mapping remains a project-supplied attribution even though the custody counts also matched the explorer. [N1], [N2] **I — IMD method.** At Ethereum block **26,398,619** (`0x192cf9b`), query `ownerOf(uint256)` (`0x6352211e`) for **every ID 0–1999** on the exact collection; aggregate the returned lowercased addresses. I used batches of 50 calls, checked all 2,000 results and recorded **zero failures**. SHA-256 of lines `decimalTokenId,lowercaseOwnerAddress\n`, in ascending token-ID order with a final newline, is **`72721b820fbafb89c178d6900944261069323d0b63e81ad4e34da2b8f2badeb5`**. This distinguishes the pinned census from the rejected 2,002-sum moving holder list. [ETHRPC], [N3], [M3] **I — repository method.** The read URL was `https://api.github.com/orgs/identity-md-launches/repos?per_page=100&sort=created&direction=desc&page=N`. Count each repository once by its public name, exclude forks, and use complete UTC dates in `[2026-11-05T00:00:00Z, 2026-11-12T00:00:00Z)`. No commit, README, review or issue text was needed for this count. [GHAPI] **R / I — Robinhood daily token proxy, completed coverage.** Timestamp-based binary searches set the inclusive start at block **111,281,136** and exclusive end at **112,102,884**. I fetched all zero-sender Transfer logs in chunks of at most 100,000 blocks, filtered to exactly three topics and one 32-byte amount, deduplicated emitting addresses, and queried code at block **111,281,135**. For every address with empty pre-day code I read `decimals()` (`0x313ce567`, a 32-byte integer no greater than 255) and `totalSupply()` at block **112,102,883**. Transient HTTP 429 errors were repaired with smaller batches; **no candidate/code/metadata read remained missing** in the final count. This does not test all ERC-20 behavior, token purpose, liquidity, customers or controller uniqueness. [RPC-RH] | R: boundary block | UTC timestamp | Hash | |---:|---|---| | 111,281,135 | 2026-11-10 23:59:59 | `0xdb94e98a2e7b4759d23ddf63bd82aa62211d337d48ebc14472529e5fdd30dbcc` | | 111,281,136 | 2026-11-11 00:00:00 | `0x2ebf9ca1dc2540126acb46d396af35eab6cfed5314d2cdbd3ed3f58c384388a5` | | 112,102,883 | 2026-11-11 23:59:59 | `0x6a363c9b81ed7689e9e48ed3f72a97162ef09fd253cb49171a9aa2689ad94185` | | 112,102,884 | 2026-11-12 00:00:00 | `0x4e51a648751ff0a72060690b53f55b9dbe05edbb9597ffa779c5d93befd9c01d` | **I.** The resulting counts are **381,029 mint events → 4,862 distinct shaped emitters → 4,256 absent-before-day addresses → 4,256 passing metadata reads**. SHA-256 of the sorted lowercase passing addresses, one per line with final newline, is **`0f831ecc1e244585a41064e40238d55dfa91b318771613a384a3a34f36306ed4`**. Example passing addresses and their first observed mint block *within this window* are below; these are not claims of legitimate projects or official explorer classifications. [RPC-RH] | R / I: example address | First window mint block | |---|---:| | `0x00325c5fb31ba99adc95bc464e7a216ae1ab2ff9` | 111,727,307 | | `0x0046907841fc52e407f17dfb293962ba86e67492` | 111,824,089 | | `0x0050079dba41313c726f502e2d7b9efbba863d1e` | 111,998,781 | | `0x00545ee84e3234f5b6460669349e860205fc7777` | 111,475,428 | | `0x00647f77f2287794b41b0427d3c095864ad3adbe` | 111,580,032 | **U.** The requested explorer-derived new-token rate remains unavailable even though this RPC proxy is complete for its own definition. Absence of code before the window does not prove first-ever historical existence, and requiring a mint event and optional metadata excludes other real tokens. A future explorer measurement should enumerate creation receipts and internal creations, deduplicate by token address, pin both daily boundaries and the pre-window comparison block, distinguish ERC-20/ERC-721/other standards and separately classify official stock tokens, community launches and spam. It must not count transfers or repeat mints as births. No token-like contract count is substituted for paid demand. ### Source register **R.** Sources were retrieved on 12 November 2026 unless a response timestamp is specified. “Matched” below describes the verification service's compiler-match result; it is not an independent security opinion. Links to inaccessible explorers identify attempted sources and are not offered as proof of facts that could not be retrieved. Chain values are attributed to their actual RPC/provider and exact block above. - **C1–C6:** [migration guide][C1]; [snapshot distribution status][C2]; [NFT payment transaction, explorer access limited][C3]; [current Hooks description][C4]; [read-only migration operations][C5]; [deployment source manifest][C6]. - **C7–C11:** [public build guide][C7]; [legacy chain-1 hook health][C8]; [public chain-control/configuration disclosure][C9]; [NFT inference and house rights][C10]; [published deployment source archive][C11]. - **D1–D3:** [Ethereum CLAUS deal terms][D1]; [Ethereum ETH deal terms][D2]; [public capabilities and arbitration limits][D3]. These endpoints are explicitly identified by the public build guide; none requires login. - **M1–M3 / E1:** [IMD participant/token-allocation docs][M1]; [IMD oracle API][M2]; [IMD collection explorer][M3]; [ERC-8004 specification][E1]. - **G1–G2 / U1:** [The Graph tokenomics][G1]; [The Graph service-eligibility explanation][G2]; [UMA mechanism documentation][U1]. These distinguish documented mechanisms and project-reported historical observations from my design inferences. - **K1–K5:** [Doges on Trial observations][K1]; [bribery/no-show analysis][K2]; [lazy/ambiguous voting analysis][K3]; [Curate incentives and challenger's dilemma][K4]; [Kleros's account of V1 limitations][K5]. - **Chain observations:** [Robinhood official connection/provider docs][RH]; [Robinhood public Pinax RPC][RPC-RH]; [Ethereum Pinax provider description][EP]; [Ethereum public Pinax RPC][ETHRPC]; [current matched token source][TS]; [matched administrator source][AS]; [Claus NFT dataset][N1]; [Claus NFT explorer holders][N2]; [IMD explorer holder endpoint][N3]. - **Activity and access limits:** [public launch organization][GH]; [repository metadata API][GHAPI]; [Claus project offers][DIR]; [Blockscout token list][RL]; [Robinhood CLAUS explorer][RT]; [Robin Etherscan list][RE]; [earlier proposal, inaccessible][OLD]. [C1]: https://claus.si/docs/migration [C2]: https://claus.si/api/migration/snapshot [C3]: https://robinhoodchain.blockscout.com/tx/0xbd404685574a53e58e2ebd5b5887486a3330a7b3d32121b4cc8364e6761c9046 [C4]: https://claus.si/Hooks [C5]: https://claus.si/api/migration/status [C6]: https://claus.si/contracts/robinhood/manifest.json [C7]: https://claus.si/build.txt [C8]: https://claus.si/hook-health.json [C9]: https://claus.si/about.json [C10]: https://claus.si/docs/nft-inference [C11]: https://claus.si/contracts/robinhood/sources.zip [D1]: https://claus.si/__network/v1/public/chains/1/deals?asset=CLAUS [D2]: https://claus.si/__network/v1/public/chains/1/deals [D3]: https://claus.si/__network/v1/public/capabilities [M1]: https://imd.fun/docs/ [M2]: https://imd.fun/docs/api/#ask-the-oracle [M3]: https://etherscan.io/token/0x0000ec93127baa929e58e97dd0095a2bfb38ec1d [E1]: https://eips.ethereum.org/EIPS/eip-8004 [G1]: https://thegraph.com/docs/en/resources/tokenomics/ [G2]: https://thegraph.com/blog/rewards-eligibility-oracle/ [U1]: https://docs.uma.xyz/protocol-overview/how-does-umas-oracle-work [K1]: https://blog.kleros.io/observations-from-our-decentralized-curated-list-pilot/ [K2]: https://blog.kleros.io/cryptoeconomic-deep-dive-doges-on-trial/ [K3]: https://blog.kleros.io/doges-on-trial-pt3-cryptoeconomics/ [K4]: https://blog.kleros.io/choosing-parameters-in-kleros-curate/ [K5]: https://blog.kleros.io/the-kleros-vision/ [RH]: https://docs.robinhood.com/chain/connecting/ [RPC-RH]: https://robinhood.rpc.pinax.network [EP]: https://pinax.network/chain/eth [ETHRPC]: https://eth.rpc.pinax.network [TS]: https://sourcify.dev/server/v2/contract/4663/0xd68727ddd50e8ba07e4be20d37a5ecf3dd814bff?fields=all [AS]: https://sourcify.dev/server/v2/contract/4663/0x979d5810d32292358639f25294472f01c95bdc46?fields=all [N1]: https://claus.si/nfts.json [N2]: https://eth.blockscout.com/api/v2/tokens/0x80396c7131159eb92e7e84839e51c9887a7a95d9/holders [N3]: https://eth.blockscout.com/api/v2/tokens/0x0000ec93127baa929e58e97dd0095a2bfb38ec1d/holders [GH]: https://github.com/identity-md-launches [GHAPI]: https://api.github.com/orgs/identity-md-launches/repos?per_page=100&sort=created&direction=desc [DIR]: https://claus.si/read/network [RL]: https://robinhoodchain.blockscout.com/tokens [RT]: https://robinhoodchain.blockscout.com/token/0x163015276f3244bC2CcB78b4D69cc8A4069EfFAC [RE]: https://robin.etherscan.io/tokens [OLD]: https://swarmcard.imd.site/