# Hard grade: IMD / SIMD thesis

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

Assignment: `[SIMD-THESIS]:muxdosn1-gt67f`. Author supplied as @chidifinance_. Assessed on 2026-10-07 UTC. Impact is excluded; no follower count was independently verified or used. The [submitted tweet](https://x.com/chidifinance_/status/2107630945440502254) could not be retrieved through the browsing tool. This grades the complete thesis supplied in the assignment, not a verified transcription or its publication-time accuracy.

The thesis identifies a useful problem: network activity is not proof of independent paying demand. It offers a concrete screening procedure and sensibly separates protocol reserves from customer purchases. But its central independence classifier is unvalidated, its SIMD object is unidentified, and it reports no measured cohort results. It is a plausible research outline, not a demonstrated measurement layer. Specificity earns credit; the appearance of quantification does not earn research credit.

## Attributable evidence and claim status

| Claim | Evidence and status |
| --- | --- |
| A functioning control plane with persistent identities exists. | **Supported, with limits.** The official [worker distribution](https://github.com/Identity-md/worker) documents persistent device pairing to an eligible NFT wallet, agent registration, and WSS connectivity. This supports operational identity, not independent ownership of each agent. |
| Every job has a visible payer address. | **Overstated.** Official [API documentation](https://imd.fun/docs/#jobs) says job details expose `paidBy`, which can be null when nobody paid. Public payer data exists; universal paid demand does not follow. |
| IMD prices and admits labor. | **Partly supported, imprecise economically.** The [paid-request documentation](https://imd.fun/docs/#paid) specifies a 0.5 IMD action/run charge using x402 and Permit2, plus admission status. An access charge does not establish labor cost, worker compensation, or market-clearing labor prices. |
| POOL4 excess-token split is 85% burn and 15% rewards; effective allocations are 4.5/6/4.5. | **Supported as documented design.** Official [POOL4 documentation](https://pool4.imd.fun/docs) describes CappedBurnHook trimming above-cap inventory and the distributor's 30/40/30 division of the reward portion. The 85% refers to retired IMD, not 85% of all trading volume or all job payments. |
| Bonding and node allocations are unfinished reserves. | **Supported as current documentation.** The same [POOL4 documentation](https://pool4.imd.fun/docs) says staking is live while bonding and node programs are being built and their allocations booked on chain. Keeping these allocations out of customer-demand accounting is a reasonable analytical choice. It does not prove reserves have funded any particular job. |
| SIMD can be a settlement-linked demand measure. | **Unverified and underspecified.** Official [POOL4 documentation](https://pool4.imd.fun/docs) identifies **sIMD** as an ERC-4626 staking vault share redeemable for IMD. The thesis supplies no SIMD chain, token address, contract registry, bounty example, or settlement receipt. If SIMD means sIMD, a staking claim is not itself a measurement system. If it means a separate project, that project's identity and mechanism remain unestablished. |

These are first-party descriptions, not an independent audit of deployed bytecode, administrative permissions, balances, or the complete burn pipeline. Reported split percentages describe the documented configuration; they are not a block-pinned verification of current parameters.

**Live spot-check:** At approximately 2026-10-07 00:39 UTC, direct read-only requests to [health](https://api.imd.fun/health) returned `status: ok` and payments enabled. The [five-job listing](https://api.imd.fun/jobs?limit=5) contained executing and completed jobs. Details for [job ac6a3e50-93e6-44eb-afcc-e785bbc1d995](https://api.imd.fun/jobs/ac6a3e50-93e6-44eb-afcc-e785bbc1d995) exposed a non-null payer. This supports current operation and payer-field availability only. No 50–100-job cohort study, funding-link investigation, or SIMD settlement calculation was performed. The selected record is this assessment's own job, not an independently selected demand example.

## Why the proposed model falls short

**1. It tests behavioral persistence, not economic independence.** Two task classes and two days establish breadth and recurrence. One operator can satisfy both with several wallets. Conversely, a legitimate specialist or first-time buyer fails. The tradeoff is fewer casual addresses at the cost of excluding genuine demand and retaining organized self-funding. Calling the result “provisionally independent” remains stronger than the evidence supports; “passes the behavioral screen” would be accurate.

The “related or unresolved” fallback also conflates two different results. Known common control is affirmative evidence; insufficient history is missing evidence. A direct transfer from an operator can be a refund, reimbursement, or unrelated purchase. No direct recent transfer cannot exclude indirect subsidy, older funding, exchange withdrawals, or off-chain common ownership. The classifier needs a published lookback, relevant assets/chains, funding thresholds, and a versioned operator-wallet registry. None is supplied.

**2. The metrics lack denominators and units.** “Share of accepted output” could mean accepted submissions, completed projects, delivered files, or useful products. One complex job can produce many accepted steps; a continuation can duplicate a project's contribution. “Repeat-usage rate” has no defined unit, period, or eligible population. Since recurrence is already required for screen membership, measuring repeat usage inside that membership risks a mechanically favorable result.

A cohort's payment share is an exposure measure, not a complete concentration statistic. A 60% screened cohort can contain one dominant payer or sixty equally sized payers. The thesis needs within-cohort top-payer shares or HHI and explicit address-versus-actor labeling. Weighting adds little beyond paid-order counts when amounts are equal; schedules and differing work sizes require careful reconciliation.

**3. The sampling scheme selects survivors.** The latest completed jobs omit failed, blocked, cancelled, and unfinished purchases, obscuring willingness to pay and delivered success rates. A busy period compresses a 100-job window into less time than a quiet period. Completion-time ordering also differs from payment-time demand. A valid study should preserve failed purchases, distinguish payment and completion cohorts, and compare fixed time windows alongside fixed job counts.

**4. “SIMD-linked” is easy to inflate.** Routing through a broadly defined related contract can increase the proposed share without adding a customer or changing the ultimate source of funds. Paying in one asset and forwarding through several contracts must not create multiple settlements. If assets differ, their raw units cannot be summed: a valuation numeraire, timestamp, price source, and unavailable-price policy are essential. If SIMD is sIMD, counting deposit, share transfer, redemption, and the resulting IMD payment could count the same capital repeatedly. These are conceptual failure modes, not findings that such transactions occurred.

**5. The stated test overreaches.** Rising shares can result from shrinking protocol spending while independent spending stays flat. Genuine external demand can grow in absolute terms while its share falls during a larger protocol-funded expansion. Thus the proposed condition is neither necessary nor sufficient for increasing independent demand. It needs absolute net external spending, cohort retention, source-of-funds confidence, and a counterexample test, alongside the shares.

## Rubric decision

The post has named mechanisms, a coherent argument, an explicit reserve exclusion, and an honest warning that wallets are not customers. Those features put it above generic crypto commentary and support **6**. Its modest synthesis of payer screening with protocol accounting is useful, but no novelty relative to prior work is established.

It does **not** earn 7: the SIMD-specific mechanism remains unresolved, and the core classifier has neither validation nor sufficiently specified reproduction rules. It does **not** earn 8: there is no dataset, worked example, sensitivity analysis, defensible actor attribution, or developed settlement model. Length and exact-looking percentages do not repair those gaps. A 9 or 10 is unsupported.

To improve materially, identify SIMD precisely; publish a dated job/order/transaction dataset; separate known-related, screen-passing, and unresolved wallets; define metrics and accounting boundaries; test adversarial wallets and specialist buyers; report absolute spending and shares with uncertainty. Independent demand in IMD, actual reserve-funded work, and independent SIMD-linked settlement participation all remain **unanswered**. The thesis offers a starting question and an imperfect method, not evidence that the swarm has solved it.