# Hard grade: subsidy changes who bears the IMD fee

**Quality: 6/10. Below the 8/10 pay bar. Flag: thin.**

The thesis identifies a useful IMD/SIMD mechanism: a job admission fee has different incentive effects when its beneficiary pays and when a sponsor pays. It names the fee, the vault, the accessibility tradeoff, and an observable funding mix. That earns credit beyond generic crypto commentary. But the core economics are familiar, the evidence is undated in the post, and its chosen metric cannot establish its behavioral conclusion. The argument is a competent outline, not a developed original analysis.

Assessed text: the thesis supplied in this assignment, attributed to @chidifinance_ and linked to [the submitted X post](https://x.com/chidifinance_/status/2107203457736192177). Direct X access failed; independent confirmation of the post's wording and publication time was unavailable. Research date: **2026-10-05 UTC**. Follower count and audience size do not enter this grade. No reach or engagement estimate is made.

## Claim audit

| Submitted claim | Evidence and status | Assessment |
| --- | --- | --- |
| Every `job.open` costs a fixed 0.5 IMD. | **Verified current advertised price.** The primary [capabilities API](https://api.imd.fun/requests/capabilities) returned `job.open` payment amount `500000000000000000`, with 18 decimals, on Ethereum mainnet. [Official API docs](https://imd.fun/docs/) corroborate the price and x402/Permit2 payment method. | Correct for the current paid action. “Fixed” should mean the present token-denominated tariff, not an immutable future price or constant fiat cost. |
| Early activity was about 115 paid jobs and 57.5 IMD. | **Attributed historical report; not independently reconstructed.** [Bankless's September 25 snapshot](https://www.bankless.com/read/inside-imd-ethereum-s-new-ai-swarm-experiment) reports 115 paid **orders** at 0.5 IMD each. Its supporting link leads to the live health API, not an archived snapshot. | The arithmetic is right: `115 × 0.5 = 57.5`. “First tracked period” lacks boundaries. Orders are not necessarily `job.open` jobs, completed jobs, unique customers, or proven unsubsidized demand. |
| SIMD holds hundreds of IMD and has moved toward 100% coverage. | **Operator-reported balance and policy, with important conflicting qualifications.** The primary [SIMD state endpoint](https://www.si-md.xyz/api/state), fetched at 20:33 UTC with state timestamp 20:31:54 UTC, reported a vault balance of **1,341.982757 IMD**, share `100%`, and named payout `0.5 IMD`. It also marked the reward rule **NOT IN FORCE**, `inForce: false`. | A funded vault and a stated full-coverage target are supported. Universal, automatic, consistently delivered reimbursement is not established. The current balance does not prove the balance at the tweet's publication. |
| Subsidy removes friction from the submitter. | **Conditional inference.** [SIMD's public docs](https://www.si-md.xyz/#docs) describe Hire opening jobs through a hot payer wallet, followed by vault refunds to that wallet. The docs also describe daily caps and a freeze after more than five paid opens in five minutes. The primary [burst status endpoint](https://www.si-md.xyz/api/hire/sent?burst=1) returned `limit: 5`, `windowMs: 300000`, and `locked: false`. | Direct sponsorship can remove the visitor's IMD payment. It does not remove every admission constraint. Reimbursement of a user who first pays would leave upfront funding, authorization, delay, and refund uncertainty. These flows need separate treatment. |
| 500–1,000+ IMD could cover thousands of jobs. | **Correct hypothetical arithmetic, subject to assumptions.** At 0.5 IMD per admission, 500 covers 1,000; 1,000 covers 2,000. The reported current balance would cover at most 2,683 whole admissions if wholly spendable for that purpose. | This is fee capacity, not demonstrated demand, completed work, or an estimate of total compute funding. Caps, other spending, and replenishment change the result. |
| User-paid/vault-covered ratio is the cleanest behavioral metric. | **Useful funding metric; weak behavioral inference.** The thesis supplies no time series, classification rule, spam measure, or comparison cohort. | Funding composition alone cannot show whether the fee filters behavior or whether the subsidy creates useful incremental activity. |

The primary API captures and documentation excerpts are saved in [evidence/](evidence/). They record public server responses, not an independent audit of balances, transfers, or enforcement.

## Why it stops at 6

**The mechanism is real, but the synthesis is thin.** “Someone else pays, so the beneficiary faces less monetary cost” is standard subsidy economics. The IMD numbers make it concrete; they do not make it novel. Repeating the fee-versus-accessibility tension adds length without a model of which users change behavior, under what policy, or by how much.

**It treats distinct payer relationships as a binary.** A visitor can submit a job paid by SIMD's payer, which the vault later reimburses. Alternatively, a user can pay IMD and subsequently receive a refund. The initial on-chain payer and ultimate economic bearer can therefore differ. Partial refunds complicate the division further. Counting the former as “user-paid” because the initial payer is a wallet, or counting transfers as jobs, would corrupt the proposed ratio. The official IMD docs expose `paidBy` and wallet order history, but those identify payment provenance rather than ultimate net cost. [IMD API docs](https://imd.fun/docs/).

**The available telemetry itself demands skepticism.** SIMD's captured state reports 17.25 IMD distributed, 69 `jobsPaid`, and zero `fullPrice` transfers, alongside its 100% label. Its frontend describes one outgoing transfer as one job paid. These fields do not demonstrate 69 fully covered jobs: 17.25/69 is 0.25 IMD per counted transfer, and transfers require reconciliation to specific admissions. The same state marks the release rule not in force. This is evidence of ambiguous counters, not proof that sponsorship never occurs. [SIMD state](https://www.si-md.xyz/api/state), [public frontend source](https://www.si-md.xyz/app.js?v=56).

**Its proposed test does not test spam.** Consider two hypothetical weeks with 90 fully subsidized jobs and 10 unsubsidized jobs. One has 90 useful first-time projects; the other has 90 duplicate junk submissions. The funding ratio is identical. Conversely, spam could rise through self-paying bots while the subsidized share falls. The metric measures reliance on a sponsor; usefulness and abuse require separate observations.

**The conclusion understates the remaining economic role.** Even if the visitor's net IMD cost is zero, each paid opening consumes the sponsor's funds. A finite budget can constrain aggregate admissions, while caps ration individual access. “More an accounting mechanism” is a plausible interpretation of the visitor's experience, not a demonstrated disappearance of economic scarcity. Nor does paying 0.5 IMD prove that fee equals the real cost of running agents.

These omissions prevent a 7: the post has named mechanisms and a tradeoff, but neither a strong evidential argument nor a sufficiently specified behavioral test. An 8 would require substantial original development. The grade does not credit this report's added analysis to the author, and does not assume the author had access to later snapshots.

## A test that would resolve the useful question

This is a proposed design, **not a completed empirical result**:

1. Define weekly cohorts of admitted `job.open` orders. Separate other paid actions, attempts, completions, and project continuations. Record the subsidy policy and eligibility in force for each cohort.
2. Join order/job IDs, initial payer, intended beneficiary where observable, and settled refund transfers. Classify fully sponsored, partially subsidized, unsubsidized, and unresolved jobs. A generic vault transfer is not sufficient linkage. Avoid treating multiple wallets as proven distinct people.
3. Report both the fully sponsored share and subsidy amount divided by total admission fees. Track counts, partial coverage, unresolved attribution, and reimbursement delay. Use admission cohorts with an explicit refund observation window so late refunds do not distort trends.
4. Alongside funding mix, measure duplicate/abusive submissions under a published labeling rule, completion and useful-output rates, repeat use, unique observable beneficiaries, and queue latency. A legitimate sponsored onboarding surge should differ from a duplicate-job surge even when funding shares match.
5. For causal claims, compare eligible and comparable ineligible cohorts around a dated policy change, accounting for growth, marketing, price changes, job mix, and caps. A simple before/after rise is not sufficient identification.

The primary [IMD health capture](https://api.imd.fun/health) at 20:33 UTC reported **1,179 paid orders across enabled actions**, not a historical sponsorship breakdown. It cannot prove that later growth was caused by SIMD or that the original 115 orders were all unsubsidized. No reconciled payer/refund time series was produced in this bounded review.

Unanswered questions remain: which jobs actually qualify for reimbursement; whether policy is discretionary or automatically enforced; what fraction receives full, partial, delayed, or no reimbursement; how much of the reported balance is available for job fees; and whether sponsored jobs improve useful output or increase abuse. Current operator docs and mirrored promotional updates describe evolving flows, so they should not be silently treated as one stable historical policy. [Mirrored @SuperIMD_eth updates](https://www.sotwe.com/SuperIMD_eth) corroborate the public coverage announcements, but their relative timestamps and inaccessible transaction-proof links limit historical verification. No independent Ethereum balance or transfer reconciliation was completed.

**Discourse impact:** the post usefully directs attention from nominal fees toward who ultimately bears them. Its main contribution is a monitoring question. It does not yet establish an anti-spam effect, subsidy-driven adoption, or an accounting-only fee regime. That is a useful 6/10 contribution, below pay grade.

```json
{"quality":6,"impactNote":"Directs IMD/SIMD discourse toward who ultimately bears admission fees and dependence on subsidies; supplies no measured behavior or causal result.","notes":"Concrete 0.5 IMD mechanism, correct conditional capacity arithmetic, and a clear accessibility tradeoff. Familiar subsidy economics, undated historical baseline, ambiguous payer/refund classification, and a funding ratio that cannot identify spam or usefulness limit originality and depth. Current primary telemetry does not establish universal automatic 100% reimbursement. Below the quality-8 pay bar.","flags":["thin"]}
```