PROOF POSIT PP-003 — THIRD ISSUE (v3)

Subject — Robinhood Chain (chain 4663), Arbitrum Expansion Program remittance

Title — Every payment Robinhood Chain has made to the Arbitrum ecosystem, read from the chain

Date of this issue's text — 2026-09-29

Why a third issue. The second issue (JFD-20260920-PP-00001) was sealed as a one-page cover, whose seal fields could not carry their own outcome, and its sealed source text left its date, three identifiers and its seal table blank. This issue carries the full text of the second with every figure unchanged. The block hash and the method hashes are now stated from the Glass Hull ledger. The re-run hash is not stated, because no ledger record carries it. The second issue's seal and Bitcoin anchor are stated from the record at the end of this document.

Measurement window	30 April 2026 (block 19, contract deployment) to 7 September 2026 (block 57,033,100)
Pinned at	Block 55,927,945 for the sealed window · block hash 0x06caf60a2e7f1d592940c579fa77d7c86be00e375b005a8ea0ec200a5057f8b4 (Glass Hull ledger, kind-3 3121913) · the 7 September settlement read at block 57,033,100
Chain	4663 — Robinhood Chain, Arbitrum Orbit
Surface	archive for genesis → 6 September, both arms in agreement. The 7 September settlement was read from a public archive endpoint and is pending re-derivation on our own instrument. Stated here, not in a footnote.
Method hash	Settlement figures: e366edc8628237f9082c183f794c3ebde711a25dd10378eeeb2055a21995e842 (method PP-003-settlements-exact-v1, frozen 2026-09-16 00:30:44 UTC before computation; Glass Hull ledger, kind-3 2931182) · distributor population: method B ea38543f7c4f15e6a2ef6cec5b2f4e2a8134cff0f9c3c01cb5fa9c8b6314e320, amendment 37cd87c97e0ae619507fa478242b3e02ba23c90ff450d0ffac202192e5cf692c (two independent routes compared, CLEARED; Glass Hull ledger, kind-3 2999355)
Re-run hash	Not stated: no Glass Hull ledger record carries it. Every figure is reproducible with the method in The Full Proof.
Supersedes / Corrects	PP-003-v2 (JFD-20260920-PP-00001), the second issue, whose text this issue carries with every figure unchanged. The second issue superseded PP-003 (JFD-20260916-PP-00001) and corrected the fee-distributor population from "two" to the measured set of three current distributors within a nine-contract family (four contracts have ever been fee accounts), and the associated ownership and configuration counts. All settlement figures, the 9,000/1,000 split, and every total are unchanged and re-read to the wei.
Subject notice	The Arbitrum Foundation was given sight of this finding before publication. Robinhood was not contacted.
How to cite	Jet Fyul Dynamics LLC, Proof Posit PP-003-v3: Arbitrum Expansion Program remittance, chain 4663, 2026-09-29, pinned at block 55,927,945

---

I. SUMMARY

Robinhood Chain owes the Arbitrum ecosystem 10% of its net protocol revenue under the Arbitrum Expansion Program. The Arbitrum Foundation's own H1 report describes that income as unaudited. Nobody outside the operator had verified it.

We read every payment the chain has made, from the deployment of its fee contracts to 7 September 2026. There are nine settlement dates (carried from PP-003's sealed run record).

The split is exact. The chain's fee distributors are standard reward-distributor contracts, weighted 9,000 / 1,000 in basis points, and every settlement in the chain's history lands on that split to the basis point, with no exception. Three such contracts are the current fee distributors, configured in the chain's first blocks and never altered thereafter.

On Monday 7 September the chain settled 12,725.576405 ETH — 2.86 times everything it had remitted in the eight settlements before it, combined. The ecosystem's share of that single day was 1,272.557641 ETH.

Three things remain unexplained. Fees wait in the distributors between weekly splits, and the ecosystem's share waits again before routing onward to L1; both balances were substantial at our pinned height. And a single address has held the power to redirect the ecosystem's share from all three current distributors since 24 June 2026, before the chain opened to the public on 1 July.

II. METRIC OBJECTIVE

The question. Does the chain remit what the Expansion Program requires, and can that be established without relying on the operator's own reporting?

What we measured. Every settlement the chain's fee-distributor contracts have paid, across the full history of the chain — the settling contract, the receiving address, the exact value, and when — and from those, the realised split per settlement.

Why the question is open. The Arbitrum Foundation's H1 2026 update describes Expansion Program income as unaudited. Robinhood Chain is now the largest contributor to that line. No independent verifier is named anywhere in the programme's public materials.

Declared before computation. The method fixed, in advance, that the realised split would be reported whatever it turned out to be — including a result showing the split is honest. It is.

III. FINDINGS

Nine settlement dates.

Settled	Ecosystem share (ETH)	Realised split
11 Jul 2026	32.319	10.000%
20 Jul 2026	69.470	10.000%
27 Jul 2026	76.931	10.000%
3 Aug 2026	21.113	10.000%
11 Aug 2026	36.206	10.000%
17 Aug 2026	47.404	10.000%
24 Aug 2026	18.375	10.000%
31 Aug 2026	142.738	10.000%
7 Sep 2026	1,272.557641	10.000%

Cadence is weekly, Mondays at 18:01 UTC, unbroken from 20 July, with one departure: no Monday settlement on 10 August, and one on Tuesday 11 August instead.

The 7 September settlement, in full. 12,725.576405 ETH across the fee accounts — 11,453.018764767607188000 to the operator's recipient and 1,272.557640529734132000 to the ecosystem's, exactly nine to one, in wei. That single week is 2.86× the eight prior settlements combined.

The distributor family. Nine contracts on the chain carry identical fee-distributor code. Four of them have ever served as ArbOS fee accounts. Three are the current fee distributors — the network distributor 0xbc5c3a7adecf54d34169fd90dbd1b7d3142df067, the infrastructure distributor 0x5a2b80a9b7effc06129bd5462d77bc20a8a59be7, and the L1-surplus distributor 0x8f516b99… — each configured in the chain's first blocks with the 9,000/1,000 split to the same two recipients, and never altered. One further contract, 0xe6cf1072…, served as the infrastructure fee account earlier before that role passed to 0x5a2b…; it holds a residual balance and has never settled.

Control. Ownership of the three current distributors passed from the deployer 0xc8451c…30ca to a single address, 0x2a153c6a1b66dbc930a8d7017230ab0253005c09, within twenty-seven seconds of one another, on 24 June 2026, before the chain opened to the public on 1 July. It has never moved since. That address can change where the ecosystem's share is sent from all three current distributors. It never has. The earlier contract 0xe6cf1072… remains on the original deployer key and was never transferred; single-key control therefore holds for the three current distributors, not across every contract that has ever been a fee account. Neither key is named in any governance record we could locate.

Waiting balances. At the pinned height the two settling distributors held substantial balances awaiting the weekly split, and the ecosystem's recipient held a further amount already split but not yet routed to L1 — its last routing event was block 45,065,698, on 24 August.

i. Proofs (Abridged)

Contracts Network fee distributor 0xbc5c3a7adecf54d34169fd90dbd1b7d3142df067 Infrastructure fee distributor 0x5a2b80a9b7effc06129bd5462d77bc20a8a59be7 L1-surplus distributor 0x8f516b99… Operator recipient 0x8b3511b4c4b68fcafdce7e6f91b925458bb64ea6 Ecosystem recipient 0x92433c650785397386a0cc347ec2489718e45307

Counts — nine settlement dates to date (carried from PP-003, DCN JFD-20260916-PP-00001, kind-15 2949424); the two settling distributors carry them, and the L1-surplus distributor and the earlier fee-account contract have never settled. Nine contracts carry the distributor configuration event. Ownership of the three current distributors moved once each to 0x2a153c6a… on 24 June. The precise per-contract configuration and transfer counts are in The Full Proof.

Both arms. The control finding rests on a negative — that ownership of the current distributors never moved again — so it was established in two directions. The pass arm returns the known ownership events where they exist; the refute arm returns empty across the remaining history. An empty result is evidence only because the same query provably finds the events when they exist.

Exact arithmetic. Every amount was computed in exact integer arithmetic; no floating-point conversion appears in the evidentiary path. The method and queries that produce these figures are in The Full Proof; running them yields these figures, or our figure is wrong.

IV. LIMITS OF OBSERVATION

What this measurement does not establish, stated as plainly as what it does.

The 7 September figure was read from a public archive endpoint, not from our own instrument. Our archive did not reach that block at sealing. Every figure from the chain's deployment to 6 September is on our own archive with two independent surfaces in agreement; the 7 September figure is pending re-derivation and will be restated when it lands. We would rather say which surface answered than have a reader discover it.

The split is exact. That is not the same as the base being correct. We measured what the distributors split. We did not measure what entered them here. Fees that never reach a fee-distributor contract are outside this measurement, and are treated in the companion reconciliation.

The waiting balances are not evidence of anything being owed late. They are a measurement at a height. Whether they are consistent with an intended cadence is a question the chain alone cannot answer.

We did not measure intent, conduct, or compliance, and nothing here should be read as a finding about any of those.

V. CONCLUSION

The mechanism is honest. The 9,000 / 1,000 split is executed by code rather than reported by a party, was configured in the chain's first blocks across the three current distributors, has never been altered, and lands on the basis point in every settlement the chain has made. That is the part nobody had confirmed, and it holds.

What remains open sits elsewhere: what enters the base before the split, what waits between the rails, and who holds the key that can redirect the ecosystem's share.

A revenue line that moved by a multiple rather than a margin, in the final month of a quarter, inside a figure its own publisher describes as unaudited, is a figure worth verifying. It now has been, for the portion the chain can answer.

VI. GLOSSARY AND DEFINITIONS

Arbitrum Expansion Program (AEP) — the licence under which an Orbit chain remits a share of net protocol revenue to the Arbitrum ecosystem.

Basis point — one hundredth of one percent. A 9,000 / 1,000 weighting in basis points is a 90% / 10% split.

Both arms — our standing requirement that a finding be established in two directions: the pass arm showing the query returns the result when it exists, and the refute arm showing it returns nothing when it does not. A negative finding without a refute arm is not a finding.

Frozen method — a measurement specification written and hashed before it is run, so the method provably predates the result and cannot be fitted to it.

Net protocol revenue — the base on which the AEP share is computed, as defined by the AEP Terms. The definition and the reconciliation to it appear in The Full Proof and the companion Posit.

Pinned height — the block at which a figure was read. A figure without a pinned height is not reproducible.

Re-run hash — confirms a result obtained with the method. The method is in The Full Proof.

Refute arm — see both arms.

Settlement — one execution of the fee-distribution split, paying each configured recipient.

VII. ADDITIONAL RESOURCES

If you would like to read more of our series, Proof Posits, they are the receipts of our proofs and they are free: glasshull.jetfyul.com/#posits.

If you would like to test yourself against a locked call, our prediction series — The Hull Against the Machine — seals each prediction before the event and reveals it on a timer: glasshull.jetfyul.com/hull-against-the-machine.

If you would like the full report — the frozen method, the queries, every row, and the certification — that is The Full Proof, and it can be licensed: glasshull.jetfyul.com/full-proof.

---

SEAL AND VALIDATION OF THE SECOND ISSUE, FROM THE RECORD

This issue's own seal, witnesses and Bitcoin anchor are recorded on its verification page, https://gh.jetfyul.com/v/JFD-20260929-PP-00001, and in its certificate of authenticity, JFD-20260929-CERT-00001, not in these pages. The second issue's, as recorded:

DCN	JFD-20260920-PP-00001
Title of record	Robinhood Chain fee distributions
Document SHA-256	b15b75eedd5fb3a3acd64178b994494911cf524fc9e4c145e30970da006841ed (4,287,866 bytes)
Source text SHA-256	6b9d852546b04f343a7d59018c449b113a044cacae10202ce03c3b28bb83b5ac
Seal event	Glass Hull ledger kind-15 3024863, 2026-09-20 02:30:08 UTC, event hash 0bdb2d14b709d8bf222ed1df552eaf6fb2e75c2bf73ad745ff5a2b9c5fabe5d3
Seal authorisation	Notary lane check, ledger event 3024861
Witnesses	A9: ATECC608B serial 0123ADA0711D8597EE · A10: ATECC608B serial 0123DC797ACC1195EE · ECDSA-P256 (secp256r1). Two independent hardware witnesses; one signature is not a seal.
Seal request SHA-256 (signed by both chips)	6a7152e801e0eef8e735317fda497c0ce784ae724c1fc95fc559c7b0175c5995
Bitcoin anchor	Block 967784, hash 000000000000000000001d6ff7dc3535248d3fe4af4d7cb636d8baaf6c2a2125, mined 2026-09-20 02:46:41 UTC; confirmation recorded as ledger event 3027428
Certificate of authenticity	JFD-20260928-CERT-00007 (SHA-256 d3abda1db13133b7f028ed44d23f95b835a9540298c6796f6e0140699d4ae115), sealed as kind-15 3218920; it certifies kind-15 3024863
Verify page	https://gh.jetfyul.com/v/JFD-20260920-PP-00001

How to validate the second issue yourself

1. Compute the SHA-256 of the file JFD-20260920-PP-00001.pdf, over its bytes exactly as delivered. It must equal b15b75eedd5fb3a3acd64178b994494911cf524fc9e4c145e30970da006841ed, or the file has been altered.
2. Look up the seal event, kind-15 3024863, on the Glass Hull ledger. Its payload carries the document's identifier and that SHA-256. The ledger is append-only; an entry cannot be revised after the fact.
3. Verify the two hardware signatures against the published witness public keys. Either signature alone is insufficient by design.
4. Verify the Bitcoin anchor. It establishes that the document existed no later than block 967784.
5. Verify the signatures on-chain if you prefer: chain 4663 verifies P-256 signatures natively at address 0x100 via RIP-7212, so a contract can check our seal without an oracle.
6. Reproduce the finding with the method in The Full Proof.

---

Glass Hull · a product of the Founder's Unfair Advantage · Jet Fyul Dynamics LLC We hold no token and take no position in anything we measure.