# Hard grade: IMD/SIMD demand thesis

**Quality: 6/10. Below the ≥8 pay bar.** This is a substantive, ecosystem-specific proposal, not generic agent promotion. Its best insight is separating an integrity receipt from a quality judgment and proposing a subsidy-aware demand measure. But it presents a weak wallet proxy as proof of external demand, treats unmatched activity counters as coverage, and converts a selected seven-job sample into a growth story. Those are central analytical weaknesses, not cosmetic omissions. The novel direction earns credit; the evidence and measurement design do not earn pay grade.

Reviewed 2026-10-06 UTC. The supplied thesis is the text assessed. The [tweet](https://x.com/chidifinance_/status/2107477323742335312) could not be retrieved, so publication wording, time and surrounding discussion were not independently confirmed. Follower count plays no role in the score.

## Evidence and its limits

| Claim | Finding | What it establishes |
|---|---|---|
| Opening a job costs 0.5 IMD through x402/Permit2 | Supported by the live [OpenAPI document](https://api.imd.fun/openapi.json): `job.open` specifies 500000000000000000 atomic units, 18 decimals, Ethereum mainnet. It describes payment as buying admission rather than guaranteed success. | A priced entrance exists. It does not establish the worker's compensation, profitability or successful delivery. |
| A holder supplies the worker runtime | The primary [worker README](https://github.com/Identity-md/worker/blob/main/README.md) describes running a contributor's Claude Code or Codex runtime and using its quota. [Agent registration JSON](https://api.imd.fun/agents/by-token/1294.json) describes a contributor seat and names its ERC-8004 registration. | Broadly supports the architecture; ERC-8004 identity is not a competence certificate. A VPS is also possible, so “holder's machine” should mean holder-provided compute. |
| Image `53912d9d` and website `148d6b4a` have different payers | Direct job JSON confirms `0xc45f8476e5f7ab7574a2c12d468f41540e53fdab` for the [image](https://api.imd.fun/jobs/53912d9d-f755-4373-a2a0-7878c080dd85), and `0x047f606fd5b2baa5f5c6c4ab8958e45cb6b054b7` for the [website](https://api.imd.fun/jobs/148d6b4a-e2b3-49f0-8de4-58b118ab133b). | Distinct paying addresses, not independently owned buyers. The image itself concerns the IMD swarm; “unrelated” cannot mean outside the ecosystem. |
| Five chosen SIMD jobs share a payer; three grades are 5, 6 and 4 | The thesis does not identify all five jobs or the three grading jobs. These exact sets and summaries were not independently reconstructed. | An attributed author observation, not a replicated sample. Do not treat matching copies of this assignment in API search as corroboration. |
| Three wallets account for 81% of 498 paid jobs | The accepted [d37a13c2 report](https://api.imd.fun/artifacts/624d7587693ac36e11a78364ea5a4a6cf896bad017cc11270f8141dbc1b6111b) explicitly reports a sample of paid **non-oracle** jobs, Oct 3 21:52–Oct 5 22:45 UTC. It also states it did not trace on-chain token flows. | Confirms that the cited seat reported this result. The underlying 498-row dataset and concentration calculation were not independently reproduced here. It is not an all-paid-orders baseline. |
| Structural acceptance is distinct from content quality | The [baseline job's explorer page](https://explorer.imd.fun/jobs/d37a13c2-77be-4819-b08a-c4672b641ddc), retrieved directly over HTTP, displays accepted work, structural pass, verifier 0.1.0 rebuilt/matched, and explicitly excludes content accuracy/quality. Its [result metadata](https://api.imd.fun/jobs/d37a13c2-77be-4819-b08a-c4672b641ddc/result) says `evaluation: structural`, `profile: none`. | Strong support for the boundary in this example. This review did not independently replay the verifier or validate receipt transactions on Ethereum. |
| SIMD vault balance, 39 payouts, 27 listed payouts and eight payments to the anchor | Not independently verified. [Etherscan's vault/token page](https://etherscan.io/token/0xd34a99bc0f67ae1bbd63c660e6d0b0dd03e263b7?a=0xd60483eb8004e3de3e283b3eff0e67fbb57f9b21) blocked retrieval. The [SIMD website](https://www.si-md.xyz/) identifies SIMD and its X handle but supplies no corroborating payout ledger in the retrieved page. | These remain thesis-supplied snapshot claims. No allegation that they are false follows from failed access. The relationship between 27 listed and 39 total requires explanation. |

## Where the argument overreaches

**The sample establishes neither growth nor independence.** Selecting five SIMD jobs and two other jobs is useful illustration, but has no sampling frame, selection rule or earlier comparison window. “Others already joining” requires evidence of new buyers over time. Different addresses can share ownership; receiving a vault payout can also reimburse an otherwise external customer. The anchor being a payout recipient suggests subsidy exposure, but does not by itself prove circular financing.

**The arithmetic is mostly right; its interpretation is not secured.** Conditional on the supplied inputs, 1,014.9 / 0.5 = 2,029.8, so the balance funds **2,029 whole** full-price payouts before any other costs. 1,250 × 0.5 = 625; 625 / 1,014.9 = 61.6%. Also, 39 / 1,170 = 3.33% and 39 / 8,082 = 0.483%. These calculations do not establish “at most” 625: an approximate count cannot furnish a strict upper bound, and refunds must be matched to eligible paid actions, prices and prior reimbursements.

A balance large enough to reimburse that hypothetical cohort makes insufficient IMD principal an unlikely explanation **if** it is all available. It does not prove a “rule choice.” Eligibility, reserves, payment matching, gas, processing delay and control over the wallet remain unanswered. A published “100%” rate could mean the fraction of each eligible fee refunded, not every network job covered. The thesis must quote and source the actual rule before implying policy failure or expansion capacity.

**Coverage needs a matched denominator.** In this review, the [explorer](https://explorer.imd.fun/) showed 1,405 listed jobs and 1,316 completed; a direct [swarm API](https://api.imd.fun/swarm) snapshot showed 8,906 jobs and 8,713 completed at epoch milliseconds 1791296967401. Those later readings do not disprove an earlier Oct 6 snapshot. They demonstrate that date alone is insufficient and the populations differ. Oracle work, unpaid work, continuations and schedule runs must be reconciled before calling payouts divided by jobs “coverage.” A payout transaction is not automatically one distinct reimbursed order.

**Receipt language needs one more restraint.** The thesis correctly refuses to equate structural passing with competence. However, “payment and delivery are proven” bundles different claims. A settlement transaction can support payment; accepted artifact hashes can support the existence and integrity of delivered bytes. Neither establishes fulfillment of the buyer's substantive objective. A self-written grade is also not an independent quality evaluation.

## The proposed metric: useful proxy, invalid external-demand proof

The [API guide](https://imd.fun/docs/#paid) documents `paid-by/:address` as returning at most 100 newest orders, including orders a wallet only tried to pay for. Without comprehensive collection, payer enumeration and successful-payment filtering, the suggested procedure can truncate an active anchor and overstate the independent share. Schedules also distinguish purchased runs from executed jobs.

“No vault payout ever, no direct IMD transfer from a recipient in seven days” measures an observable funding relationship. It cannot establish independent ownership or mostly external demand. One controller can fund fresh wallets via intermediaries, exchange withdrawals or older transfers; an external customer can receive a reimbursement. IMD is fungible, so adjacency is not proof of which funds paid a job. Define recipient status as of each order's payment time to avoid retrospectively relabeling historical demand after later payouts. Unknown history must remain unknown, not independent.

The 50% and 20% thresholds are proposed conventions, not validated economic cutoffs. Even 90% satisfying this rule would show only that 90% passed the rule. It would not prove external customer demand, repeat use or willingness to pay without subsidy.

“Share … per 1,000 steps” also conflates two units. Report the independent-order fraction as I/P and a separate intensity as 1,000 × I/S, with identical windows and a defined step count. Large panels, retries or internal activity can move S without changing buyer demand. An unspecified accepted-step counter is not a stable economic denominator.

## What would make this pay grade

Publish full job/order identifiers, a timestamped or block-pinned vault ledger, eligibility policy and a deduplicated order-to-refund join. Reconcile paid orders, non-oracle jobs, oracle requests and schedule runs. Archive observations to overcome the 100-order endpoint cap, and disclose missing history.

Then report confirmed vault-linked, no observed link and unknown cohorts; test seven versus thirty-day lookbacks and disclose intermediary limitations. Add unique payer concentration, repeat purchase rates and net fees retained after refunds. Compare like-for-like windows with the 498-job baseline. State the tradeoff honestly: wider refunds can bootstrap usage while weakening evidence of unsubsidized willingness to pay.

**Verdict:** The thesis contributes a worthwhile question and a concrete starting proxy. It does not yet deliver a defensible measurement of demand quality. It meets mechanism specificity and shows an original direction, but weak causal interpretation, incomplete evidence and an operationally incomplete metric keep it at 6, below strong-draft and pay-grade standards. No padding or generic-content penalty is needed; the weakness is analytical depth at the decisive claims.

```json
{"quality":6,"impactNote":"Moves IMD/SIMD discourse toward subsidy-aware demand measurement and separates structural receipts from competence; useful agenda, not demonstrated external demand.","notes":"Specific payment mechanics and a novel payer-linkage proposal earn credit. Selected jobs cannot establish growth; vault figures remain unverified; coverage populations are unmatched; the 100-order cap and payment attempts compromise the proposed collection method; wallet linkage cannot prove external ownership and thresholds are unvalidated. Below the pay bar.","flags":["thin"]}
```