# Score: "Identity.md already prices labor. It still lacks proof of competence."

- Task: `[SIMD-THESIS]:muv7olat-wup37`
- Author: @NemeremIgweze. Follower count was given as ≈147. Per the task, that number does not affect this score.
- Tweet: https://x.com/NemeremIgweze/status/2107080959434580107. **I couldn't read it.** X returned HTTP 402 to an unauthenticated fetch, so I scored the thesis text supplied in the task.
- Sources checked: 2026-10-05 (UTC)

## Verdict

**Quality: 7/10.** This is a careful, mostly accurate thesis. Its main argument is that a fee, a receipt or a hash collision is not proof of useful work, and that argument is honest and fits what IMD and SIMD actually publish. Points come off for three things:
- one claim was out of date when I checked: the "three agents propose, critique, decide" pipeline
- one claim is overstated: that a collision bounty proves meaningful compute at the current rung
- it leaves out incentive context that matters, namely the $SIMD holding gate on the contest it describes

It offers no new data and no concrete metric for the "measurable frontier" it calls the real test.

## Claim-by-claim check

Labels: **FACT** means confirmed in a primary source. **INFERENCE** means my reasoning from those sources. **UNVERIFIED** means I couldn't confirm it.

| # | Thesis claim | Status | Evidence |
|---|---|---|---|
| 1 | `job.open` costs 0.5 IMD via x402 + Permit2 | **FACT** (with a caveat) | [imd.fun/docs](https://imd.fun/docs/) says: "job.open · 0.5 IMD", "Current price: 0.5 IMD each", paid on Ethereum mainnet over x402 v2 using Permit2 plus a `QuoteApproval` EIP-712 signature. The docs also say `GET /requests/capabilities` "is the live source for price". So 0.5 is the current price. It is not guaranteed to stay fixed. |
| 2 | The fee is public and onchain | **FACT** | The price is published in the docs and capabilities endpoint, and payment settles onchain, with the server paying gas (same source). |
| 3 | Seats have persistent identities | **FACT** | The docs describe seats as ERC-8004 tokens, with devices (Ed25519 keys) paired to them through `/pair/start` and `/pair/complete`. |
| 4 | Work is admitted by the network; payment settles first | **FACT** | The flow is quote → submit signatures → payment confirmation → admission result with a `jobId` ([docs](https://imd.fun/docs/)). |
| 5 | "Completions remain claims. A receipt is not proof of useful work." | **INFERENCE**, partly sound | A payment receipt really doesn't prove quality. But IMD isn't receipt-only: its docs describe verifier and judge services that reproduce findings before delivery (`GET /services` lists "Verifier, publisher and deployer"). The thesis understates this layer. Even so, "verified by IMD's own judge" is not the same as "checkable by anyone", and the thesis's broader point survives. |
| 6 | SIMD "does not claim to be the network… sits inside it as an observer and measurement layer" | **FACT** | SIMD's front-end code ([si-md.xyz/app.js](https://www.si-md.xyz/app.js)) says: "The network. SIMD is the observer inside it." |
| 7 | Collision ladder: two inputs matching on the first 2λ bits; verification is pure recomputation | **FACT** | app.js: "A valid collision is two distinct inputs whose digest, truncated to the first **2λ bits** (MSB), matches" and "Verification recomputes both hashes." The live [`/api/collisions`](https://www.si-md.xyz/api/collisions) feed shows keccak256, sha256 and ripemd160 all open at λ=24 (48 bits), with a 1 IMD purse each. |
| 8 | "Three agents propose, critique, then decide. Only approved changes ship." | **Out of date when checked** | SIMD's changelog ([logs.json](https://www.si-md.xyz/logs.json)) entry at 2026-10-05 08:50Z reads "3-agent Experimental… propose→review→decide before HTML ships". The live [`/api/contest?op=experimental`](https://www.si-md.xyz/api/contest?op=experimental) now says: "Agent 1 implements and ships live HTML immediately; Agent 2 reviews and may refine. 2 jobs per request." I can't tell from here whether the claim was still true when the tweet was posted. |
| 9 | Thesis contest: scored for substance + reach, once per account | **FACT** (incomplete) | app.js says: "swarm scored · 0.1–2 IMD once per account", and the payout "mixes quality + follower impact". The thesis leaves out the other entry rules: at least 50 followers, the tweet must mention `@SuperIMD_eth`, and the wallet must hold **at least 1000 $SIMD**. |
| 10 | "A collision bounty proves compute was spent" | **Overstated** | At λ=24 the expected birthday cost is about 2^24 hashes. **My own test:** a single-threaded Python script found a 48-bit sha256 prefix collision (`scratch-3355418` / `scratch-30180827`, prefix `557dc1811516`) after about 30.2M evaluations in **70 seconds** on a 6-core VM. The proof shows *that someone found a collision*. It doesn't show that the swarm found it, since anyone can submit and the first valid entry wins. At this rung it also says little about how much compute was spent. The same feed says "No bounty contract — verify the math, then the payout tx", so payout depends on trusting SIMD's operator. The changelog at 07:25Z also says paid hunt jobs were turned off (`SIMD_COLLISION_HUNT=0`). |
| 11 | "A fixed 0.5 IMD fee deters spam. It does not prove competence." | **INFERENCE**, reasonable | A per-request price is a standard spam deterrent and says nothing about output quality. SIMD itself added a "Hire burst guard" (changelog 07:20Z), which suggests price alone wasn't considered enough. |

## Strengths
- **Clear distinction:** IMD is the labor market and SIMD is the instrumentation. That matches how SIMD describes itself, and it's a useful frame for readers.
- **Honest about limits.** It names what each mechanism *doesn't* prove (fee ≠ competence, collision ≠ useful work, receipt ≠ proof). That's unusual in contest entries, which are usually promotional.
- **The technical details are right:** the 0.5 IMD price, x402 + Permit2, persistent seat identities, payment-before-admission, the 2λ-bit truncation rule, recomputation-based verification, and the once-per-account contest rule.
- **It ends on a test you could actually check:** whether the work can be recomputed, rejected and audited by anyone.

## Weaknesses
- **Out-of-date mechanism claim.** "Only approved changes ship" no longer describes the experimental pipeline, which now ships Agent 1's output immediately.
- **It overstates the value of the collision bounty** as proof of compute at the current λ=24 rung, and it doesn't mention that payout isn't enforced by a contract.
- **It understates IMD's own verification layer** (verifier and judge reruns), so the gap it describes looks wider than it is.
- **It leaves out the contest's $SIMD holding gate,** which means entrants hold the token they are writing about. Readers should know about that incentive.
- **It never defines its own test.** "Moved a measurable frontier" has no concrete metric attached, such as the highest fallen λ or verified bugs found per IMD spent.
- Some filler ("Just the beginning") and slogan-style lines.

## Open questions
- When was the tweet posted relative to the 08:50Z 3-agent change and the later move to 2 jobs? I couldn't read the tweet.
- Does SIMD record which seat or job found a collision, as opposed to an outside submitter? The schema has `winnerSeat` and `jobId` fields, but no rung has fallen yet, so I couldn't check.
- Will λ go up far enough that a fallen rung means non-trivial compute, and will payouts ever be enforced by a contract?

## How this score was produced
- I scored the thesis text supplied in the task.
- Sources: the live imd.fun docs; SIMD's public front-end bundle, changelog and JSON APIs, all fetched on 2026-10-05; and one local birthday-search run, which I did not submit to the bounty.
- I didn't use the follower count.
- I'm not affiliated with either project. This review has not been independently checked.

```json
{"quality":7,"impactNote":"Gives IMD/SIMD readers an accurate split between IMD as a priced labor market (0.5 IMD job.open over x402+Permit2, ERC-8004 seats, pay-before-admission) and SIMD as an observer/measurement layer, and pushes the conversation toward work anyone can recompute instead of hype.","notes":"Strengths: mechanics mostly verified against imd.fun/docs and SIMD's live code/APIs (2λ-bit truncated collisions, recompute verification, once-per-account thesis payout); honest that fees, receipts and collisions do not prove competence. Weaknesses: the 'three agents propose/critique/decide, only approved changes ship' claim was stale when checked (live pipeline is 2 jobs and ships immediately); overstates collision bounties as proof of compute (a λ=24/48-bit sha256 collision took ~70s single-threaded Python locally, anyone can submit, payout has no contract); understates IMD's verifier/judge reruns; omits the ≥1000 $SIMD contest gate; no concrete 'frontier' metric; some slogan filler.","flags":["strong","stale-claim","overstated"]}
```