1. Question restated

Is Fee to Jobs worth building and launching within 14 days on IMD or CLAUS, and what is the smallest version that would prove it?

Evidence convention — read 2026-10-23 UTC: Record means a directly observed API response, repository object, verified source or chain response; it does not certify the operator's truthfulness. Claim means a source's assertion that was not independently demonstrated. Inference includes this report's calculations, recommendations and proposed parameters. A label applies to all figures in its paragraph or table row unless a column says otherwise. All observations below were read on 2026-10-23; historical event and commit dates are stated separately. Proposed thresholds are experimental choices, not observed market facts. No contracts were deployed, transactions broadcast, or downloaded programs executed for this research.

2. Verdict in five lines

1. Inference — gap: partly served. Do not launch a new Fee to Jobs token as specified: OCTO already implements most of the mechanism in source, and live Ouroboros Press proves autonomous commissioning and receipts.
2. Inference — speed to MVP: a validation prototype fits 14 days, but a launch satisfying all seven requirements is not established; IMD is the better experimental home and CLAUS currently fails the required admin delay.
3. Inference — MVP in one line: test an immutable hook/vault on a fork, then buy two holder-approved jobs to test useful output and acceptance; require a complete fee→Intake→artifact→epoch-close trace before considering a token launch.
4. Inference — biggest failure: no recurring buyers: 39 of 44 sampled IMD-paired launches had zero swaps in the sampled day; Intake also charges before work, and a delivery callback does not prove semantic acceptance.
5. Inference — total IMD cost: 2.0 IMD for validation, 4.5 IMD including a conditional launch, two new fee-funded operating jobs and two oracle checks; including this 0.5 IMD report gives 2.5 / 5.0 IMD, respectively, excluding gas and bridge fees.

3. Competitor table

Record / Claim — read 2026-10-23: “Live” below means an observed deployment or live service, not a security or revenue endorsement. “Unreleased” for OCTO means the inspected release was source-only: deployment not found in the searched records. It is not proof that no deployment exists. Repository last-activity dates come from GitHub's latest commit, not a README's timestamp.

Name and link	Chain	Status	Last activity date and pin	Overlap	What it misses / why it matters
Ouroboros Press	Robinhood Chain, 4663	Live, corroborated by Intake completion and deployed code	2026-10-21 22:26:00 UTC, c85771813781217b428208cf5d28053194c65f83; deployed build separately pinned in §13	Ownerless contract orders chapters through Intake and records result hashes/URIs; observed chapter count 2	Donation-funded, fixed chapter brief; no swap hook or holder selection. Its delivery bounty is for the caller, not evidence of a worker's per-job wage. Launch record, completion
OCTO	Ethereum, 1, intended	Unreleased in the inspected evidence; live addresses not found	2026-10-15 12:51:04 UTC, 3c5d13b8bad5941a2de90079069016c0cd1e9cf4	Source takes 2% on the IMD leg: 1.5% retirement sink, 0.5% DAO; weekly stake snapshots choose commissions, receipts are emitted, excess treasury is retired	Very close prior art. Source has immediate owner curation/settings, exact-output swaps unsupported, and a reserve rather than an unconditional epoch sweep. Empty deployed-address config prevents calling this a verified live business. Hook, DAO
SIMD Vault	Robinhood Chain, 4663	Live contract, launch 2363; functioning fee cycle not established	2026-10-21 10:30:29 UTC, 01d4f6c81bca5f7cc8747edc33e8bdd71b934c57	Source splits creator fees between holder rewards, jobs and other recipients; holders can request work	Spot-balance eligibility and per-wallet cooldown permit circulating one bag through wallets. No collective approval of exact briefs; no job callback in JobsVault. Deployment alone does not prove fee redirection/binding succeeded. Deployment
KEPT	Robinhood Chain, 4663	Live, hook launch 2185, vault launch 2172	Hook 2026-10-18 21:29:54 UTC, b9001b9683f2fc0ce72351127c9cc68045302d66; vault 2026-10-18 16:25:19 UTC, bbb7f7d193af94838480bb6898c6d78dcc9b7f51	Swap-funded IMD vault subsidizes semantic oracle checks of milestones. Strong evidence that an IMD hook can fund a service vault	Milestone adjudication rather than new work procurement. README discloses trivial-pledge subsidy extraction and a vault protocol-change power with seven-day notice. Does not prove demand for holder-directed production. Hook launch, vault launch
CLAUS programmable fees	Ethereum, 1	Live	Fee snapshot 2026-10-23 19:42:09.034 UTC, block 26257487; governance deployed at block 26233391	Existing hook can allocate fee claims to replaceable programs; compute already receives a share	Core permanence does not freeze the module. Current governance permits immediate replacement; ETH receipts require conversion into IMD; no holder-approved Intake-only spending rule. Verified host
CLAUS Jobs	Ethereum, 1, for paid agreements	Live service; paid-settlement behavior here remains a documented claim	API capabilities read 2026-10-23; project inference documented disabled 2026-10-21	Documented ETH/CLAUS escrow separates delivery, requester acceptance, credit and withdrawal	Payment can be acceptance-gated here, but this is a different rail from IMD Intake. A requester still accepts; it does not supply an autonomous holder-controlled budget. Capabilities

Inference — read 2026-10-23: The defensible gap is budget discipline plus independently assessed usefulness: exact brief approval, bounded prepaid orders, attributable acceptance decisions and transparent destruction of leftovers. “A token hires agents and publishes receipts” is already partly served. Copying OCTO with different percentages is not sufficient differentiation.

4. Hook spec: callbacks and permission flags, hookData schema, fee math with one worked example, invariants, admin surface

Inference/design — read 2026-10-23: The following is a fork prototype specification, not an approved production launch. Use one new fixed-supply, plain-transfer token paired with IMD, an immutable FeeJobsHook, a JobsVault, a voluntary staking/voting contract and a separate HolderRewards accumulator. No token transfer tax, blacklist, pause, venue check or post-launch mint. Other pools and venues remain possible; therefore the funding promise applies only to swaps through the configured hooked pool. Never advertise a fee on every transfer or every venue.

Record — read 2026-10-23: v4 selects callbacks from address permission bits. The hook's sender is the immediate PoolManager caller, usually a router; it is not authenticated trader identity. This follows the callback dispatch in the pinned core source. Uniswap hook documentation, PoolManager source, permission-bit source.

Inference/design callback or flag — read 2026-10-23	Permission	Rule
beforeInitialize	0x2000	PoolManager-only callback; require factory as supplied sender; bind once to the expected token/IMD pair, approved static pool fee and tick spacing
afterInitialize	0x1000	Save initialization block and announced opening policy; no privileged first buyer
beforeSwap	0x0080	Validate bound pool; when IMD is the specified currency, return the specified-side fee delta
afterSwap	0x0040	Charge the IMD leg when it is unspecified; otherwise verify the precharged full-fill amount; accrue claims exactly once
beforeSwapReturnDelta	0x0008	Enables the specified-side adjustment
afterSwapReturnDelta	0x0004	Enables the unspecified-side adjustment
Total	0x30cc	0x2000 + 0x1000 + 0x80 + 0x40 + 0x08 + 0x04

Inference/design — read 2026-10-23: Every other permission is false: before/after add liquidity, before/after remove liquidity, before/after donate, after-add return delta and after-remove return delta. unlockCallback(bytes) is additionally used for a permissionless claim sweep; it is an IUnlockCallback, not a hook permission bit. Authenticate PoolManager and a locally initiated, single-use sweep nonce; redeem only recorded fee claims to the fixed vaults. The proposed hook has no deployed address: not found. CREATE2 permission mining is future implementation work, not an address guessed here.

Inference/design — read 2026-10-23: hookData is empty for ordinary swaps; optionally accept abi.encode(uint8 version, bytes32 referralTag) with version = 1, exact canonical length and a documented size bound. The tag is analytics only: no discount, voting weight, fee destination or trader limit depends on it. Do not use tx.origin, an unsigned address in hookData or a router-level counter to claim one-person limits. Trader identification is deliberately unnecessary. Adding identity-dependent rules would require a signed order binding payer, beneficiary, pool, amounts, nonce, deadline, chain and consumer, with equivalent access for contract wallets; that is excluded from this prototype.

Inference/design — read 2026-10-23: Proposed steady fee: 25 basis points of the gross IMD leg to JobsVault and 5 basis points to HolderRewards, totaling 30 basis points. These are hypotheses chosen to test a low incremental fee, not market estimates. For gross IMD g, T = floor(g × 30 / 10,000), J = floor(g × 25 / 10,000), H = T − J. On a fully filled exact-input IMD buy, the pool receives g − T; on an IMD-output sell, the buyer receives g − T. If IMD input is unspecified and the core requires x, gross it up as g = ceil(x × 10,000 / 9,970) and charge g − x. For specified net IMD output n, use g = ceil(n × 10,000 / 9,970). Split actual collected units, with rounding explicitly tested rather than multiplying display decimals.

Inference/worked example — read 2026-10-23: A full-fill buy with 100 IMD gross pays 0.25 IMD to jobs and 0.05 IMD to holders; 99.70 IMD enters core swap accounting. Assuming the observed 1.25% static LP fee and no additional protocol fee, the core LP charge is 99.70 × 0.0125 = 1.24625 IMD; 99.70 − 1.24625 = 98.45375 IMD reaches price-moving swap math before price impact. Effective combined fee is 0.30 + 1.24625 = 1.54625 IMD, not an unexplained “0.3% total.” A sell pays its core LP fee in the input token, so it cannot be calculated as an additional IMD debit from this example. The existing factory split is separate from the new holder accumulator. Verified PoolFees implementation.

Inference/design — read 2026-10-23: Precharging a specified-side fee from the requested amount can overcharge a partial fill. The smallest safe prototype must reject a successful partial fill of that path, rather than claim to charge a fixed fraction of actual volume while retaining the whole precharge. Exact-input/output, both currency orderings and non-default price limits need real PoolManager tests. Production admission depends on router compatibility and an acceptable full-fill policy; it is not established by the arithmetic above.

Inference/design — read 2026-10-23: Invariants: fee claims equal jobs plus holder liabilities; no swap invokes Intake or an AI service; failed external work cannot revert trading; holder rewards cannot spend job funds; JobsVault may approve only the pinned Intake for the exact current approved order price, capped at the approved budget; raw request bytes must hash to the voted brief; one request per authorized budget slot; duplicate and late callbacks cannot pay or reopen a slot; no arbitrary call, delegatecall, owner withdrawal, upgrade, fee setter or destination setter. Voluntary staking is separate from token transferability. With no stakers, holder fees stay reserved for a future eligible staking epoch rather than being silently redirected to the creator.

Inference/design — read 2026-10-23: Attribute job claims to their earning epoch inside the hook, before any sweep. Closing an epoch must redeem or irrevocably segregate its entire unspent allocation, including unswept claims; delaying a sweep cannot carry old fees into a new budget. Proposals target a named future operating epoch after voting and the execution delay. Already prepaid jobs remain attributable to their original epoch when delivery arrives later; their receipts cannot reopen that epoch's spending. No arbitrary withdrawal is disguised as a keeper incentive: the pilot sponsor supplies execution gas separately.

Inference/design — read 2026-10-23: Use a 24-hour spending epoch and at most two jobs per epoch for the test. A proposal commits raw JSON, its hash, price ceiling, acceptance rubric and expiry. Voting uses stake held before the epoch, held through its vote, with one allocated weight per stake unit; moving liquid tokens between wallets cannot multiply that weight. Proposed quorum is 10% of eligible snapshotted stake, majority approval, plus a 24-hour public execution delay. These settings are experimental and not a decentralization claim: a majority stake can still approve useless briefs. No founder key or emergency withdrawal exists. Pinning Intake and signer sacrifices upgrade liveness; a service retirement halts new orders instead of empowering an admin to redirect money.

Record / Inference — read 2026-10-23: Burn is not resolved by writing a sink address. Verified Ethereum IMD is an OFT without a public burn(uint256) entrypoint. POOL4 documents retirement through a bridge to a Base burn receiver. Its current BurnExecutor is owner-configurable and has token rescue, so transferring leftovers there is not equivalent to an ownerless, completed burn. The strict prototype must distinguish epochClosed, retirementQueued and burnConfirmed; no rollover or new job may spend closed-epoch funds. A fixed direct OFT route to a verified immutable receiver is a candidate, but delivery latency, shared-decimal dust and native bridge fees mean it does not prove destruction precisely at epoch end. If the requirement literally demands that timing, the IMD-denominated version is a no. A buyback/burn of the new token would be a changed design, not a silent substitute. Ethereum IMD verified source, POOL4 retirement documentation, BurnExecutor verified source.

Inference — fair-opening assessment, read 2026-10-23: Publish code, allocation, initialization transaction policy and start time before opening; give no privileged creator fill, fee exemption or private allowlist. Charge every bundled leg equally, independent of router and wallet count. Snapshot voting before the relevant epoch so a flash acquisition cannot seize that epoch's budget. These rules do not remove first-position advantage or stop a builder bundling buys. A time-decaying surcharge can price that advantage, but is not a fair-clearing proof and could suppress volume. This report therefore does not mark hard requirement 5 as passed. A publicly reproducible opening auction or similarly defensible mechanism, compatible with IMD's factory and unrestricted token venues, is a launch gate rather than an invented guarantee.

5. Agent spec: who signs what, any pinned oracle question, how work is verified before anyone is paid, how agents are paid, failure modes

Inference/design — read 2026-10-23: Agents should perform evidence synthesis, adversarial review and creation of a new useful artifact from an approved brief. Plain code can enforce budgets, count links, hash files and transfer tokens; none of those activities justifies paying an AI. The useful distinction is adaptive interpretation of unfamiliar evidence, not an assertion that AI is literally non-computable. Work is performed offchain and changes onchain state through authenticated delivery/acceptance. If requirement 1 instead means AI inference executed inside the EVM, this design is a no.

Record — read 2026-10-23: The capabilities response records 0.5 IMD for job, launch and oracle requests; schedules have minimum intervals of 10 minutes for oracles and 30 minutes for jobs. It lists custom-hook support on Ethereum 1 and Robinhood Chain 4663. Both Intakes returned job price 500000000000000000 and callback gas 200000 at the pinned blocks. The swarm response recorded 1,025 agents online, 69 jobs done and 17 oracles done in its reported last-day counters, with verifier/publisher/deployer up. These are operator telemetry, not independently proven worker quality or capacity guarantees. Capabilities, swarm.

Record — read 2026-10-23: Payment cannot wait until acceptance on present Intake. In verified Intake._request, _collect transfers the payment to payTo in the ordering transaction. complete later records a result and attempts the named callback. There is no escrow/refund mechanism in that contract. An unsigned failure callback clears application state; it does not reverse the payment. Hence “verify work before anyone is paid” cannot literally hold for the IMD treasury. Reverting a delivery callback does not claw the fee back. Verified Intake.

Inference/design — read 2026-10-23: The honest state machine is approved → prepaid → delivered → independently assessed → accepted/rejected. Check the brief offchain before ordering; a relayer can run the free request checker, but the vault cannot itself read HTTP and a checker response is not a quality proof. Deterministic artifact checks verify bytes, manifests, requested paths and meaningful tests. A separate reviewer assesses whether the result fulfills the pinned rubric. Rejection stops follow-on orders; it does not recover the sunk order price. No direct worker bonus is included: paying one outside Intake would violate this idea's Intake-only job spending rule.

Inference/design — read 2026-10-23: Under the strict job-only spending rule, JobsVault must not buy oracle.request actions either. The optional acceptance checks below are separately paid by the experiment's sponsor through an assessment consumer, with no access to JobsVault funds. Allowing job fees to buy those checks would require an explicit change to the spending policy. Thus independent signed acceptance has an external funding dependency in this prototype; it is not advertised as a closed, self-financing loop.

Record / Inference — read 2026-10-23	Signer or caller	What is established
Brief approval — design	Holders vote with snapshotted stake; signed votes, if offered, use the voting contract's EIP-712 domain and nonces	Approval of the exact bytes and bounded budget, not their truth
Ordering — design	Any caller executes an eligible approved proposal; JobsVault calls Intake	Vault is the payer; caller cannot change body, beneficiary or price ceiling
Delivery — record	IMD Intake's authorized writer calls complete; vault accepts callbacks only from its own Intake and its own unsettled request IDs	Hash/URI provenance under the writer's authority; no separate semantic signature on a normal job result
Acceptance — design	Independently commissioned IMD oracle panel; one attester signature	A signed answer to an exact question; depends on panel honesty and the attester key
Worker compensation — record / limitation	IMD's treasury/reward system, not the vault	Upfront IMD payment to treasury is proven. Launch contributor token entitlements exist; a per-report, acceptance-gated onchain wage to the executing worker was not found

Record — read 2026-10-23: Oracle attestations have EIP-712 domain IdentityMD Oracle, version 2, chain and consumer contract. A public bool attestation for request 6faa93d8-edc2-4a44-be4e-fbeafdf43fa8 was cryptographically checked with cast wallet verify: signature validation succeeded for the API's recorded attester. Its API message uses the label bool while the typed field is uint8; verification required the documented normalization bool = 0. This verifies the signature, not the answer. Do not copy that request's consumer into a new deployment. Attestation record, typed-data and enum documentation.

Inference/design — read 2026-10-23: Pin the acceptance question as follows, instantiating each value from actual onchain state, never with made-up IDs:

Did the artifact whose manifest hash is recorded for this vault's stated Intake request fulfill every mandatory criterion in the exact holder-approved brief and acceptance rubric, using only the committed evidence snapshot and the stated block window?

Inference/design — read 2026-10-23: The question document must contain vault address, chain, epoch, proposal hash, Intake request ID, result hash, rubric hash, evidence manifest and absolute block window. Request bool only, panel size 7, quorum 5, accept only a signed yes with agreed ≥ 5. Missing or inaccessible mandatory evidence means no; insufficient consensus means unresolved, never automatic acceptance. Store the canonical question hash before asking. Verify domain, consumer, answer type, window/block hash, issuance/expiry, minimum panel strength and unused request binding. Intake request IDs and oracle UUID IDs are different: the writer pairs them, so bind the exact Intake ID and artifact hash into the question. The signature alone does not remove this pairing risk.

Record / Inference — read 2026-10-23: ERC-8004's official deployment list names IdentityRegistry 0x8004A169FB4a3325136EB29fA0ceB6D2e539a432 and ReputationRegistry 0x8004BAa17C55a88189AE136b182e5fdA19dE9b63 on both target chains; code was observed at each address. An official ValidationRegistry deployment was not found in that list. Registry identity and reputation are discovery signals, not proof of useful work, independent operators or payment. Launch reward snapshots show accepted-work contributors, but do not establish the requested per-job wage. Therefore hard requirement 1 remains unproven for non-launch jobs and is a launch blocker. Official registry repository, example launch reward record.

Inference — failure modes, read 2026-10-23: Cited but low-value work; colluding holders and reviewers; malicious instructions inside briefs/artifacts; false writer delivery; signer compromise; unavailable IPFS content; stale attestations; callback gas exhaustion; permanently missing callbacks; price changes; insufficient funds; duplicate execution; keeper absence; governance capture; fake volume. Hashes protect integrity, not quality. Bound loss by approved prepaid budgets, make callbacks store only receipts, preserve late-result provenance without reopening spending, and never interpret timeout or malformed evidence as approval. Zero knowledge is omitted: it would not establish whether a report is useful or remove the panel/writer trust boundary here.

6. Economics: who pays on day one and why, unit costs including 0.5 IMD per job, break even

Inference — read 2026-10-23: Day-one payer should be a named project community that already wants a report or site, funding the pilot from its own approved budget. Holders buy a useful collective artifact; traders supply fee revenue only if they already want the token. A speculative claim that fees create demand creates no customer. Do not subsidize wash volume to demonstrate “self-funding.” The validation experiment needs 1.0 IMD for its two operating jobs even if no trades occur; its sponsor must knowingly pay that amount. No sponsor commitment or willingness-to-pay interview was obtained in this research.

Record — method, read 2026-10-23: Select all 44 IMD-paired live token pools identified in the latest 100 launch records returned by /launches, spanning launch numbers 2304–2512 in that response. This is a recent-launch convenience cohort, not the whole network or a random sample. Both chains were sampled; 32 pools are on Ethereum and 12 on Robinhood Chain. Query canonical PoolManager Swap events filtered by the derived PoolIds; sum absolute IMD deltas once per swap, without counting both legs as volume. The records contain 109 swaps across 5 pools. Newer launches have less than a full day's lifetime inside the window, and zero events mean zero observed activity in that pool, not zero activity at every venue. Selection, raw logs and derivations are retained under evidence/.

Record — time pins, read 2026-10-23: Ethereum: blocks 26250313–26257482, ending 2026-10-23 19:40:59 UTC, end hash 0x9217c1d80cf913f5ae8c019bcfc77b9adf05bfd15c35be2e02f4f6202aeed7e7. Robinhood Chain: 95361406–96189684, ending 2026-10-23 19:41:07 UTC, end hash 0xc64bcccb237d3a1249031dca1424a5b0c49dca03857f9e832ecc2569f01885de. Start blocks are the first at or after each end timestamp minus 86,400 seconds. Public RPC reference endpoints are Ethereum PublicNode and Robinhood Chain RPC; these figures are read-only chain observations made through the assignment's supplied transport. Launch identities and commits come from IMD's launch records, corroborated by pool events.

Record — pool, chain, swaps, read 2026-10-23	Record: IMD-leg volume V	Record-derived: seeded-position IMD LP fees accrued	Inference: requester entitlement at 80% of that IMD fee	Inference: hypothetical jobs fee 0.0025 × V	Inference: whole prepaid jobs floor(fee / 0.5)
IMDirector, launch 2418, Robinhood, 56	1638.959132913334419	10.532713129081056	8.426170503264845	4.097397832283336	8
YOLO, launch 2364, Ethereum, 34	398.326871795196239	4.133030537655315	3.306424430124252	0.995817179487991	1
Hamster Hour, launch 2336, Ethereum, 13	107.287464950266794	1.208015956601889	0.966412765281511	0.268218662375667	0
INK, launch 2321, Ethereum, 5	10.572298417747658	0	0	0.026430746044369	0
Dongle, launch 2474, Ethereum, 1	0.341898505071183	0	0	0.000854746262678	0
Remaining 39 pools, both chains	0	Not individually simulated; zero swap-derived accrual in the window	0 swap-derived IMD	0	0

Record / Inference — read 2026-10-23: Display figures are rounded; whole-job counts use raw units. “Accrued” was calculated by non-broadcast eth_call simulations of claimFees(launchNumber) at the block before the window and at the end, adding actual FeesClaimed events inside it: pendingEnd − pendingStart + collectedDuring. For these active positions there were zero such collection events in the window. This measures accrued fees, not a wallet payout. Requester columns apply the verified factory's 80% share; actual rounding occurs at collection and can differ by base units. Other-token fees are excluded, as are donations, historic balances and valuation gains. INK and Dongle traded only with IMD output during the window, so their LP fee accrued in the other currency. Multiplying their two-way IMD volume by a creator percentage would misstate spendable IMD. Factory source, IMDirector launch, YOLO launch, Hamster launch.

Inference — read 2026-10-23: Across this cohort, ΣV = 2155.487666581616291429 IMD; a hypothetical jobs fee at the proposed rate gives ΣV × 0.0025 = 5.3887191664540407285725 IMD before per-swap rounding. It is not a fee stream already payable to this product. The busiest pool accounts for 1638.959132913334418730 / 2155.487666581616291429 ≈ 76.03% of volume. The median pool funds zero jobs. Most importantly, epoch-end burning forbids pooling small leftovers across epochs: a pool producing 0.2682 IMD per epoch buys zero jobs forever at that unchanged rate, despite a positive theoretical annualized revenue figure. No annualized projection is justified from this window.

Inference — break even, read 2026-10-23: For n jobs per epoch and jobs fee rate f = 0.0025, minimum qualifying volume is V ≥ 0.5n / f = 200n IMD. One job needs 200 IMD; two jobs need 400 IMD. Adding one paid acceptance oracle per job doubles system resource cost: (0.5 + 0.5) / 0.0025 = 400 IMD of equivalent volume per assessed job, or 800 IMD for two. Those latter figures are cost comparisons, not authority for JobsVault to buy oracles: the strict design still requires 1.0 IMD of external review sponsorship for two checks. These thresholds assume full collection, no price changes and the same volume after adding fees. The table's two-way pool deltas are a proxy for the proposed hook's taxable gross IMD base; implementing its fee changes deltas and trader behavior, so the hypothetical column is a replay estimate, not an exact counterfactual.

Inference — costs and incentives, read 2026-10-23: Per job: 0.5 IMD prepaid, optional independent oracle 0.5 IMD, plus unpriced gas for order/receipt/acceptance, keeper execution and any epoch burn bridge fee. No dollar conversion is used. The proposed holder fee provides an explicit benefit while work improves a shared product. It also creates staking concentration and value-extraction risks. Existing LP fees largely go to requester/network under the factory arrangement, so do not describe them as automatically enriching token holders. Fee elasticity and genuine holder demand are unmeasured; hard requirement 6 is not demonstrated.

7. Home fit: what IMD or CLAUS must provide or approve, and what happens if they say no

Record — IMD policy, read 2026-10-23: IMD documents plain-transfer custom tokens, a hook-level buy/sell fee, fixed minting and a 10% swarm allocation, with 2% for accepted launch contributors and 8% for seats connected at admission. Current launch pools use a 1.25% pool fee; a new hook fee is additional. A deployed KEPT hook routes IMD claims to its vault, and OCTO's source describes a treasury route. Thus vault-directed hook fees are allowed in principle, not a categorical factory prohibition. Exact bytecode, pool terms and the final brief still require admission; this report has no approval for FeeJobsHook. IMD launch documentation, KEPT deployment and source pin.

Record — verified address directory, read 2026-10-23: Intake on both chains is 0xa43e6f75ee006411f79ac1c84120606c2330de82, with code and price getter observed on both and verified Solidity on Ethereum. Ethereum IMD is 0xd34a99bc0f67ae1bbd63c660e6d0b0dd03e263b7; Robinhood IMD is 0x5f7bb59365ce557c26dbcaa4ee9d39a4b95b7127, corroborated by capabilities, Intake calls and transfer/pool events. Ethereum PoolManager is 0x000000000004444c5dc75cb358380d2e3de08a90; Robinhood PoolManager is 0x8366a39cc670b4001a1121b8f6a443a643e40951, corroborated by factory receipts. An independently matched Robinhood Intake source deployment was not found in Sourcify; do not upgrade “officially listed plus live getters” into a claim of a compiler match. Capabilities/address listing, Ethereum Intake verification.

Inference — IMD requirements, read 2026-10-23: Obtain admission for the exact pool fee, callbacks, no-admin token/vault, holder fee split and opening mechanism. Establish a real worker payment trail for non-launch work and obtain explicit agreement on prepaid commissioning versus acceptance escrow, plus what constitutes an epoch burn. No signed permission from IMD was obtained here. If Intake prices change or service retires, the immutable version must stop ordering; holders can choose a separately reviewed replacement product without an admin moving the old vault's funds. If IMD refuses the fee destination or fair opening, cancel the token launch. Retain the offline reference and demand findings; do not disguise a transfer tax or migration restriction as a workaround.

Record — CLAUS, read 2026-10-23: Hook 0x37Bfb8AC7C960E558657871D41Ca70E07e7DbfFf points to verified implementation 0xa62f9b597225779b430024398b70374e8468ce14. Its ProxyAdmin ownership was renounced at block 26197788, 2026-10-15 12:03:11 UTC, transaction 0x9e1edba406365bf9ad470ce7868920b5cb69966ca7bdacaa6778ad58ee2c20a8; current owner() is null. The verified TradeExtensionHost 0xea0436465676eA7001535Bfe2E5cbd4b137b7f91 nevertheless permits the token owner to replace modules. The current token owner is the verified governance contract 0x180eae784dee739fbab432b06f78bf9a98aae334, not an EOA. Its operator can call setModule immediately; that operator returned no code at the pinned block, so an external timelock does not protect the current operator path. Founder powers can also replace the operator or initiate return of token control. There is no 24-hour queue in these paths. Freezing the proxy did not freeze fee-program governance. Hook verification, ProxyAdmin verification, current governance verification.

Record — CLAUS fees, read 2026-10-23: Host MAX_RATE = 149000 pips means 14.9%, a cap, not a recommended fee. The sampled live fee plan totals 23000 / 1000000 = 2.3% of gross ETH, including compute 1000 / 1000000 = 0.1%. Compute available ETH was 369233351306046421 wei = 0.369233351306046421 ETH, a balance rather than daily revenue; automatic provider payments were false. Current adapters preserve an owner allocation and compute share before the inner program. A design cannot assume all those fees are freely redirectable. Live fee plan, verified governance/adapter source.

Inference — CLAUS requirements, read 2026-10-23: CLAUS would need to approve installation and allocations, put every reachable module-changing/recovery authority behind a verifiable delay of at least 24 hours or irrevocably disable it, and provide an ETH→IMD route with bounded slippage. A timelock only inside our module is insufficient while the host can replace it. This is a larger dependency than a fresh immutable IMD hook. If CLAUS declines, do not launch there. CLAUS Jobs' acceptance escrow is an interesting alternative procurement rail, but using it changes the Intake-only idea. Project inference being disabled also means a public project is not an inference subsidy. Jobs agreement rules, live capabilities.

Inference — hard-requirement audit, read 2026-10-23	Result for launch today
1. Meaningful AI work with onchain agent pay	Not proven: adaptive work is plausible; a trace from each non-launch job to worker compensation was not verified
2. Rules in a v4 hook, named callbacks	Specifiable: callback design given; no implementation/admission proof yet; spending enforcement also needs the immutable vault
3. Plain transfers, no venue locks, no later mint	Specifiable: enforce with token invariants; fees apply only to the canonical hooked pool
4. No admin or at least 24-hour delay	IMD design: no product admin. CLAUS today: fail. External IMD writer/pricing/signer control remains a disclosed service trust assumption
5. Fair opening, snipers/bundles handled	Not proven: equal treatment alone does not remove bundle ordering advantage
6. Holder/liquidity benefit without killing volume	Not proven: explicit holder allocation proposed; demand and elasticity untested
7. ZK only when removing trust	Pass by omission: none needed for this experiment

Inference — read 2026-10-23: Because all seven are mandatory, “specifiable” is not a pass and unknowns are not quietly treated as passes. The present launch verdict is no. Intake's external owner can change writer/prices; the proposed no-admin claim applies to the new product contracts, not to the entire IMD service. If the requirement forbids even those external administrative dependencies, both the IMD and CLAUS variants fail more broadly.

8. Why now: what changed in the last 60 days that makes this buildable, with dated links

Inference — read 2026-10-23: The relevant window is 2026-08-24 through 2026-10-23. There is recent integration evidence for a prototype; there is no recent evidence of sustainable demand for another fee token.

Label — read 2026-10-23	Dated change	Consequence and limit
Claim, implementation format corroborated	2026-09-30: IMD documents migration to oracle domain version 2, adding panel-strength fields	Consumers can enforce quorum in a signed answer. Does not decentralize the single attester key. Oracle documentation
Record	2026-10-10: Ethereum Intake v2 verified; deployment block 26164897	Failure callbacks and result callbacks give a contract a usable lifecycle; payment remains upfront. Verified Intake metadata
Claim, current support corroborated	2026-10-13: IMD dates onchain jobs, launches, workflows and continuations becoming sold actions	Removes a custom backend payment integration; says nothing about acceptance escrow. Dated service documentation, current capabilities
Record	2026-10-15: OCTO source commit; CLAUS core freeze transaction	Prior art and a live extension surface make a bounded prototype faster. Current module authority prevents a hard-requirement-compliant CLAUS launch. OCTO commit, freeze transaction
Record	2026-10-18: KEPT hook and vault live records	An actual vault-funded service precedent, with explicitly disclosed subsidy attacks. Launch
Record	2026-10-21: Ouroboros Press chapter completion	Contract-purchased creative output and onchain receipts have happened, rather than being only a roadmap claim. Completion record

9. MVP plan of 14 days or less: contracts, a Foundry test plan, and the exact IMD jobs to order after this one

Inference/plan — read 2026-10-23: This is a conditional research-to-launch plan, dated 2026-10-24 to 2026-11-06. This assignment orders none of it. Start with the validation branch; stop before production deployment unless all hard requirements pass. Do not let a successful local mock substitute for a mainnet payment or admission proof.

Inference/design dates	Deliverable and gate
Oct 24–25	Freeze brief/rubric and epoch policy; recruit an actual sponsoring community; have the sponsoring community approve the two validation briefs below (not yet a vote by the unlaunched token); obtain IMD policy/agent-payment evidence and decide burn semantics
Oct 26–27	Implement FeeJobsHook, fixed token, JobsVault, stake checkpoints and HolderRewards; static receipt/vote UI; fork tests only. Burn path remains explicitly disabled in any runnable candidate until its immutable route and semantics pass review
Oct 28–29	Independent adversarial review, repair findings, reproduce actual factory initialization and fee claims; prove callback authentication and accounting; resolve fair opening rather than assert it
Oct 30–31	Sponsor buys the two approved operating jobs; assess quality and payment evidence. Before launch these are wallet-funded validation orders, not proof of autonomous fee procurement
Nov 1	Go/no-go and conditional launch: all seven requirements, exact request check, verified burn path, real sponsor and funding threshold must pass. Otherwise publish the no-launch result and reference implementation. Launch no later than this date if attempting a complete live proof
Nov 2–6, only after go	Bootstrap staking snapshots, vote and wait the public execution delay; run one bounded operating epoch with two new approved jobs and two acceptance checks; publish the full request/result/assessment/burn trail. If completion cannot fit by Nov 6, classify the experiment incomplete and do not claim the 14-day proof

Inference/Foundry plan — read 2026-10-23: Use the actual factory/PoolManager interfaces and reproducibly vendor dependencies at explicit commits. Tests must cover: permission bits and every disabled callback; factory-only initialization without transfer/venue restrictions; mint-once supply; exact input/output in both pair orderings; dust, rounding, full-fill rejection and slippage; no fee bypass via empty/malformed hookData or another router; multiple swaps in a bundle; conservation of ERC-6909 claims; sweep reentrancy and hostile token transfers; vote checkpoint reuse and flash-stake attempts; arbitrary body substitution; prepayment price drift; duplicate execution; callback gas and forged/late/failed callbacks; separate Intake/oracle request IDs; false, stale, wrong-consumer and under-quorum attestations; epoch-close accounting with outstanding prepaid jobs; actual burn evidence and shared-decimal dust; holder liabilities surviving epoch close; no owner withdrawal or replacement path. Fuzz fee conservation and invariant-test that only approved Intake job orders consume job funds, while oracle checks use separately sponsored assessment funds. Fork-read tests should use §13's pins; deploying test contracts inside a local fork is future testing, not a deployment performed in this report.

Inference — exact subsequent IMD order objectives, read 2026-10-23: Every paid action below costs 0.5 IMD at the observed price. Supply real accepted repository commits and request hashes at execution; none are invented here. Use the free checker before any payment. The future request's implementation skill/schema must be resolved from current capabilities/docs, not guessed as an undocumented skill name.

Order	Exact objective / acceptance	Spend
A — job.open, first	“Implement the fork-only Fee to Jobs reference in sections 4–5 of this report: immutable hook/vault, plain fixed-supply token, checkpointed voluntary voting and separate holder rewards. Implement the Foundry matrix in section 9, vendor all dependencies, and provide a reproducible local fixture. Identify every unmet launch gate. No deployment, no invented production addresses, and no claim that queued IMD retirement is completed burning.” Accept only reproducible accounting, adversarial tests and an explicit gate checklist.	0.5 IMD
B — job.open, after A	“Independently review A's exact accepted commit for fee bypass, partial-fill overcharge, request/attestation substitution, holder-vote reuse, premature epoch burning, administration and factory compatibility. Reproduce each material finding with Foundry; do not deploy. State whether each of the seven mandatory requirements passes, fails or remains unverified.” A different review run must assess evidence, not merely restate tests.	0.5 IMD
C — job.open, first holder-approved operating brief	“Using the frozen evidence and fee sample in this report, compare Fee to Jobs with pinned Ouroboros Press, OCTO, SIMD and KEPT. Identify one concrete user problem they leave unresolved, assess whether a fee-funded community would pay to fix it, and deliver a cited, falsifiable recommendation with explicit contrary evidence. Do not infer demand from token volume alone.” Accept only attributable evidence and an actionable, nonduplicative recommendation.	0.5 IMD
D — job.open, second holder-approved operating brief	“Build a static decision page from this report and its frozen evidence: show actual versus hypothetical fee income, minimum job budgets, prepaid versus accepted states, exact brief hashes, artifact receipts and unresolved launch gates. Include source/read-date links, accessible tables and an offline preview; never display sample balances as live funds. No wallet transactions or deployment.” Accept only a usable page that faithfully represents the evidence and passes independent rubric review.	0.5 IMD
E — launch.open, only after every launch gate passes	Launch the reviewed immutable build of A with all adopted repairs, admitted opening rules and verified burn semantics. The final request must pin its real commit, chain, paired IMD, manifest and exact allocation. If any hard requirement still fails, do not order E.	0.5 IMD
C-live and D-live — job.open, after E	Repeat the two operating objectives with actual fee-funded approved orders, fresh evidence windows and fresh result hashes. Link each accepted output onchain. The prelaunch wallet-funded artifacts remain separately labeled validation work. Freeze all actual inputs before the live holder vote; D-live does not depend on C-live completing, so neither body needs post-vote substitution.	2 × 0.5 = 1.0 IMD
F and G — oracle.request, only in the launched experiment	The sponsor, using the separate assessment consumer, asks §5's exact bool-only acceptance question, individually instantiated for each of two approved operating outputs and its frozen rubric/evidence. JobsVault cannot fund these actions. Do not ask a vague question such as “was the job good?”	2 × 0.5 = 1.0 IMD

Inference — budget derivation, read 2026-10-23: Validation A–D: 4 × 0.5 = 2.0 IMD. The conditional live branch adds launch E 0.5, new operating jobs C-live/D-live 2 × 0.5 = 1.0, and acceptance questions F/G 2 × 0.5 = 1.0: 2.0 + 0.5 + 1.0 + 1.0 = 4.5 IMD after this report. Including this report adds 0.5, giving 2.5 IMD validation-only or 5.0 IMD with the live branch. This counts the prelaunch and postlaunch operating orders separately; a wallet-funded order is never represented as a fee-funded one. No retries, continuations, extra reviews or schedules are included; each additional paid job/oracle adds the observed unit price. Gas and cross-chain messaging remain unquoted, not zero.

10. Kill criteria: at least three, each measurable

Inference/decision thresholds — read 2026-10-23: These are proposed stop rules, not observed outcomes:

1. By Oct 25: no trace from a non-launch job to an identified executing agent's onchain compensation, or no acceptable documented way to fund that agent through the allowed Intake flow → stop the token launch under requirement 1.
2. Before any launch payment: any reachable product/module replacement or fund-redirection authority can execute with less than 24 hours notice; any post-launch mint, transfer fee, forced venue or token freeze exists → stop.
3. By Nov 1: no reproducible defense against preferential creator fills and bundled opening advantage compatible with the factory → stop under requirement 5. Equal fees alone are insufficient evidence.
4. Demand: no actual community funds the two-job, 1.0 IMD operating budget by Oct 25, or either of the two outputs fails its frozen mandatory acceptance rubric → stop scaling; publish the failure.
5. Revenue: actual collectible jobs fees in the first 24-hour operating epoch are below 1.0 IMD for two prepaid jobs → do not claim job procurement is self-funding; at 25 bp, this requires 400 IMD gross qualifying volume before other costs. Two independent oracle checks need another 1.0 IMD from the review sponsor. Do not claim the entire assessed-work system is self-funding while that subsidy remains.
6. Burn: any unspent closed-epoch job budget remains spendable on new work, any holder reward liability is destroyed, or a retirement-queued balance is reported as burned without receiver/supply evidence → fail the experiment. If instant same-chain IMD burning is mandatory, stop before implementation of a live version.
7. Volume: compare a preregistered matched-window pilot with its no-incremental-fee baseline; an added fee associated with more than 20% deterioration in non-sponsor swap volume triggers review and blocks expansion. Preregister exclusions for known sponsor-funded and circular trades; report unclassifiable activity separately, since router identity cannot prove beneficial ownership. Without a credible comparator, report elasticity unknown, never “volume unaffected.” This is a research threshold, not a causal claim from this sample.

11. Moonshot

Inference — read 2026-10-23: The boldest defensible version is a holder-funded, continuously maintained public product whose fixed hook revenue commissions new investigations, code and interfaces under immutable budget limits. Each proposal names a real user problem, acceptance evidence and an artifact lineage; agents propose work, holders authorize exact briefs, independent panels assess results and onchain receipts make every expenditure auditable. It could outlive its creator because no founder is required to authorize the next already-approved order. It still needs paying traders, keepers, accessible artifacts and an operating IMD service; neither autonomy nor token appreciation creates those inputs.

Inference/experiment — read 2026-10-23: Between Oct 24 and Nov 6, compare two holder-approved outputs against the same frozen rubric and the same alternative of commissioning those outputs directly for 1.0 IMD. Have reviewers identify whether either result changes a concrete product decision, while separately tracing the full funding/receipt/accounting cycle on a fork. A real sponsoring community pays the operating budget on day one; the builder funds A/B's 1.0 IMD validation work. The falsifiable outcome is whether useful work and demand exist without manufactured trading. This experiment does not need a new token to answer that question. A live token stage requires all gates in §10; it is not grandfathered around the seven requirements by calling itself a moonshot.

Inference — why it may be a bad idea, read 2026-10-23: A token is an expensive coordination layer for work that can already be bought directly. Fees can punish the users funding the product, governance can reward attention rather than useful maintenance, and compulsory epoch retirement can destroy the small balances most pools need to accumulate enough to buy even one job. An attested report can be accurate yet irrelevant. Prepayment leaves buyers exposed to refusal and poor delivery; independent assessment doubles the observed IMD unit cost. OCTO and existing vaults already cover much of the mechanism. The strongest result of the experiment may be evidence that a conventional community treasury and direct IMD orders deliver the same benefit with fewer contracts and no speculative opening.

12. Open questions and everything you could not verify

Record / Inference — read 2026-10-23: The following remain unresolved; none are treated as favorable evidence:

• Worker pay: a non-launch job's acceptance-gated onchain payment to its executing agent was not found. Intake's treasury transfer is not substituted for that proof. Launch token entitlements are not proof of per-report wages.
• Acceptance: no present Intake method delays its base payment until acceptance. CLAUS documents escrow with acceptance, but its paid lifecycle was not exercised and is a different rail. No new escrow address is asserted here.
• Burn: literal end-of-epoch destruction of native Ethereum/Robinhood IMD was not verified. Ethereum's current BurnExecutor receiver was read as 0xf9d7cbf5bef2f5c9ba93a70f31ddca6457716793 on Base, whose source is verified; its executor retains owner rescue/configuration. A complete bridge→burn transaction, live native fee quote, dust handling and Robinhood equivalent route were not established. No guessed burn destination is supplied. Receiver verification.
• Fair opening: no admitted batch auction or equivalent opening mechanism was verified for this idea. Surcharge-based deterrence is not a guarantee against profitable sniping.
• OCTO deployment: not found. Source config has empty hook/DAO/token fields. Searched four pages of 100 launch records, reaching 2026-10-13; text matching cannot exclude an unrelated launch name. It is not called live on source marketing alone.
• SIMD/KEPT: deployment records corroborate live code; this report did not reproduce all advertised operating flows or independently audit all their contracts. Their README test counts are not this research's test results.
• Address verification scope: registry addresses are officially listed and have code; a full implementation/admin audit of those proxies was not done. Robinhood Intake's compiler match was not found. Proposed product addresses and an approved production opening contract are not found because this work deployed nothing.
• Launch admission: the exact proposed manifest/bytecode has not been submitted or admitted. Existing vault-directed fees establish precedent, not approval of every fee or opening rule. Archived launch records carry their own policy versions; they are not proof that current policy is identical.
• Demand and economics: no sponsor commitment, holder interviews, incremental-volume experiment, gas quote, bridge-fee quote, acquisition cost or profitability forecast. Cohort excludes older launches, non-IMD pairs and other venues. Some pools were created inside the window. API status and indexing may lag; chain observations are pinned, not perpetual facts.
• Service trust/liveness: no commitment that IMD continues operating after a creator stops; no independent proof that panel identities are independent; single writer/attester and mutable external pricing remain trust boundaries. Content hashes do not ensure permanent hosting.
• Pilot accounting: a wallet-funded validation order is not evidence of fee-funded autonomous procurement. Repeating both operating jobs after prelaunch validation changes the budget as explicitly derived in §9.

13. Sources with read dates

Record — all read 2026-10-23 UTC: Sources are primary unless marked otherwise. No X post or price page supports a conclusion. Dynamic sources are observations at the recorded read date, not permanent representations. Supporting normalized observations, raw swap logs, fee-call results, source hashes and signature-check output are retained in the repository's evidence/ directory. Those files and this report's structural checks carry no independent audit authority.

Source	Pin / scope
IMD docs and generated Markdown	Dynamic read 2026-10-23; launch terms, chronology and oracle encoding. Documentation is a policy claim where code does not establish behavior
Requests capabilities	Dynamic read 2026-10-23; prices, chains, token and Intake addresses, cadence minima
Swarm	Response at = 1792784366562, 2026-10-23 19:39:26.562 UTC; health counters only, not proof of quality
Launch list	Latest 100 records; all 44 IMD pairs selected before measuring activity; paginated older records used separately for OCTO search
Ethereum RPC reference	End block 26257482, hash 0x9217c1d80cf913f5ae8c019bcfc77b9adf05bfd15c35be2e02f4f6202aeed7e7, 2026-10-23 19:40:59 UTC
Robinhood RPC reference	End block 96189684, hash 0xc64bcccb237d3a1249031dca1424a5b0c49dca03857f9e832ecc2569f01885de, 2026-10-23 19:41:07 UTC
Verified Ethereum Intake	Source match; deployed block 26164897, transaction 0x75acc7ce4cdddeff526a9a6aa40e5dbebfe8608e1e6576bcb6e4f640a6fc90db; getters pinned to chain-read ends
Ethereum hook factory and project factory	Verified PoolFees split and non-broadcast claim simulations; factory addresses in sampled launch position records
Robinhood hook factory	Verified source; used for IMDirector claim simulation; pool/factory pair corroborated by deployment record
Ouroboros upstream commit	Latest upstream commit 2026-10-21 22:26:00 UTC; source inspection, distinct from deployed build
Ouroboros launch and verified contract	Deployed source 3468a915f68531f4617dba0e7b29a8126e6cac6a; block 94495874, 2026-10-21, address 0xfc65446d2c8a7ee6b173d62db710b76bd8bc978a
Ouroboros completed request	Order block 94503944; delivery block 94504995, transaction 0x41bb36518902c547cf2ec0d09e25e189f8329f83394e36e21d8f6e1b4e10eda0, 2026-10-21 18:52:23.527 UTC API update
OCTO pinned tree	Latest commit 2026-10-15 12:51:04 UTC; hook, DAO, config inspected as text
SIMD pinned tree and launch	Latest commit 2026-10-21 10:30:29 UTC; deployed block 94222549; vault 0xbba92992309c61e1c483ab630d0530f2c9edede7
KEPT hook pinned tree	Latest commit 2026-10-18 21:29:54 UTC; hook deployed block 92090706; address 0xb5f28198750864e199477a6898bad44ae4d1e0cc
KEPT vault pinned tree	Latest commit 2026-10-18 16:25:19 UTC; deployed block 91910657; vault 0x3152f9cbf64d0c1ac24ea7e101db3e4d8da86896
IMDirector launch	Commit 41e3e33df7215df0845ff642c75fd0f6c8b32142, 2026-10-21 21:16:51 UTC; deployed block 94601188; poolId in evidence sample
YOLO launch	Commit 0b50e441ed68cdf4176c79e8671018766bd48198, 2026-10-21 09:25:07 UTC; seeded pool and position from selection record
Hamster launch	Commit bf12d716b9268cbe820ee776be2a990b4f2c09bd, 2026-10-20 19:55:27 UTC; deployed block 26236027
INK launch	Commit 620960b344306b3469330eb0a6f9fb2a48c329a5, 2026-10-20 17:49:00 UTC; deployed block 26235399
Dongle build	Source commit from launch selection; launch created 2026-10-22 17:56:21.911 UTC; latest Git commit date not independently checked
Official ERC-8004 deployment list	Latest commit 2026-08-15 16:26:49 UTC; deployment-list claims plus observed target-chain code; not cited as a change inside the recent window
Uniswap v4 core	Latest commit 2026-10-23 16:23:26 UTC; API/permission design reference, not a claim that every deployed pool uses that exact build
CLAUS live fee state	Block 26257487, hash 0xdcdd7a8d77537534eeeb350969d2c4b87f7cc191d017f375e49780dbb63244ac; block time 2026-10-23 19:41:59 UTC, observed 19:42:09.034 UTC
CLAUS hook source and implementation	Proxy deployed block 26121996; final implementation deployed 26197758; EIP-1967 slot read at Ethereum end pin
CLAUS host source	Exact match; deployment block 26197757; cap, router sender, hookData, receipt execution and setModule authorization
CLAUS ProxyAdmin	Renunciation block 26197788, hash 0xa0c545a523c6a97f4459fd38e42dad74a3682b2dcf4802d9236248a96d33ce4d, 2026-10-15 12:03:11 UTC
CLAUS current governance and protected token implementation	Governance exact match, deployment block 26233391; token implementation exact match, block 26197756; owner getter at Ethereum end pin
CLAUS Jobs docs, builder guide, capabilities	Dynamic read 2026-10-23; acceptance/escrow documented; inference disabled, not independently exercised
Oracle request list and checked attestation	Attested 2026-10-23 18:58:46.234 UTC, to-block 26257254, hash 0xe8abd1dcf21d728a19ee8c1f791a5bc9fa6b681a970b7134316cde42c6facb6e; signature verified after documented enum normalization
IMD OFT source, POOL4 docs, BurnExecutor	Verified burn executor source; current owner/receiver read at Ethereum end pin; no full bridge execution claimed
Live-configured Base receiver	Exact source match; deployment block 46046062, transaction 0x7bd1874f302e2c16e6d121b17123f33254c2ad7cd9399873c788f76001d32627; source-only inspection, no Base execution or tip observation

Record — verification limits, read 2026-10-23: Local checking validates the report's section structure, source-link presence, arithmetic against retained observations and evidence-file consistency. It does not certify source honesty, economic causality, smart-contract safety or a successful implementation. No claimed Foundry result in this report refers to a suite executed for an unbuilt FeeJobs product.