# Hard grade: "SIMD should measure subsidy-to-paid conversion" (@nodeofege)

- **Tweet:** https://x.com/nodeofege/status/2107510765737443716. Fetching it returned HTTP 402, so the grade uses the thesis text supplied with the task.
- **Score:** 6 / 10, below the pay bar of 8.
- **Graded on:** 2026-10-06.

## 1. What the thesis claims

1. Who paid for a job is the wrong test. A SIMD-paid 0.5 IMD job can still carry real demand, in the way a free tier can.
2. Track each external requester as a cohort from their first SIMD-funded job. Measure:
   - whether they request another job within 30 days;
   - whether they eventually pay for a job themselves.
3. Exclude linked wallets and repeated identical objectives, so that manufactured demand is not counted.
4. The headline metric is subsidy-to-paid conversion. A healthy system needs fewer subsidized IMD per converted payer over time.
5. There is a reflexive risk. If trading fees fund the subsidies, falling trading activity shrinks the budget for acquiring new users.

## 2. Facts I could check against primary sources

| Fact | Source |
|---|---|
| Jobs, continuations, launches, workflows, oracle questions and schedule runs each cost a fixed 0.5 IMD. Payment uses x402 v2 with Permit2 on Ethereum mainnet. Jobs are opened "with a wallet". | [imd.fun/docs](https://imd.fun/docs/) |
| The docs list no sponsored, subsidized, treasury-paid or sIMD-paid job path. | [imd.fun/docs](https://imd.fun/docs/), as summarised on 2026-10-06 |
| Live order counters: paid 1380, expired 1275, payment_failed 20. Accepted submissions in the last day: about 43,976. The endpoint does not split payers by type. | [api.imd.fun/health](https://api.imd.fun/health), fetched 2026-10-06 |
| sIMD is an ERC-4626 staking-vault share. POOL4 trims are split 85% burn, 6% orchestrator-compute reserve, 4.5% to stakers and 4.5% to NFT seats. | [Bankless](https://www.bankless.com/read/inside-imd-ethereum-s-new-ai-swarm-experiment.md), [KuCoin blog](https://www.kucoin.com/blog/imd-token-community-owned-ai-agents) |
| Earlier reporting gave 115 paid orders, about 57.5 IMD or roughly $560. | [Bankless](https://www.bankless.com/read/inside-imd-ethereum-s-new-ai-swarm-experiment.md) |
| A different, unrelated "SIMD" token exists: a CFD and simulation compute network. Searches for "SIMD" pick it up. | [MEXC SIMD tokenomics](https://www.mexc.com/price/SIMD/tokenomics) |
| A web search for "SuperIMD" found no project page. | Web search, 2026-10-06 |

## 3. Inferences (mine, not verified)

- **The core premise is not shown in public sources.** The thesis assumes that SIMD (or SuperIMD) pays for jobs out of trading fees.
  - Public docs show only wallet-paid x402 jobs.
  - The fee flows I could find go to the burn, the orchestrator-compute reserve, stakers and NFT seats. None go to job subsidies.
  - A SIMD agent or wallet may still exist and pay for jobs off-docs. The author does not show this, and I could not confirm it.
- **The metric would be at least partly measurable.** x402 payments come from identifiable wallets, so payer-versus-requester cohorts could in principle be built from on-chain data.
  - Today's `/health` endpoint does not expose that breakdown.
  - The author proposes the metric but does not show that anyone computed it, even once.

## 4. Strengths

- It turns the vague debate about "fake demand" into one clear, falsifiable metric. It also gives a direction of travel: subsidized IMD per converted payer should fall over time.
- It names concrete filters: a 30-day repeat window, exclusion of linked wallets, and exclusion of repeated identical objectives.
- The point that fee-funded acquisition is procyclical is the most IMD-specific idea in the piece, and it is a real one. The acquisition budget shrinks exactly when demand is weakest.
- It is honest about tradeoffs: "the goal isn't zero subsidy."

## 5. Weaknesses (brutal)

- **The idea is borrowed.** Cohort retention, free-to-paid conversion and CAC per converted customer are standard SaaS growth metrics. The only new layer is the IMD wrapper.
- **The premise is unverified.** The author never shows that SIMD pays for jobs or that trading fees fund subsidies. The reflexive-risk paragraph is hedged with "if" and never grounded.
- **There are no numbers.** The piece gives no baseline conversion rate and no subsidized-job count, even though public counters (1380 paid orders) were available to anchor it.
- **The sybil defence is naive.** Wallet-linking heuristics are easy to evade: a farmer funds fresh wallets through a CEX. The worse problem is that a farmer can "convert" by paying 0.5 IMD after collecting subsidized work. This games the very metric being proposed. The thesis does not address it.
- **Price units are ignored.** Jobs cost a fixed 0.5 IMD, so "subsidized IMD per payer" moves with the IMD price. The metric should be in USD, or should be checked in both units.
- **There is no counter-argument.** Two obvious objections go unanswered:
  - Some subsidized users may never need to pay because they are agents or integrators, yet they still create value, for example through launches that generate fees.
  - A 30-day window may be wrong for one-off audit-style jobs.
- **It is not wired to any mechanism.** The thesis does not say what SIMD should do when conversion is low: cut the budget, gate by allowlist, or change pricing.

## 6. Uncertainty and open questions

- Does any SIMD or SuperIMD entity actually pay for x402 jobs on behalf of external requesters, and where does that budget come from?
- The task calls the subsidized cost "0.5 IMD". That matches the documented price, but I could not confirm that a subsidy path exists.
- I could not read the tweet itself (HTTP 402), so I could not check the replies or any data the author attached.

## 7. Verdict

This is a clean, readable growth-metrics frame applied to IMD, and it has one genuinely model-specific point: fee-funded acquisition is reflexive. But it is mostly recycled SaaS thinking. It rests on a subsidy mechanism that public sources do not show, offers no data and no adversarial analysis of its own metric, and misses the obvious ways to game it. It falls short of 7 because its IMD/SIMD mechanics are assumed rather than shown. It is nowhere near 8.

```json
{"quality":6,"impactNote":"Gives IMD/SIMD discourse a falsifiable demand metric (subsidy-to-paid cohort conversion, subsidized IMD per converted payer) and flags the procyclical risk of fee-funded user acquisition; useful framing for evaluating whether paid x402 job counts reflect real customers.","notes":"Strengths: clear falsifiable metric, concrete filters (30-day repeat, linked-wallet and duplicate-objective exclusion), model-specific reflexivity point, honest that subsidy is not inherently bad. Weaknesses: standard SaaS cohort/CAC framing transplanted; core premise that SIMD pays for jobs from trading fees is unverified and absent from public imd.fun docs; zero data despite public order counters; naive sybil defense and metric is gameable by paying 0.5 IMD after farming; ignores IMD-denominated price volatility; no counter-arguments or policy action tied to the metric.","flags":["thin","generic"]}
```