Hard grade: IMD prices work; SIMD observes and subsidizes it

Quality: 6/10 (integer, not /100). Task: [SIMD-THESIS]:muyg5wlq-racmg. Assessed 7 October 2026 UTC against the supplied thesis and live primary sources.

This is a substantive mechanism explanation with a useful economic objection. It earns more than the default 2–5 range because it separates observation, payment, verification and reimbursement, names addresses, recognizes inactive policies, and proposes falsifiable checks. It does not earn 8: the subsidy economics are asserted without matched receipts, the token discrepancy lacks the cited original announcement, and the opening describes an admission fee too strongly as a price for labor. Detail is not evidence.

The rubric supplied in the assignment ends midway through its 3–4 description. This verdict applies the supplied low-score default and rare-8+ standard; it does not invent missing official score bands.

What the evidence establishes

Thesis claim	Finding and attributable evidence	Evidentiary limit
A job.open costs 0.5 IMD on Ethereum.	Confirmed at the API level. The live capabilities route returned network: eip155:1, asset 0xd34a99bc0f67ae1bbd63c660e6d0b0dd03e263b7, amount 500000000000000000, decimals 18, and recipient 0x4e0fa57bde726079356537e2f34d671e9f41adbc.	A current tariff, not a lifetime fixed price or proof of completed labor.
IMD sells labor at a fixed price.	Overstated. The IMD payment documentation specifies x402/Permit2 and describes the purchase as action-admission, with resultGuaranteed: false.	Paying opens work; it does not guarantee a useful deliverable, buy specified worker hours, or establish worker compensation.
Seats use their own machines/accounts; ERC-8004 pairing and standing are IMD functions.	Documented. IMD docs describe worker operation with the holder's Claude/Codex account, NFT pairing, ERC-8004 binding, public standing reads, and reputation-batch routes.	“Key” is narrower than “account/paid plan.” Exact 2,000 supply and the assertion that every accepted submission becomes an onchain reputation event were not independently established here.
SIMD is an observer and payer, rather than the control plane or a named-seat dispatcher.	Strongly supported as published architecture. The SIMD Docs view, implemented in its published application, distinguishes public reads from paid Hire and says job.open has no destination seat. The state route contains public-route statuses and stored observation/inference/execution records.	Product documentation and observable state corroborate the separation, not every backend execution path. A cycle's “verify” stage is not an independent audit of task quality.
Missing values should say UNAVAILABLE.	Published policy, with observable missing fields. SIMD's application documents the label; state includes a null chain-head field explicitly marked unavailable.	Not a comprehensive UI conformance test.
Hire opens real paid IMD jobs.	Corroborated directly by IMD. SIMD identifies payer 0x9fAdAB91f6FA03DBd7F4F8a08A338704bAAcf63f. The IMD paid-by route returns admitted orders and confirmed payment transaction hashes for that address.	Payment/admission does not prove completed work or independent customer demand. The 100 returned orders are capped, not an all-time total.
The vault refunds the payer.	Reported by the project API, not independently confirmed from chain logs here. The state snapshot names vault 0xd60483Eb8004e3DE3e283b3efF0e67FBb57f9B21 and reports two 0.25 IMD outgoing rows to the payer, at 14:39:23 and 18:02:35 UTC.	Outgoing transfers alone do not identify the job, reimbursement entitlement, revenue origin, or coverage ratio. Etherscan could not be fetched in this research session.
A standing 100% reward policy and Contest automatic payouts are not in force.	Confirmed as published policy/status. SIMD Docs states both limitations; /api/state returns reward.inForce: false.	This concerns the named policies. It does not mean every payout path is disabled: the same docs separately describe Thesis automatic payments as enabled by default.
More than five opens in five minutes triggers a freeze.	Documented and corroborated by configuration. The burst-status route returned limit: 5, windowMs: 300000, lockMs: 3600000, count: 1, locked: false.	The live freeze transition was not exercised. Call it a documented guard with visible configuration, not fully tested enforcement.
Website still publishes the older SIMD address.	Confirmed. The homepage and published app name 0xbb0c1f82a2ea0253ea3d91c2f05caded82133415.	The alleged 7 October announcement naming 0xb6026518463ec80f1cb33e4318e802487c27f995 was not retrieved. Different address strings do not establish migration, token equivalence, or fee flow.
verifier-rerun and verifier-paths have the stated behavior; sIMD is a staked IMD share.	Unresolved in this review. The retrieved IMD documentation does not establish those exact verifier profiles or sIMD semantics.	No credit for treating these as independently verified facts; absence from retrieved docs is not proof they are false.

The missing receipts matter

One direct paid-order example is order 963252e5-84e9-4777-bdaa-8463e6c7425c, admitted to job c5590743-bb22-486c-8bc4-7316d094a329, with payment hash 0xfbddceec64c612d4e70dc3aff9c1401853c263179691ed0001264dbf8e36aec2. Both IMD's paid-by response and SIMD Sent identify it. Sent describes a SIMD stress test and reports it executing. That proves a platform-recorded paid job, not an external builder purchasing useful output.

The vault snapshot reports a 0.25 IMD transfer to the same payer at 18:02:35, transaction 0x0d3e8812ff29c981c42f766356d3f5c160cf1ad9bfc822f1356a162cd85947b4. This is a plausible reimbursement candidate after the 18:01:25 paid open, but temporal proximity is inference, not a matched refund. The thesis supplies neither this match nor any stronger allocation record. Source: SIMD state.

The sampled Sent response contained 20 tasks, two marked complete, and zero tasks with listed files. Its resultAvailable flags were false throughout. This establishes a visibility gap in that response, not missing or useless artifacts across IMD. I did not successfully inspect a completed artifact or independently reproduce its work. The thesis's “files a stranger can open” test therefore remains unanswered. Source: SIMD Sent.

The “fees cover about 50%” and “some batches refunded in full” account claims are unverified attribution: no original post URL, calculation window, matched job set, or transaction reconciliation is supplied. A 0.25 IMD transfer equals half the listed opening fee arithmetically; it does not establish that fees funded it or that this is the average coverage rate.

Where the argument is strong—and where it fails

Useful inference: if the requester receives reimbursement, its net cost can fall and paid-open counts become a weaker proxy for its own capital commitment. That is a reasonable economic hypothesis. The author deserves credit for asking about circular funding and public outputs rather than equating activity with value.

Unproven inference: subsidized opens increase demand or make work cheaper for outside builders. Hire uses SIMD's payer; a visitor may not bear the original fee. The argument must specify whose commitment is measured: visitor, payer, vault funder, or token trader. Subsidized execution can still produce valuable files. Conversely, a paid receipt can represent a self-funded stress test or thesis review. Neither automatically measures external demand.

Weak falsification: if the vault stops refunding tomorrow, that does not disprove yesterday's subsidy. Use a defined period. Failure to see jobs on one explorer page can also reflect indexing or availability, rather than false payments. Check the authoritative paid-order response, payment transaction, and artifact together. Documentation address changes alone cannot prove a migration: require deployment identity and a traceable fee mechanism.

Originality: the control-plane/observer split and many operational details closely follow SIMD's own published Docs. They are useful explanation, but paraphrasing documentation is not original research. The subsidy-versus-demand-signal tradeoff is the contribution that moves this above generic promotional content.

What would justify a higher grade

Provide a small reconciled ledger: payer, admitted order, payment hash and amount, completed job, accessible artifact, vault transfer, beneficiary, and net payer cost, all within a stated time window. Verify transfer logs independently. Retrieve the original contract announcement and establish whether it describes a replacement, a separate fee-generating token, or something else. Separate outside requests from SIMD-originated experiments and contest administration. Support the verifier and sIMD terminology with primary technical sources.

Until then, 6/10 is the defensible quality grade: specific and thoughtful, but mostly documented architecture plus an unmeasured economic thesis. Impact is separate and not scored. The assignment's followers≈59 is supplied metadata, not verified audience or engagement. No follower count, reach, originality certainty, or payout eligibility has been invented.

Research limits

The submitted tweet returned HTTP 403 through the web tool; its actual text, publication metadata and engagement were not verified. The supplied thesis is the object graded. SIMD loaded through direct HTTPS requests after browser-tool retrieval failed; /docs was 404 because Docs lives at #docs. Direct API reads succeeded. The naked identity.md domain returned unrelated passport-service content, so technical research used the IMD project's linked imd.fun documentation.

Published app code and API responses are primary project evidence, not independent certification. The report includes no successful onchain log audit, completed-artifact inspection, enforcement test, or original 7 October account-post verification. Those limits lower the grade; failed retrieval is not reported as completed research.