Section · AVAT — Nasdaq: AVAT · CIK 0002092446
Avalanche Treasury Corp.
The treasury that borrowed. It holds roughly 3.5% of all circulating AVAX, has drawn $33M against it, and disclosed a going-concern note in its first 10-Q — withdrawn in its second. Its paper trail stops one inch short of the chain. Its twin, AVAX One, is covered separately.
11 — Custody
Nine vaults on a monthly clock
The Q2 10-Q names three custodians. The Wallets traces the Anchorage side, where two C-Chain wallets held 10,802,161 AVAX at the last block of 30 June 2026 against 10,802,156 filed. BitGo is history: 5,249,578 AVAX and 1,180,516 stAVAX at 31 December 2025, and nothing at either by 30 June. The third name is Coinbase Custody Trust Company, and its balance fell by 3,819,664 AVAX over the same six months. This chapter is where that money was, how it moved, and how far the evidence reaches.
Figure 01 — Filed
Two numbers, six months apart
The filing states the position at two dates and says nothing about addresses. The measured side is nine P-Chain staking keys, read from the transactions that placed and released them.
Nine P-Chain vaults account for 99.45% of the filed Coinbase Custody balance at 31 December 2025 and six of them for 98.85% at 30 June 2026
Each pair of bars is one filed date. The outline is the filed figure; the solid fill is the measured stake. Both dates share one axis that starts at zero and ends at the larger filed figure, 8,329,871 AVAX, so the fall between them is drawn to scale. The December measurement is the nine stakes placed on 1 January 2026, one day after the filed instant and before any reward accrues. The June measurement is the six survivors' exports of 30 June 2026, taken 24 hours before the filed instant, after which that AVAX sat on the C-Chain in the same vaults.
the nine stakes placed on 1 January 2026, which is the same money one day after the filed instant and before any reward accrues.
the six survivors' exports of 30 June 2026, taken 24 hours before the filed instant, after which the AVAX sat on the C-Chain in the same vaults.
The 98.85 percent is a subset-sum fit from a 63-second export window, and an export window cannot see a stake that spans the filed instant. At least four keys carrying the same fingerprint held 1,647,622 AVAX across 30 June 2026 23:59:59 and are excluded by quantity alone. The balance method is read at the instant itself and repairs part of this. The blind spot for a key that never exports stays real, and closing it needs a walk of every stake transaction between 1 January and 30 June 2026.
The decline between the two rows is 3,825,730 AVAX measured against 3,819,664 AVAX filed, which is 100.16 percent. The margins are thin: 45,623 AVAX in December and 51,689 AVAX in June. One more member above about 50,000 AVAX in either window ends the match. The decline row follows from the two rows above it once both are fixed, so it is bookkeeping rather than a third confirmation.
Source: Filed figures: AVAT Form 10-Q for the six months ended 30 June 2026, filed 19 August 2026 (SEC CIK 0002092446). Measured figures: a walk of platform.getBlockByHeight over three windows on the Avalanche P-Chain, 10,670 blocks and 3,869 transactions, keeping every AddPermissionlessValidatorTx and ExportTx of 100,000 AVAX or more that carries the batch fingerprint.
The same question, asked from balances
Stakes are one way to reach the filed figure. Balances are another, and they are read at the filed instant itself rather than a day either side of it. Fourteen C-Chain addresses carry this batch’s money: the five cold wallets, the six vaults’ own C-Chain counterparts, and the three wallets funded alongside them by one vault on 5 November 2025. Read with eth_getBalance at the three filed blocks, they come within about a thousand AVAX of every figure the company filed for Coinbase Custody.
| Filed date | Fourteen wallets | Filed | Covered | Gap | Filing |
|---|---|---|---|---|---|
| 31 Dec 2025 | 8,328,855.1650 | 8,329,871 | 99.9878% | 1,015.84 | 10-Q of 19 Aug 2026 |
| 31 Mar 2026 | 8,460,942.4590 | 8,462,227 | 99.9848% | 1,284.54 | 424B3 of 27 Jul 2026 |
| 30 Jun 2026 | 4,509,210.9947 | 4,510,207 | 99.9779% | 996.01 | 10-Q of 19 Aug 2026 |
The fourteen
Five cold wallets, listed in Figure 04. Six counterparts, listed against their keys in Figure 03. Three more, all funded by 0xC9CEDe65, the 75jvtcl vault, inside 115 minutes on 5 November 2025:
- 0x8da8c161a493e957c0a3e44a3ce920b8ab0a2c4512,572.36 AVAX at 31 December 2025. a batch member in its own right, missed because it stakes below the 100,000 AVAX floor the P-Chain walk used: P-avax1gfddpgxxld3qgkl53evddup6ly2c95356524mu on NodeID-145uWMw7TbEZuhU4EaerESBBaDsmLneiu, re-staked 3 July 2026 to the same 25 September end as the four survivors.
- 0xf9059a012fa69b3e8d7d0488b30d0c7db36a84ef21,894.07 AVAX at 31 December 2025. the same case: P-avax13ajllh5hujtrmuhds67nz0twa0yjxxzdxywtus on NodeID-2VwCwLusvnfiZ55tQY6g7UEZdpBVAgs6W, also re-staked 3 July 2026 to 25 September.
- 0x03ff64055443b1228549793e9e49391b0710175110,140.35 AVAX at 31 December 2025. has never touched the P-Chain and has held the same 10,140.35 AVAX for ten months.
Both filed sentences say “approximately”. A residual near a thousand AVAX on a figure of 4.5 to 8.5 million is 0.012 to 0.022 percent, which leaves no room for a missing vault of more than about a thousand AVAX and is consistent with one idle sub-balance or with the filing's own rounding. The blocks are 74,824,633 (2025-12-31 23:59:56 UTC), 81,789,468 (2026-03-31 23:59:59 UTC) and 89,166,730 (2026-06-30 23:59:59 UTC), each confirmed with eth_getBlockByNumber against https://avalanche.drpc.org.
Figure 02 — The clock
What a custodian's staking job looks like from outside
The nine were found by shape, not by label. Every cycle read on these keys does the same five things, and the custodian sets all five. A client can change its own habits; it cannot change the job its custodian runs.
An ImportTx brings the whole C-Chain balance onto the P-Chain.
An AddPermissionlessValidatorTx stakes all of it, at a delegation fee of 100 percent, which leaves an outside delegator nothing to earn. It does not shut the node: platform.getCurrentValidators on the node carrying the largest vault's stake returned delegationFee 100.0000 alongside delegatorCount 2 and delegatorWeight 260,911.20 AVAX at the 17 September 2026 read.
The reward owner is a single address with a threshold of 1, and it is the staking key itself. That is a one-key vault, not the 2-of-3 output the BitGo wallets used.
Through June 2026 the stake ended at an exact month end, 00:00:00 UTC on the last day of the month. The July cycle broke that: the four surviving vaults re-staked on 2 and 3 July to 25 September, a Friday.
An ExportTx returns the stake to the C-Chain within minutes of that midnight, and the siblings act inside about a minute of each other.
The seconds give the job away. On 30 December 2025 three of these keys exported at 00:47:05, 00:47:06 and 00:47:07 UTC. On 30 June 2026 six of them exported inside 63 seconds, from 00:08:02 to 00:09:05. On 30 April 2026 two stakes returned in the same block. A threshold of 1 on every output makes each key a one-key vault, which is a different arrangement from the 2-of-3 outputs the BitGo-era keys used.
Figure 03 — The nine
Three tied by flow, six by arithmetic
Three of the nine have a cold wallet that paid into the June delivery, and that delivery reached AVAT’s treasury wallet. The other six were found by arithmetic, and each of them now has a C-Chain counterpart as well, printed in the first column. A counterpart is the address the key imports from and exports to; it carries the balance method below, and it pins nobody on its own. Their standing stays collective.
| P-Chain vault key | Staked 1 Jan 2026 | Exported 30 Jun 2026 | At the 16 Sep 2026 read | Tie |
|---|---|---|---|---|
| P-avax1ahc32g2x340g2r09m2u4vv7lcr6lsqhdxjnjjaJan node NodeID-Cgdu1Ghr… · cold wallet 0x8de2cf20… | 1,505,055 | left 10 Jun | swept the other four cold wallets on 9 June and sent the delivery on 10 June; C-Chain balance zero since | flow |
| P-avax1px56m7z0sugl45pj8ykt3uyx2wfw80rm82dgsgJan node NodeID-4gp6z23D… · cold wallet 0x720efa4a… | 719,474 | left 10 Jun | paid 1,259,389 AVAX into the 9 June consolidation; its C-Chain wallet was refilled to 483,924 AVAX between 1 July and 3 August 2026 by six payers, which is a separate question | flow |
| P-avax1fdlvwr4zjl9kd6r4rtppqnlunvct7x97geds7jJan node NodeID-C5sarWUt… · cold wallet 0x628eef4d… | 1,122,680 | left 10 Jun | paid 1,245,363 AVAX into the 9 June consolidation; C-Chain balance zero since | flow |
| P-avax1xpedzqexwkzjhuj4dp5dcmc0nvphvwld2mkcwaJan node NodeID-CgTe7m2x… · C-Chain 0x1c90D5440c0E768CF7Ed984770E0D51A6fe8E417 | 2,524,681 | 2,534,260 | 2,500,060 AVAX staked on NodeID-DVh35F3R… to 2026-09-25 00:00:00 UTC | arithmetic, collective |
| P-avax1lhewhmafsejpyj5dt2crtvnpculex5vhf39zmdJan node NodeID-8QDaNCzY… · C-Chain 0x4d8708374a1Dc42126534Ba989C796313c0fD0cc | 1,170,222 | 1,174,845 | 1,158,985 AVAX staked on NodeID-8QDaNCzY… to 2026-09-25 00:00:00 UTC | arithmetic, collective |
| P-avax175jvtclx70razxnfxq6kf7vdw50kvj4mkwlqwmJan node NodeID-DQdJGdg9… · C-Chain 0xC9CEDe653245689928AE513209F2215138E5aE53 | 644,165 | 351,986 | 166,676 AVAX staked on NodeID-DQdJGdg9… to 2026-09-25 00:00:00 UTC | arithmetic, collective |
| P-avax1gd8jekzy9q4j5tr937mygd77kjactxna6890y8Jan node NodeID-FTbhiFzL… · C-Chain 0x8CADa8839c830e6f67a5E6837CBD0477E9c2384E | 278,192 | 100,068 | not validating at the 16 September 2026 read; where its 100,068 AVAX went is not established | arithmetic, collective |
| P-avax1j9tgjlu7xl40p68u5gp2gc32g59te6mgfwqczxJan node NodeID-F5Ecwd5q… · C-Chain 0x2444CE0428Ac2aEd55B663E664966D942eF3Ca62 | 169,093 | 145,555 | not validating at the 16 September 2026 read; where its 145,555 AVAX went is not established | arithmetic, collective |
| P-avax1wyd54qc3fen63q6hvzzzg9rjp9rct8ewdjs8t2Jan node NodeID-CS2Rnk3P… · C-Chain 0x75B0870CE99Bb97cA5313C75c3fCDad062aeE82c | 150,686 | 151,803 | 149,493 AVAX staked on NodeID-CS2Rnk3P… to 2026-09-25 00:00:00 UTC | arithmetic, collective |
| Nine vaults · 3 by flow, 6 by arithmetic | 8,284,248 | 4,458,518 | 4 staked, 3,975,214 AVAX, all ending 2026-09-25 00:00:00 UTC |
Four of the six survivors are staked today. They re-staked on 2 and 3 July 2026 and all four end at the same second, 2026-09-25 00:00:00 UTC, carrying 3,975,214 AVAX between them. That is the next event on this side of the record, and it is a Friday rather than a month end.
The other two stopped validating after June. Their June exports were 100,068 and 145,555 AVAX, 245,623 AVAX in all, and where that went has not been established. It is not on any key this research has read.
Figure 04 — The C-Chain side
Five cold wallets, read at the two filing blocks
Three of the nine vaults stake from a C-Chain wallet. Two more wallets never touch the P-Chain and only pay into the group. Each balance is eth_getBalance at block 74,824,633 (2025-12-31 23:59:56 UTC) and block 89,166,730 (2026-06-30 23:59:59 UTC), read against an archive node.
| Wallet | 31 Dec 2025 | 30 Jun 2026 | 16 Sep 2026 |
|---|---|---|---|
| 0x8de2cf20044bcef9067d38dd086410334616a142the hub: funded once by the Prime hot wallet on 7 Nov 2025, five monthly cycles, then the consolidation and the delivery | 1,505,055 | 0 | 0 |
| 0x628eef4de1b5d6db3fa183b6ce9d3a5f28c4a791cycled from 15 Oct 2025; paid 1,245,363.0931 AVAX in at 18:42:01 on 9 June | 1,057,169 | 0 | 0 |
| 0x720efa4a6193f1d25234674f8b041009c2608f35cycled from 10 Oct 2025; paid 1,259,388.6891 AVAX in at 18:47:03 on 9 June; refilled by six payers between 1 July and 3 August 2026 | 719,474 | 0 | 483,924 |
| 0xb5244a69813ffa75f8218011c582b9b88a923eb7never touches the P-Chain; paid 18,253.2760 AVAX in at 18:38:02 on 9 June | 50,071 | 0 | 0 |
| 0x97387b9b292add0317c615d97f1f15dd408ee2f9never touches the P-Chain; paid 11,177.6401 AVAX in at 18:36:02 on 9 June, and paid the Prime deposit forwarder 5,615 AVAX on 1 July 2026 | 15,439 | 5,615 | 19,701 |
| Five wallets | 3,347,208 | 5,615 |
A zero on this table means the AVAX was on the P-Chain, not that the wallet was empty. The December figures are readable at all because the job exports on 30 December and re-stakes on 1 January, so at the filed instant the money had just come home. Two days either side the same query returns nothing. The five wallets’ C-Chain balances at block 74,824,633 sum to 3,347,208.4909 AVAX and the three staking wallets’ 1 January stakes to 3,347,208.4910, a difference of 0.0001 AVAX. That is the check that the C-Chain reading and the P-Chain reading agree. The two sums cover different wallets: the three staking wallets’ own December balances come to 3,281,698.3119 AVAX, which is 65,510 AVAX short of the five-wallet total.
The two wallets that never stake are tied to the group by arithmetic on two measured numbers. They held 65,510.179 AVAX between them at the December block, and the vault they pay into staked 65,510.1791 AVAX more on 1 January than its own December balance. They were swept in between the two dates, so the December sum counts them once and the five wallets are one book.
Figure 05 — The delivery
Two days in June, and 4,060,744 AVAX changes custodian
On 9 June 2026 the four other wallets paid into the hub over eleven minutes. On 10 June the hub sent the lot to a Coinbase Prime deposit address, which forwarded it to the Prime hot wallet, which paid AVAT’s treasury wallet 79 minutes after the first hop. Two rehearsals bracket it: 2,000 AVAX down the same route 44 hours ahead, and a 5 AVAX test into the treasury on the day.
| When (UTC) | From | To | AVAX |
|---|---|---|---|
| 2026-06-09 18:36:02 UTC | 0x97387b9b… | 0x8de2cf20… | 11,177.6401 |
| 2026-06-09 18:38:02 UTC | 0xb5244a69… | 0x8de2cf20… | 18,253.2760 |
| 2026-06-09 18:42:01 UTC | 0x628eef4d… | 0x8de2cf20… | 1,245,363.0931 |
| 2026-06-09 18:47:03 UTC | 0x720efa4a… | 0x8de2cf20… | 1,259,388.6891 |
| 2026-06-08 19:17:01 UTC | 0x8de2cf20… | 0x7f744757…the rehearsal, 44 hours ahead of the delivery and down the same route | 2,000 |
| 2026-06-10 15:42:01 UTC | 0x8de2cf20… | 0x7f744757…the Prime deposit address, which forwards everything it receives and holds nothing at either filed blocktx 0xc105c97947198dfd… | 4,060,686.5278 |
| 2026-06-10 15:42:58 UTC | 0x7f744757… | 0x1d67bd45…into the Coinbase Prime hot wallet, 57 seconds latertx 0x2149fe2eceac2e2f… | 4,060,686.5278 |
| 2026-06-10 17:02:11 UTC | 0x1d67bd45… | 0x98a90680…into AVAT's treasury wallet, 79 minutes after the delivery startedtx 0x346cda5ab494e6e1… | 4,060,744.2591 |
One hop of that route is not a transfer anybody can follow. The Prime hot wallet is an omnibus: it held 87,941 AVAX at the December block and 79,275 at the June one, so a four-million payout needs a same-hour inflow, and an internal Coinbase shuffle of 3,848,581 AVAX ran through it in the meantime. What is measured is 4,060,687 AVAX in at 15:42:58 and 4,060,744 AVAX out at 17:02:11, 57.7313 AVAX apart. That is an amount-and-timing match across an exchange hop, and it is the one link in this chain that is not a direct transfer. Every other hop on the table has a hash and two endpoints.
What the delivery settles is where the Coinbase Custody decline went. The filing puts 3,819,664 AVAX of decline over six months; the measured departure is 4,060,686.53 AVAX, and it arrives at a wallet that the same filing’s Anchorage figure pins to the unit at 30 June. The decline is a custodian transfer. Whether any part of the position was also sold in the period is a separate question that these transactions do not answer.
The departure is 241,022.53 AVAX larger than the filed decline, because the account took AVAX in during the half as well as sending it out. Two of those top-ups were measured: 519,545.39 AVAX into 0x720efa4a on 1 April and 103,947.88 into 0x628eef4d. Both wallets paid into the hub on 9 June and left in the same batch, so the June transfer carries more than six months of decline. Reading the gap that way is an inference from two measured top-ups, not a reconciliation of the account.
Figure 06 — One date
What 30 June 2026 adds up to
Both sides of the filed position, measured at the same instant, against the total the 10-Q reports for that date.
The residual on the Coinbase side is 46,074 AVAX, and the six vaults’ C-Chain counterparts are what would close it or break it. Nothing in this arithmetic mixes dates. Every figure is read at the same block, or at the export 24 hours before it, and no residual is carried across from one filing date to another.
Figure 07 — The nodes
Three machines both operations used
Two nodes carried a vault’s stake and later carried a key the research already places on the AVAT trail. A third is a prefix match that has not been confirmed.
| Node | On the Coinbase Custody side | On the trail |
|---|---|---|
| NodeID-Cgdu1GhrbwSD4x3vRTp9fFe3sitWckyPDmeasured | carried 0x8de2cf20's stake for five consecutive monthly cycles, 7 November 2025 to 30 April 2026 | carries 1,133,999 AVAX for P-avax1qjukw4p93pyygk0zdgw2zqsyspatw829d5gtqn, one of the two keys AVAT's counterparty funded on 9 September 2026 |
| NodeID-F5Ecwd5qinZSTxacdEfWWsnz2eXyp8uK6measured | carried batch member P-avax1j9tgjlu's 169,093 AVAX from 1 January 2026 | carries AVAT validator 1, P-avax16rq0ey0yezqax6cwwry4kvh2grrw58mdp2tct0, 1,999,999 AVAX at a 5 percent fee |
| NodeID-DVh35F3R11HAAxzeSTbk3eR9w2BzPrJQtmeasured | carries the largest vault's 2,500,060 AVAX to 25 September 2026 | the full node ID matches one the registry lists among those that ran AVAT's BitGo-era stakes, 30 October 2025 to 1 April 2026, confirmed by three routes. The same node has carried an unrelated private delegator since January 2022, so the overlap is weak |
A node identifies the provider rather than the customer, and these are rented commercial validators. A single private delegator unrelated to either company has used NodeID-DVh35F3R since January 2022 and NodeID-3kUUeZB7 since September 2021, so a shared node carries almost no weight on its own. All three overlaps are read as weak corroboration beside the filed sums, never as a tie.
The same job, a different client
One more key rides the same 63 seconds on 30 June 2026 and wears the same fingerprint: P-avax1n7sgakzfml8ea4fg76qm7pnsn0v06gw5r5wc0l, with the C-Chain wallet 0xa0b25c31…. It exported 616,217 AVAX at 2026-06-30 00:09:01 UTC and stakes 619,983 AVAX today on NodeID-Eeg2Huip… to 2026-09-30 00:00:00 UTC. It rides the same 63-second window and wears the same fingerprint, and its cycle has run unbroken since August 2023, years before AVAT existed. Adding it to the June sum overshoots the filed figure by 12.5 percent; leaving it out lands within 1.1 percent. Same batch job, different customer. It is named here so that nobody re-finds it and reads it as AVAT's.
Figure 08 — Standing
The claims, and the strength of each
The standing in one sentence. The cluster is matched to the filed quantity at three dates, as a set, and that is an inference rather than a proof. Three filed dates are reproduced on one method to within about a thousand AVAX, and no individual member is pinned on its own.
Correction, 17 September 2026: this chapter first said the cluster was proven by filed quantity. On this site proven means a filed quantity matched to the unit, or a transaction identified by its hash. The cluster matches to about a thousand AVAX on figures the filings call approximate, so the word was wrong and the sentence was changed. No measurement changed.
The decline across the first half of 2026 is the difference of the two date sums. It follows arithmetically once both of those are fixed and cannot fail on its own, so it counts as bookkeeping rather than as a third confirmation.
The December C-Chain balances and the 1 January stake amounts are the same AVAX read on two sides of one atomic swap. Their agreement to 0.0001 AVAX checks that two readings of one quantity agree; it is not independent evidence of the filed figure.
The filings call each figure approximate. Ruling another key out because adding it would overshoot by a few thousand AVAX reads that word as exact to about a thousand AVAX, which the filings do not promise.
The set was fixed by flows and by one staking pattern, not by searching for a sum. Other stakes of the same pattern were live at the June instant under keys nobody has tied to this set, and not all of them have been examined.
Tied by flow. Three of the nine are tied by flow. Their C-Chain wallets paid into the consolidation of 9 June 2026, and that money reached AVAT's treasury wallet the next day. One hop of that route runs through a Coinbase Prime omnibus, so it is an amount-and-timing match rather than a single transfer, and it is labelled so.
Tied by arithmetic. The other six are tied by arithmetic, and collectively. Their stakes make up the rest of two sums that come to 99.45 percent and 98.85 percent of the filed figures, and their C-Chain counterparts carry the balance method to within about a thousand AVAX at all three filed dates. No single one of the six is pinned on its own, and none should be described as if it were. The cluster is matched to the filed quantity as a set. No key is proven on its own, and the set is an inference.
What breaks it. What would break it: a tenth member in the hours of 1 January 2026 that were not walked, or an eighth in the June window. On the stake method the margins are 45,623 AVAX in December and 51,689 AVAX in June, so there one more member above about 50,000 AVAX in either window ends the match. The balance method is the tighter of the two and so the easier to break: its residuals are 1,016 AVAX in December and 996 AVAX in June, and an unlisted member above roughly a thousand AVAX ends the match there.
What settles it. What would settle it: a walk of the 2 and 3 July 2026 re-stake cohort, read the way the 1 January cohort was, which would give a stake-side June figure to set against the export-side one. The six keys' C-Chain counterparts have since been found and read at all three filed blocks, and that is what the balance method rests on.
How far the reads reach
- Delegators were not read at all. platform.getCurrentValidators no longer returns a delegators array, and 58 nodes carrying 100,000 AVAX or more of delegated weight returned no delegator rows. Any member of this batch that delegates instead of validating is outside the scan. That is the largest hole.
- Three windows were walked out of a chain that ran about 1.5 million blocks across the period: one hour around 30 December 2025, seven hours of 1 January 2026, and 26 hours around 30 June 2026. The month ends of January through May, July and August 2026 were not walked.
- The nine-key sums rest on P-Chain stake and export amounts rather than on balances. The control that shows the two readings agree is on the cold wallets: the five wallets' C-Chain balances at block 74,824,633 sum to 3,347,208.4909 AVAX and the three staking wallets' 1 January stakes to 3,347,208.4910, a difference of 0.0001 AVAX. The three staking wallets' own December balances come to 3,281,698.3119 AVAX, and the 65,510 AVAX between the two sets is the satellite sweep. Those two readings are opposite sides of one atomic swap, so their agreement checks the reading rather than the filing.
- The 31 March 2026 date, where the 424B3 puts 8,462,227 AVAX at Coinbase Custody, is covered by the balance method alone. The P-Chain around it was not walked for this chapter, so there is no stake-side figure here to set beside the other two.
- No explorer tag was read, and nothing here depends on one.
Not claimed
Nobody has filed an address for any of these wallets or keys, and no explorer tag was read for any of them. The six vaults are described as a set because that is the only standing the evidence gives them. The 10 June delivery is a custodian transfer on the measured record, and it says nothing about whether AVAX was also sold in the period. Coinbase Prime funding, a round size, a monthly cadence and a shared node do not identify a customer on their own, and none of them is used here as if it did.
Sources and method
Historical C-Chain balances are eth_getBalance at blocks 74,824,633 and 89,166,730 against the archive node https://avalanche.drpc.org; the pruned public nodes answer these heights with “missing trie node” and cannot be used. P-Chain stakes and exports come from platform.getBlockByHeight walked over three windows and from platform.getCurrentValidators, read 16 September 2026 at 22:12:32 UTC. Transfer histories come from Routescan v2, which returns the atomic rows the EVM explorers omit. Filed figures come from EDGAR under CIK 0002092446.