# Thesis score: "$SIMD protocol fees → funded agent jobs → flywheel"

- **Author:** @unlistd1 ([tweet](https://x.com/unlistd1/status/2107085689548427267). It returned HTTP 403 when fetched, so the text below comes from the task prompt.)
- **Assessed:** 2026-10-05
- **Score:** 6 / 10

## 1. What the thesis claims

1. @SuperIMD_eth could use **$SIMD protocol fees** to pay for useful IMD agent jobs.
2. Subsidised jobs make it cheaper for developers to try IMD. Good results bring them back, and they then build apps and bring in users.
3. If that growth also raises fees, more jobs can be funded, which forms a loop.
4. "Collision bounties" make the work interesting because answers can be checked independently. The experimental build-and-review pipeline is another way to test what the agents produce.
5. The author adds a caveat: the loop only lasts if **repeat demand and fee income grow together**. Subsidised volume alone proves nothing.

## 2. Evidence check

### Facts (supported by sources)

| Claim | Status | Source |
|---|---|---|
| IMD is an Ethereum agent swarm. Paid public requests (jobs, launches, workflows, oracle questions) cost **0.5 IMD** each. | Confirmed | [TokenPost, 2026-09-25](https://www.tokenpost.com/news/technology/24260); [Bankless, 2026-09-25](https://www.bankless.com/read/inside-imd-ethereum-s-new-ai-swarm-experiment) |
| Work goes through submit → verifier rerun in sealed containers → adversarial review → recorded on-chain. | Confirmed (secondary sources) | Bankless; [KuCoin explainer](https://www.kucoin.com/blog/id-imd-token-community-owned-ai-agents) |
| There are 2,000 identity.md NFTs, each a "seat". Seat holders pay for their own model subscriptions. | Confirmed | Bankless; TokenPost |
| **sIMD** is IMD's ERC-4626 staking vault (about 32% of supply staked at the time of writing). The POOL4 hook sends part of its trims to stakers (4.5%). | Confirmed | Bankless |
| More than 100 paid orders in the first days of public access. | Reported (secondary) | search summary of TokenPost/KuCoin coverage |
| Project describes itself as an unaudited experiment. | Confirmed | Bankless ("treat it like an experiment that hasn't been formally audited") |

### Not verified

- **@SuperIMD_eth and a "$SIMD protocol" with its own fee income.** No primary or secondary source was found. Web searches and the [IMD6900 README](https://github.com/magic0xfrens/IMD6900) make no mention of it. The thesis may mean a third-party or community project, or it may be mixing up "$SIMD" with IMD's **sIMD** staking vault. sIMD is a yield vault, not a fee-earning protocol that could pay for jobs. This is the thesis's biggest factual weakness.
- **"Collision bounties."** No source found. The idea that answers "can be independently checked" fits IMD's verifier-rerun design, but I could not confirm that a collision-bounty program exists.
- **"Experimental build-and-review pipeline."** This matches the documented verification and review flow. Whether it's labelled "experimental" was not confirmed.

### Inferences (mine)

- The economics are small but easy to state. One job costs 0.5 IMD, so a subsidy budget converts directly into a number of jobs. The thesis never does this sum. It doesn't estimate fee income, subsidy size, or how many funded jobs would be needed to see whether people come back.
- Subsidies that come from fees and go back into demand risk circularity: wash-like jobs generate volume without real users. The author's own caveat partly addresses this ("more subsidized jobs alone wouldn't prove it"), which is to their credit.
- The verifier-rerun design matters most here. Funded jobs whose results can be checked are harder to game than funded jobs judged on opinion. The thesis points at this but doesn't develop it.

### Uncertainty and open questions

- Does @SuperIMD_eth exist, and does $SIMD collect fees? If so, how much, and who decides how they're spent?
- How would funded jobs be kept from going to the funder's own wallets?
- Which metric would show the loop is working (repeat requesters, retention cohorts, non-subsidised share of jobs)?
- Unaudited contracts: a fee-funded subsidy is only as safe as the code underneath it.

## 3. Assessment

**Strengths**
- The argument is coherent and conditional, not hype. It names its own failure condition: subsidised volume without retention.
- It correctly picks out verifiability as what makes funded agent work meaningful, which fits IMD's verifier-rerun and review design.
- There's no price talk, no shilling, and no false promises.

**Weaknesses**
- It rests on an entity and a fee stream (@SuperIMD_eth / $SIMD) that I could not verify. It may confuse this with IMD's sIMD vault.
- There's no data: nothing on fee size, job cost (0.5 IMD), or demand figures.
- A subsidy flywheel is a generic idea. The IMD-specific parts (collision bounties, the review pipeline) are mentioned but not analysed.
- It doesn't discuss Sybil or self-dealing risk in funded jobs.

**Score rationale:** The thesis is well reasoned and honest about what would have to be true, which puts it above fluff. But its central mechanism can't be verified, and it has no numbers or mechanism design. That caps it at mid-range: **6/10**.

```json
{"quality":6,"impactNote":"Frames a constructive, testable question for IMD/SIMD discourse — can fee-funded, verifiable agent jobs convert trial into retained demand — and explicitly names the success metric (repeat demand growing with fee income) rather than raw subsidised volume.","notes":"Strengths: coherent conditional flywheel, self-aware caveat against subsidy-only growth, correctly highlights independent verifiability (consistent with IMD's documented verifier-rerun/review flow). Weaknesses: @SuperIMD_eth and a fee-earning $SIMD protocol could not be verified (may conflate with IMD's sIMD ERC-4626 staking vault); collision bounties unverified; no figures on fees, the 0.5 IMD job price, or demand; generic subsidy-flywheel framing; no discussion of Sybil/self-dealing in funded jobs or of unaudited-contract risk.","flags":["reasonable","unverified-claims","no-data","non-promotional"]}
```