# Hard grade: IMD / SIMD thesis

**Quality: 6/10. Pay bar: FAIL (requires ≥8).**

Assignment: `[SIMD-THESIS]:muxdosn1-gt67f`  
Author: @chidifinance_  
Submitted post: [X status](https://x.com/chidifinance_/status/2107630945440502254)  
Research date: 2026-10-07 UTC.

The thesis identifies a worthwhile problem: visible work and distinct payer wallets do not establish independent customer demand. Separating reserve funding from customer spending is sensible. But the proposed classifier cannot establish independence, the repeat metric partly rewards its own selection rule, and the SIMD measure confuses routing and subsidy with willingness to pay. This is a competent analytical outline with specific mechanics, not a demonstrated measurement model. Its specificity earns 6; unresolved validity problems keep it below 7, and it lacks the evidence and depth required for 8.

## Scope and evidence

The supplied thesis is the text graded. Direct retrieval of the X post failed, so correspondence with the live post and publication-time conditions remain unverified. No follower count was researched or inferred; audience impact is excluded from quality. This review checks mechanisms and the proposed method. It does not perform the author's proposed 50–100-job cohort study or claim that independent demand has been measured.

Primary pages and public endpoints were read on the research date. Project documentation establishes what the project publishes; it is not an independent audit of contracts or economic outcomes. Selected live API fields and source hashes are preserved in [evidence.json](evidence.json) for offline inspection.

| Thesis claim | Evidence and status | Consequence for grading |
| --- | --- | --- |
| IMD prices and admits labor. | **Supported narrowly.** The live [request capabilities](https://api.imd.fun/requests/capabilities) returned IMD on Ethereum, 18 decimals, `500000000000000000` atomic units for `job.open`. The [API documentation](https://imd.fun/docs/#paid-requests) describes x402/Permit2 paid admission. | Establishes a request fee, not a market price for every unit of worker labor or evidence of profitable delivery. |
| Persistent identities and a real control plane exist. | **Documented.** The official [worker README](https://github.com/Identity-md/worker#start-and-pair) describes NFT enrollment, ERC-8004 registration, one active device per NFT, and WSS connections to IdentityMD's control plane. | Concrete infrastructure; does not establish independent customers or operators. |
| SuperIMD runs the control plane. | **Misattributed or at least imprecise.** SuperIMD's own [published frontend](https://www.si-md.xyz/app.js?v=66) describes polling Identity.md's public control plane and opening jobs through a SIMD payer. | Distinguish the application and funding layer from the underlying IdentityMD network. |
| Each job has a visible payer. | **Overstated.** The [job API documentation](https://imd.fun/docs/#jobs) explicitly allows `paidBy: null`. A sampled [completed job](https://api.imd.fun/jobs/c94a4567-46ea-46da-bee9-c8442df0d27e) did return a payer. | Public attribution is useful where present; missing attribution must remain missing, not become a customer. |
| POOL4 uses 85/15, with 4.5/6/4.5 allocations and unfinished reserve programs. | **Supported as published design/status.** [POOL4 docs, sections 05 and 11](https://pool4.imd.fun/docs) describe 85% burn, a 15% reward bucket split 30/40/30, and bonding/node reserves while those programs are built. The percentages follow arithmetically. The docs identify bounded owner-tunable parameters. | Accurate mechanism detail. These are documented parameters, not a block-pinned audit of current contract settings. Reserve origin supports separate accounting. |
| SIMD participation strengthens independent demand. | **Conditional inference, not established fact.** SuperIMD's [frontend documentation](https://www.si-md.xyz/app.js?v=66) describes IMD vault refunds to a hot payer and IMD-paid Hire jobs; it labels an automatic reward policy not in force pending an on-chain release rule. | Subsidy and mediated payments directly complicate the suggested interpretation. Published mechanics do not verify actual refund totals. |

## Why the method falls short

**1. The independence filter measures behavior, not economic control.** Two task classes across two days are easy for an operator to manufacture. Funding through an intermediate wallet or outside the undefined “recent” interval passes the direct-link screen. Conversely, a legitimate specialist or one-time buyer fails it. A direct transfer can be reimbursement or payment for unrelated services; it does not by itself prove common control. These are logical counterexamples, not allegations about observed wallets.

The author correctly calls the rule provisional. That caveat prevents a claim of final proof, but does not validate calling its passing group independent. “Related” and “unresolved” also need separate labels: evidence of affiliation and absence of enough observations are different states. Unknown operators and incomplete funding history require explicit coverage reporting. A fixed lookback, dated wallet registry, link direction, asset and amount rules, and a stable task taxonomy are missing.

**2. The four numbers are insufficiently specified.** “Jobs funded” requires joining payments to jobs and handling schedules, continuations, refunds and unpaid jobs. “Share of accepted output” needs a unit: accepted attempts, final artifacts, or completed projects give different results. The sampled job's acceptance verdict is structural with profile `none`; it explicitly says no suite ran. Acceptance therefore cannot universally stand for substantive quality or usefulness. That is one observed example, not a finding about all accepted work. [Sample job record](https://api.imd.fun/jobs/c94a4567-46ea-46da-bee9-c8442df0d27e).

The proposed concentration ratio is a **cohort payment share**. It cannot distinguish one customer from many customers inside that cohort. A top-payer share or payer-level HHI would add the missing distribution information. Because repeat appearances are already required for the passing group, repeat usage in the same window is partly built into classification. Use a prior eligibility window and later retention window, with explicit denominators and observation time.

**3. The SIMD share has the wrong causal interpretation.** A routing tag says how work was financed, not whether the end customer paid for it. Several users can share a service payer; one reimbursed user can retain a distinct wallet. The proposed filter can thus undercount users behind a gateway and overstate their independent spending. Higher SIMD-linked share could reflect a larger subsidy budget while customer net expenditure stays zero. Free usage can still be valuable adoption; it does not demonstrate willingness to pay at an unsubsidized price. This is an inference from the published funding mechanism, not a measured outcome.

The definition also needs chain IDs, contract addresses, eligible transfer events, and payment-to-job joins. Token amounts in IMD and SIMD cannot simply be added: use a stated common valuation basis and settlement-time prices. Avoid counting a vault reimbursement and the associated admission fee twice. Specify both overall SIMD-linked share and the independent group's share of SIMD-linked value; otherwise denominator changes can obscure what increased. Keep SuperIMD's SIMD token distinct from POOL4's sIMD vault shares.

**4. The proposed window cannot support the growth conclusion by itself.** The latest 50–100 completed jobs are a convenience sample. They omit failed and unfinished paid demand, favor faster jobs, and span a variable amount of calendar time. Completion growth can change when execution latency changes. Compare fixed time windows, include all paid admissions and their outcomes, and report the completed-job slice separately. A rising percentage can accompany falling absolute spend, so publish both levels and shares. None of these quantities is actually computed in the thesis.

## Rubric decision and the missing contribution

This exceeds generic crypto commentary: it names payer attribution, CappedBurnHook, reserve allocations and an actionable cohort proposal. Its strongest synthesis is treating activity, capital origin and customer demand as separate questions. It also avoids inventing fixed explorer totals.

However, the useful insight is mostly a standard wallet-cohort framework. The thesis repeats its conclusion without testing its classifier against even one disclosed cohort or counterexample. The named mechanics do not rescue the incomplete SIMD interpretation. A 7 would require a technically sounder argument with the subsidy and gateway tradeoffs developed. An 8 would additionally need an original, reproducible application or a substantially developed model; length and numerical-looking rules are insufficient. A 9–10 is unsupported. Copying or plagiarism was not established, so no such penalty is imposed.

To make a stronger contribution, publish one frozen dataset with job IDs, UTC windows, settlement transactions, payer attribution, task labels and dated protocol-wallet labels. Separate end user, admission payer, ultimate funder and refund beneficiary. Report gross fees, subsidies and net customer spend alongside absolute and proportional results; keep POOL4-derived financing separately attributed. Validate the classifier against independently labeled examples, show sensitivity to its thresholds, and evaluate future retention without reusing its eligibility criteria. The concrete payoff would be demonstrating how much apparent demand survives subsidy adjustment and attribution uncertainty.

## Unanswered questions

- Which known-wallet list and funding history would make the three-part filter reproducible? How many payers would remain unknown?
- Can public records reliably join each admission payment and reimbursement to a unique job and ultimate requester?
- What exact SIMD contracts and settlement events belong in scope, and what valuation source is usable when liquidity is thin?
- How much usage persists after refunds end, and what fraction of accepted work is used by customers?
- Does any reserve-funded work in the study actually exist? The reserve documentation alone does not establish that it has financed jobs.

**Final verdict: 6/10, below the pay bar.** Useful question and credible POOL4 detail; no evidence-backed measurement of independent demand, and the proposed SIMD signal can strengthen precisely when subsidy dependence grows.