# HARD GRADE — "Three signals" thesis on Identity.md (IMD) / SIMD

**Target:** [x.com/chidifinance\_/status/2107382993950224739](https://x.com/chidifinance_/status/2107382993950224739)
**Author:** @chidifinance\_ (followers ≈307 — not an input to the score)
**Graded:** 2026-10-06 · **Verdict: quality 7/10 — strong draft, below the pay bar of 8**

---

## 1. The question

Does this thesis earn a pay-grade score? That requires (a) named mechanisms, (b) tradeoffs, (c) IMD/SIMD-specific claims for ≥7, plus (d) originality and depth for ≥8. I graded by verifying every factual claim against primary and contemporaneous sources, then testing whether the argument's *mechanics* hold up — not whether it sounds sophisticated.

---

## 2. Fact-check of the thesis's claims

### 2.1 The September snapshot — VERIFIED EXACTLY

The thesis cites "100 jobs, 91 completed, 2,669 accepted submissions in 24 hours, and 38 paid orders." I traced this to a public writeup dated 24 Sep 2026 ([mtezy/imd-writeup](https://github.com/mtezy/imd-writeup)), which states:

| Claim in thesis | Source wording | Match |
|---|---|---|
| 100 jobs | "100 jobs (91 completed, 5 executing, 4 cancelled)" | ✅ |
| 91 completed | as above | ✅ |
| 2,669 accepted submissions / 24h | "2,669 submissions accepted / 24h" | ✅ |
| 38 paid orders | "38 paid" | ✅ |

**This is a real, attributable, unembellished citation.** No rounding, no inflation, no invented figures. That is more than most IMD commentary manages, and it is the single strongest thing about the post.

### 2.2 POOL4 split — DIRECTIONALLY RIGHT, MATERIALLY INCOMPLETE

The thesis: "85% of each retired batch to burn and 15% to rewards, with 6% reserved for the lead orchestrator and 4.5% for NFT inference nodes."

The actual split, consistent across three independent sources ([Bankless](https://www.bankless.com/read/inside-imd-ethereum-s-new-ai-swarm-experiment), [KuCoin](https://www.kucoin.com/blog/imd-token-community-owned-ai-agents), [CoinCodex](https://coincodex.com/crypto/identity-md/)):

- **85%** burned
- **6%** orchestrator compute reserve
- **4.5%** to **sIMD stakers**
- **4.5%** to NFT seats / nodes

The 85/15 top line is correct. But the thesis names only two of three sub-components and **silently drops the 4.5% to stakers** — then labels the entire 15% as "capital being directed toward future or existing infrastructure."

That label is wrong for roughly **30% of the bucket it describes**. Staker yield is a distribution to passive token holders; it is not infrastructure funding, not compute, and not "productive" under the thesis's own definition. The thesis built a three-signal taxonomy and then mis-sorted a third of one signal.

### 2.3 "Bonding market and NFT node program not yet live" — SUPPORTED, BUT THE PREMISE IS DECAYING

Supported. Public sources describe bonding as forward-looking: *"Bonding will open when IMD reaches $4, and the Bonding Reserve can eventually be converted into ETH to pay for AI Inference."* Reporting on 25 Sep 2026 ([TokenPost](https://www.tokenpost.com/news/technology/24260)) likewise found "no recurring production work for outside protocols." NFT **inference nodes** (a funded compute program) are distinct from NFT **seats** (live: 270 enrolled on 24 Sep, ~380 by 25 Sep), and the thesis uses the right term.

**But the thesis does not notice that its own premise is expiring.** The stated bonding trigger is $4. On 6 Oct 2026 IMD traded in the **~$11–$13** range depending on tracker (**$11.07** on [CoinCodex](https://coincodex.com/crypto/identity-md/); **≈$12.80** converted from BRL on [OKX](https://www.okx.com/en-br/price/identity-md-imd), which also reported +28% in 24h), against a reported all-time high of **$14.51 on 5 Oct**. Trackers disagree on the exact level and the token is moving fast, but **every reading is roughly 3× or more above the activation threshold.** A thesis whose load-bearing claim is "reserves cannot honestly be counted as productive compute *today*" should have flagged that the gate it relies on has been numerically cleared for weeks. Either bonding has activated and the claim is stale, or the $4 figure is obsolete documentation — and the thesis engages neither branch.

### 2.4 Omission: the 86 expired orders

The *same snapshot* the thesis cites reports, alongside "38 paid": **"7 failed (`payment_permission_expired`)", "86 expired", "1 quoted."**

So paid conversion was **38 of ~132 orders (≈29%)**. For a thesis whose entire subject is measuring paid demand, the expired-order count is the most informative adjacent number in the dataset, and it is dropped. It cuts both ways — 86 expirations could be *latent* demand lost to permit-expiry UX rather than *absent* demand — which makes the omission worse, not better: it removes the one datapoint that could distinguish "nobody wants this" from "the payment rail leaks."

### 2.5 Unresolved conflict the thesis inherits

Bankless, dated **25 Sep** (one day after the thesis's snapshot), reports **115 paid orders via x402** and "~57.5 IMD (~$560)" in total x402 revenue. 115 × 0.5 IMD = 57.5 IMD, so that figure is internally consistent. It cannot be reconciled with "38 paid" one day earlier.

The likely explanation is **differing scope** — the API prices `job.open`, `launch.open`, `workflow.open`, `oracle.request` and `schedule.create` all at 0.5 IMD ([imd.fun/docs](https://imd.fun/docs/)), so 115 may count all paid actions while 38 counts jobs only. **I could not confirm this.** The thesis presents 38 as settled fact without acknowledging that the public record disagrees by ~3× on the metric most central to its argument.

---

## 3. Where the argument is genuinely good

**The taxonomy is correct and non-obvious on a dashboard.** Paid x402 orders, accepted submissions, and reserve inflows really do render as adjacent counters that invite conflation, and they really do answer different questions. The thesis states this cleanly and without hedging.

**The refusal to count reserve growth as productive compute is honest.** Most token commentary does the opposite. This is real analytical discipline and it is the main reason this post is not a 5.

**The proposed measurement program is actually implementable** — which I checked rather than assumed. The thesis asks to track paid requests, repeat customers, and "the execution attached to each funding source." The public API supports exactly this: `GET /requests/paid-by/:address` returns per-wallet order history, `GET /jobs/:id` exposes a `paidBy` field ("the wallet that paid (null when nobody did)"), and `GET /seats/records` gives per-seat accept/reject counts ([imd.fun/docs](https://imd.fun/docs/)). An actionable proposal that survives a feasibility check is worth real credit. The thesis does not cite these endpoints or show it verified feasibility — it got lucky rather than rigorous — but the proposal stands.

**No hype, no price target, no padding.** Tight for its length.

---

## 4. Where it fails to reach pay grade

### 4.1 It misses that the reserve is funded by sell pressure, not by work

This is the big one. POOL4 is a protocol-owned ETH/IMD Uniswap V4 pool governed by a `CappedBurnHook`: the hook caps pool IMD inventory, and **when sells push inventory above the cap, the excess is trimmed after the swap**, with the cap ratcheting down ~1,000 IMD/day.

Therefore "protocol-directed funding" is **a function of trading volume and sell pressure — not of job demand, revenue, or compute need.** The thesis's third signal is, mechanically, a *speculation* signal wearing an infrastructure label.

This breaks the taxonomy the thesis is built on. The consequences it never draws:

- The reserve is **pro-cyclical and reflexive**: it fills fastest when speculative churn is highest, and starves in quiet markets — precisely when subsidy would be most needed to keep work flowing.
- "Reserve-dependent activity" versus "paid demand" is therefore not demand-vs-subsidy. It is **demand-vs-trading-volume**, which is a different and considerably less flattering comparison.

A thesis analyzing POOL4 that never mentions the mechanism that funds POOL4 is not doing pay-grade mechanism work.

### 4.2 It is addressed to SIMD without engaging what SIMD is for

The post opens by telling @SuperIMD\_eth to separate these signals. But SIMD's own site states its purpose directly: *"protocol fees are collected for one use: covering Identity.md job costs in full, so the work is paid by the protocol"* ([si-md.xyz](https://www.si-md.xyz/), [superimdc.xyz](https://www.superimdc.xyz/)).

SIMD is not accidentally blurring paid demand and protocol funding. **Paying for the work is its product.** This inverts the thesis's framing and raises a question far sharper than anything the thesis asks:

> If the protocol covers job costs in full, why would any customer ever pay 0.5 IMD via x402? The "independently paid demand" ratio is **endogenous to the subsidy** — SIMD's design mechanically drives the thesis's headline metric toward zero.

That is the genuinely original insight available in this material, it is sitting in one sentence on the front page of the project the post is addressed to, and the thesis walks straight past it. Its falsifiable test may be structurally unable to return a positive result for reasons that have nothing to do with customer willingness to pay.

### 4.3 The falsifiable test has no price control

Paid requests cost **0.5 IMD** — denominated in the token, not in dollars. At ~$9.73 (25 Sep) that is ≈$4.87; at the ~$11–$13 range seen on 6 Oct, ≈$5.50–$6.50. **The USD cost of buying machine labor rises with the token price.** A token that has roughly tripled is a rising price wall against the exact demand the thesis wants to measure. Flat paid demand amid appreciation is therefore ambiguous between "customers don't want this" and "it got more expensive," and the test as written cannot separate them.

### 4.4 Falsifiable in form, under-specified in practice

"Measure whether paid requests and repeat usage rise faster than reserve-dependent activity, while accepted output remains stable or improves" names **no threshold, no observation window, and no decision rule.** Rise faster by how much, over what period, judged when?

And the scale problem is fatal to the test at present: the denominator is **38 paid orders and ~$560 of lifetime x402 revenue**. At n=38, a handful of orders swamps any trend. The thesis proposes an elaborate ratio test on a base where noise dominates signal, and never says so. Directionally falsifiable — yes. Operationally testable today — no.

### 4.5 The framework is competent, not novel

Separating organic demand from incentivized activity is a standard crypto-analysis move (mercenary TVL, wash-traded volume, emissions-driven usage). Applying it to IMD is a correct and useful port. It is not new thinking, and the thesis does not advance the method. The rubric reserves 8+ for novel synthesis; this is skilled application of an existing lens, minus the two mechanisms that would have made it original.

### 4.6 Structural thinness

Five paragraphs, linear, no counter-argument, no worked numbers, no model. The thesis never argues against itself — e.g. that protocol-subsidized bootstrapping is a *defensible* strategy rather than a weakness, which is the obvious rebuttal from the SIMD side. Pay-grade work anticipates its strongest objection.

---

## 5. Facts / inferences / uncertainty / open questions

### Established facts (attributable)
- Snapshot figures 100/91/2,669/38 are accurate as of 24 Sep 2026 — [mtezy/imd-writeup](https://github.com/mtezy/imd-writeup).
- That same snapshot also records 7 failed, 86 expired, 1 quoted order; 263 agents online; 270 seats enrolled.
- POOL4 trim split is 85% burn / 6% orchestrator compute / 4.5% stakers / 4.5% NFT seats — [Bankless](https://www.bankless.com/read/inside-imd-ethereum-s-new-ai-swarm-experiment), [KuCoin](https://www.kucoin.com/blog/imd-token-community-owned-ai-agents).
- POOL4 is a protocol-owned Uniswap V4 ETH/IMD pool with a `CappedBurnHook`; trims are triggered by sells pushing inventory above a cap that declines ~1,000 IMD/day.
- Paid requests cost 0.5 IMD across `job.open`, `launch.open`, `workflow.open`, `oracle.request`, `schedule.create` — [imd.fun/docs](https://imd.fun/docs/).
- The API exposes `paidBy` per job and `GET /requests/paid-by/:address` per-wallet order history.
- SIMD states protocol fees exist to cover Identity.md job costs in full — [si-md.xyz](https://www.si-md.xyz/).
- Bonding is described as opening at IMD = $4. On 6 Oct 2026 IMD traded ~$11–$13 across trackers ($11.07 [CoinCodex](https://coincodex.com/crypto/identity-md/); ≈$12.80 from BRL on [OKX](https://www.okx.com/en-br/price/identity-md-imd), +28% in 24h), with a reported ATH of $14.51 on 5 Oct. Trackers disagree on the level; all readings are ≥3× the $4 trigger.
- Late-Sept comparison point: 43,800+ total accepted submissions, ~29,600 in 24h, ~380 seats, 115 paid x402 orders (~$560) — [Bankless](https://www.bankless.com/read/inside-imd-ethereum-s-new-ai-swarm-experiment), 25 Sep 2026.
- Live explorer, 6 Oct 2026: 1,328 jobs posted, 1,239 completed, 83 incomplete, 665 agents online, 43.9K steps/24h — [explorer.imd.fun](https://explorer.imd.fun/).

### My inferences (reasoned, not sourced)
- The 4.5% staker allocation is holder yield, not infrastructure funding, so the thesis's "15% = capital toward infrastructure" framing is ~30% mislabeled.
- Because trims are sell-driven, reserve growth proxies trading volume and is pro-cyclical — the thesis's third signal is partly a speculation signal.
- SIMD's full-subsidy design makes "independently paid demand" endogenous to the subsidy, potentially suppressing the thesis's metric by construction.
- IMD-denominated pricing confounds the proposed test by raising USD cost as the token appreciates.
- Between 24 Sep and 6 Oct, jobs rose ~100→1,328 (~13×) while completion share held (~91%→~93%) — execution scaled without obvious quality loss. This is weak, partial evidence *for* the thesis's first branch (execution expanding), though funding-source attribution is not public enough to run the full test.

### Uncertainty
- **38 vs 115 paid orders** one day apart is unreconciled. Scope difference (jobs-only vs all paid actions) is my best explanation; unconfirmed.
- **Whether bonding is live today.** The $4 trigger has been cleared by ~3× or more, yet sources still describe bonding prospectively. I could not find an authoritative current status page.
- **Supply figures disagree:** 7.1M (Bankless, 25 Sep) vs 4.1M / 3.88M (Oct 6 trackers). A ~3M burn in 11 days is possible under aggressive trimming but I could not verify it.
- **The tweet itself returned HTTP 402 and could not be fetched.** I graded the thesis text as supplied in the assignment; I could not independently confirm wording, thread length, or that no additional context exists in replies.

### Unanswered questions
- Is reserve-funded work separable from customer-funded work in public data today? `paidBy` is per-job, but no endpoint surfaced an aggregate reserve-vs-customer execution split.
- What share of the 86 expired orders represented genuine intent lost to permit-expiry UX?
- Has any orchestrator compute reserve actually been spent on inference, or is it only accumulating?

---

## 6. Score rationale

**Clears ≥7:** named mechanisms (POOL4, CappedBurnHook-adjacent trim split, x402, bonding reserve, NFT inference nodes vs seats); a real tradeoff ("the protocol can fund work" vs "customers fund the work"); IMD/SIMD-specific throughout; verified-exact factual citation; genuine technical honesty in refusing to score reserves as compute; and a measurement program that survives a feasibility check.

**Blocked from 8:** originality is modest — a standard organic-vs-incentivized lens, ported competently. More damningly, it misses the two mechanisms most central to its own subject: POOL4's reserve is **sell-pressure funded** (so its third signal is partly speculation), and SIMD **pays job costs in full by design** (so its headline metric is endogenous to the subsidy it critiques). It also mislabels 4.5% of the split, drops the 86 expired orders from its own source, omits a price control, and sets no thresholds on a base of n=38. Pay grade requires catching the mechanism that undercuts your own framing. This post doesn't.

**7/10. Strong draft, not pay-grade.**

---

## Sources

- [mtezy/imd-writeup — 24 Sep 2026 snapshot](https://github.com/mtezy/imd-writeup)
- [Bankless — Inside IMD, Ethereum's New AI Swarm Experiment (25 Sep 2026)](https://www.bankless.com/read/inside-imd-ethereum-s-new-ai-swarm-experiment)
- [TokenPost — Ethereum's IMD Tests AI-Agent Network (25 Sep 2026)](https://www.tokenpost.com/news/technology/24260)
- [KuCoin — What Is IMD Token?](https://www.kucoin.com/blog/imd-token-community-owned-ai-agents)
- [IMD API docs](https://imd.fun/docs/) · [IMD Explorer](https://explorer.imd.fun/)
- [SIMD — si-md.xyz](https://www.si-md.xyz/) · [superimdc.xyz](https://www.superimdc.xyz/)
- [CoinCodex — Identity.md price](https://coincodex.com/crypto/identity-md/) · [OKX — Identity.md price](https://www.okx.com/en-br/price/identity-md-imd)
- Graded tweet (unfetchable, HTTP 402): [x.com/chidifinance\_/status/2107382993950224739](https://x.com/chidifinance_/status/2107382993950224739)

```json
{"quality":7,"impactNote":"Supplies IMD/SIMD discourse with a correct and portable taxonomy - paid x402 demand vs execution throughput vs POOL4 reserve inflows are three different questions that render as adjacent dashboard counters. Its refusal to count reserve growth as productive compute while the bonding market and NFT inference-node program remain pre-live is the kind of technical honesty the IMD conversation is short on, and the five metrics it proposes are genuinely implementable against the public API (paidBy per job, /requests/paid-by/:address, /seats/records), so it can be run rather than just admired.","notes":"STRENGTHS: snapshot figures (100 jobs, 91 completed, 2,669 accepted/24h, 38 paid) verified exactly against mtezy/imd-writeup dated 24 Sep 2026 - accurate, attributable, un-inflated. Names real mechanisms (POOL4 trim, 6% orchestrator reserve, 4.5% NFT nodes, x402, bonding reserve), draws a real tradeoff between 'protocol can fund work' and 'customers fund work', correctly distinguishes NFT inference nodes from live NFT seats, and carries no price hype or padding. WEAKNESSES: (1) misses that POOL4 trims are triggered by sell pressure against a declining inventory cap, so its 'protocol-directed funding' signal is mechanically a trading-volume proxy - pro-cyclical, filling during speculation and starving in quiet markets - which breaks its own taxonomy; (2) addressed at @SuperIMD_eth without engaging SIMD's stated purpose of covering Identity.md job costs in full, making 'independently paid demand' endogenous to the subsidy and possibly driven toward zero by design - the sharpest point in the material, left untouched; (3) omits the 4.5% staker allocation and so mislabels ~30% of the 15% bucket as infrastructure capital when it is holder yield; (4) drops 86 expired and 7 failed orders from the very snapshot it cites (38/132 paid, ~29% conversion), discarding the one datapoint separating absent demand from a leaking payment rail; (5) no price control - requests cost 0.5 IMD, so USD cost rises with a token that roughly tripled, confounding the test; (6) its 'not yet live' premise is decaying unexamined, as the stated $4 bonding trigger has been cleared by ~3x or more (IMD ~$11-13 across trackers on 6 Oct 2026); (7) no thresholds, window, or decision rule, on a base of n=38 orders and ~$560 lifetime revenue where noise dominates; (8) framework is a competent port of the standard organic-vs-incentivized lens, not novel. UNCERTAINTY: 38 vs 115 paid orders one day apart is unreconciled (scope difference suspected, unconfirmed); current bonding status unverifiable; supply figures conflict (7.1M vs 4.1M); the tweet itself returned HTTP 402 and could not be independently fetched, so grading relied on the supplied text.","flags":["strong"]}
```