# Score: "Identity.md is the labor market; SIMD is the measurement layer"

Tweet: https://x.com/DrakosUnchained/status/2107036292076757067 (author @DrakosUnchained)
Research date: 2026-10-05

## Verdict (JSON)

```json
{
  "quality": 7,
  "impactNote": "Gives IMD/SIMD discourse a clear, testable frame: IMD prices and admits labor (0.5 IMD per job.open, paid over x402 + Permit2), and SIMD's job is to turn that labor into recomputable evidence. It moves the conversation from token hype toward verifiability, which is the network's own stated selling point (sealed-container verifier, onchain-scored agents). The IMD mechanics it cites match the public docs. The SIMD-specific mechanics could not be confirmed from public sources.",
  "notes": "Strengths: the IMD facts are accurate and specific (fixed 0.5 IMD per job, x402 v2 with Permit2, persistent seat identities). It makes a real argument: priced public labor is an honesty mechanism, and completions are not proof of competence. The collision-ladder description is technically coherent: matching 2λ bits costs about 2^λ hashes by the birthday bound, and checking a claim only requires recomputing hashes. The positioning is honest that SIMD 'does not pretend to be the network'. Weaknesses: none of the SIMD claims (collision ladder, three-agent propose/critique/decide site, X-gated thesis contest scored for substance plus reach) could be checked against a public primary source. The essay is promotional and never says how SIMD would measure that the swarm 'moved a measurable frontier'. A hash-collision bounty proves the agents spent compute, not that they are competent at useful work, and the essay doesn't address that gap. Calling the 0.5 IMD price an 'honesty mechanism' is asserted rather than shown: a fixed fee deters spam but does not by itself make work auditable. Minor: 'http://job.open' is an auto-link artifact, and the name 'SIMD' collides with IMD's existing sIMD staking share (ERC-4626), which can confuse readers.",
  "flags": []
}
```

Flags: the thesis isn't spam, is on-topic, and isn't low-effort. I didn't mark it "strong" because its distinctive claims (the SIMD ones) are unverified and the measurement argument is never made concrete.

## Evidence: facts checked against sources

| Thesis claim | Status | Source |
|---|---|---|
| `job.open` costs a fixed 0.5 IMD | **Confirmed** | IMD docs: "Open a job…", 0.5 IMD per job, `500000000000000000` atomic units ([imd.fun/docs](https://imd.fun/docs/)). Also Bankless: "Anyone can pay 0.5 IMD per request to open a job settled via x402" ([Bankless, 2026-09-25](https://www.bankless.com/read/inside-imd-ethereum-s-new-ai-swarm-experiment)) |
| Paid via x402 + Permit2 | **Confirmed** | IMD docs: x402 v2 envelope plus "the Permit2 payment and an EIP-712 approval of that payment for this quote" ([imd.fun/docs](https://imd.fun/docs/)). x402 is an open payment standard under the x402 Foundation / Linux Foundation ([x402.org](https://www.x402.org)). Permit2 is Uniswap's unification of `SignatureTransfer` and `AllowanceTransfer` ([Uniswap docs](http://developers.uniswap.org/llms.mdx/docs/protocols/permit2/overview)) |
| Agents hold persistent identities / seats | **Confirmed** | 2,000 identity.md NFTs, one seat each, registered under ERC-8004 ([KuCoin, 2026-09-29](https://www.kucoin.com/blog/imd-token-community-owned-ai-agents)). The explorer lists agents as "identity.md seats on machines, scored onchain for the work they get accepted" ([explorer.imd.fun/agents](https://explorer.imd.fun/agents)) |
| Work is admitted through the network and verified | **Confirmed (IMD side)** | "a verifier rebuilds each submission in a sealed-off container…" ([Bankless](https://www.bankless.com/read/inside-imd-ethereum-s-new-ai-swarm-experiment)). Job states include pending/accepted/rejected/completed ([imd.fun/docs](https://imd.fun/docs/)) |
| SIMD = "Superintelligent Identity.md", hires through IMD rails | **Unverified** | No public primary source found. Searches for the name and mechanics returned only unrelated results (e.g., the SIMD SHA-3 candidate hash) |
| Collision ladder: distinct inputs that match on the first 2λ bits | **Unverified (existence)**; **sound (design)** | No public spec found. Analysis is in the next section |
| Three-agent propose/critique/decide site | **Unverified** | No public source found |
| Thesis contest gated by X accounts, once per account, scored for substance plus reach | **Partially corroborated by this task only** | This scoring task's prompt matches that description (it separates impact from quality and supplies the follower count). No public rules page was found |

## Inferences (my analysis, not sourced facts)

- **Collision ladder soundness.** By the birthday bound, finding two distinct inputs that agree on the first 2λ bits takes about 2^λ hash evaluations, so treating λ as the difficulty setting makes sense. Checking a claimed pair takes two hash evaluations, so "verification is recomputation, not reputation" is accurate for this kind of puzzle. The caveat is that the puzzle measures compute and search effort, not the useful, domain-specific competence the thesis says agent stacks are missing.
- **"Price as honesty mechanism."** A fixed fee raises the cost of spamming jobs and leaves a public payment trail, as the 115 public x402 orders reported by Bankless show. Auditability comes from the verifier and the onchain scoring, though, not from the price. The thesis merges the two.
- **Name collision.** IMD already has `sIMD`, the StakedIMD ERC-4626 share ([KuCoin](https://www.kucoin.com/blog/imd-token-community-owned-ai-agents); [Bankless](https://www.bankless.com/read/inside-imd-ethereum-s-new-ai-swarm-experiment)). "SIMD" as a separate project name invites confusion. The thesis doesn't acknowledge this.
- **Scoring rationale.** The IMD claims are accurate, specific, and argued (+). The framing is original enough to be useful (+). The SIMD claims, which are the core of the piece, can't be checked publicly, and the "measurable frontier" claim has no metric attached (−). Overall that puts it above average (7), short of a strong 8+.

## Uncertainty

- I couldn't read the tweet itself: x.com returned HTTP 402 to an unauthenticated fetch. I scored the thesis text as given in the task and assumed it matches the tweet.
- The SIMD project pages (site, collision ladder spec, contest rules) may exist at locations that search engines don't index. Their absence from my results doesn't show they don't exist.
- Agent counts on the explorer change over time (609 agents listed on 2026-10-05). The thesis makes no numeric claims about them, so this doesn't affect the score.

## Unanswered questions

1. Where is SIMD's public spec for the collision ladder (hash function, input format, how λ is set, how payouts work)?
2. How does SIMD measure that "the swarm moved a measurable frontier"? Which metric, and against which baseline?
3. Does the three-agent revision pipeline publish its rejected revisions, so the claim that it "rejects weak updates" can be audited?
4. How is "reach" weighted against "substance" in the contest, and how is gaming by bought followers handled?