TL;DR
One of the two things that produced last month's avoid turns out to have been our mistake rather than Pacifica's. The BlockSec audit is not a 404. Pacifica's docs link a report hosted in BlockSec's own GitHub repository, it covers perp-contracts/programs/pacifica-mainnet/src/lib.rs, and that is the program holding our deposits. We read it end to end. Two Low findings, four recommendations, two notes, nothing medium or high.
The correction matters more than the upgrade. GitHub's commit history shows that file was added to BlockSec's repository on 30 April 2026, under the message "Add missing public reports", and the bridge report on 7 April 2026. Both were public more than three months before our 3 August report declared the audit absent from the firm's own site. Nothing changed this month. We looked in the wrong place last month and wrote down a red we had not earned. Give Pacifica the credit that is genuinely due, which is that the audit exists and the deployed program has never been upgraded, and take the blame that is genuinely ours.
The other red did not move, and looking harder made it worse. The upgrade authority is still BzZgjpBFHb19vWjBRoM5eVy6qnUjbGH5tMJrnN7cA47a, still an ordinary on-curve key, still not the Squads vault the Fund Security page promises you can verify. That same key is stored inside the program's own central_state account, so one plain key is both the deploy authority and the in-program admin. And there is a second key. FHFvAvN1F6wzBGVwnAFM4kJ26BmM8rq1f7zhpTewnQ8r is also an ordinary on-curve key, it also sits in central_state, and we watched it sign a BatchWithdraw on its own at 16:39 UTC today, moving 999 USDC out of the hot bridge. It signs again every few minutes. BlockSec wrote the consequence into its own report: accounts with withdraw_authority permission can withdraw user assets from the vault unlimitedly, and if those private keys are lost or exploited it could lead to losses for users. That is the auditor, not us.
Checklist
Audits and contracts. This is the section that changed, though not for the reason the calendar suggests. Pacifica's Audits page points at raw.githubusercontent.com/blocksecteam/audit-reports/main/rust/blocksec_pacifica_fi_perp_contracts_signed_20250424.pdf, which returns HTTP 200 and 1.71 MB. Client Pacifica Fi, target Perp Contracts, version 1.0, dated 24 April 2025. Scope is a single file, perp-contracts/programs/pacifica-mainnet/src/lib.rs, at commits b4c48da19e17efd015bc5c57a91e13731ee199e8 and e3631ce1efe06308e838a8785c0ac71742346b4d. The findings table has eight numbered rows: two Low risk findings (a potential DoS in batch_withdraw() and a missing minimum value check in deposit()), four recommendations, and two notes. IDs 1 through 6 all carry status Confirmed, which in BlockSec's own legend means recognised by the client but not fixed yet. The two notes carry no status at all. A separate report, "Security Audit Report for Pacifica's Bridge Contract", dated 1 April 2026, covers programs/sol-bridge and programs/spl-bridge with three non-critical issues and no criticals.
Three caveats keep this at watch rather than strong. Only the bridge report appears on the blocksec.com audit-report index; the perp report exists there as a file in the repository but is not an indexed entry, and the "Perp Contracts" entries on that index belong to HertzFlow, a different client. The repository pacifica-fi/perp-contracts is private, so there is no way to check that the bytecode deployed on 31 May 2025 was built from the audited commit, and nothing on chain is source verified. And the scope is narrow by design: the off-chain matching engine, the oracle, the liquidation and ADL waterfall, the money market and the vaults are where most of the discretionary power lives, and we found no audit of any of them. One incidental confirmation is worth keeping: the audit discusses a function called batch_withdraw(), and the live instruction we observed on chain today logs as BatchWithdraw. The report and the deployed program are the same thing.
Admin control. The crux, and it is now sharper than last month rather than softer. The program PCFA5iYgmqK6MqPhWNKg7Yv7auX7VZ4Cx7T1eJyrAMH runs under the BPF upgradeable loader with program data at FC5zcmPmqo5188fLVga2HN5QZ1T1ojhNneVkFdu6eF1E. That account has exactly two signatures in its entire history, both on 31 May 2025 at 23:01 and 23:06 UTC, and the recorded last deploy slot is still 343778047. No upgrade, ever, in fifteen and a half months, and this run we could date it exactly rather than estimate it from slot arithmetic. The upgrade authority is still present and unchanged. We re-ran the curve test against controls in both directions: the authority comes back on curve, and all four published bridge and vault addresses come back off curve. A Squads vault would be off curve. This one is not.
The new finding is inside central_state. The hot USDC bridge's token authority, 9Gdmhq4Gv1LnNMp7aiS1HSVd7pNnXNMsbuXALCQRmGjY, is a 185 byte account owned by the Pacifica program, and it is confirmed as the SPL authority on the hot bridge token account. Decoding it gives five non-zero pubkeys. The first is BzZgjpBF..., the same key as the upgrade authority, so the deploy key and the program's own admin role are one key, not two. The second is FHFvAvN1F6wzBGVwnAFM4kJ26BmM8rq1f7zhpTewnQ8r, another ordinary on-curve key. We pulled its transaction: signature 5DTLfRXYFNjtZmaK8Jp2MKLFw4MS9DNEvshQdKuGjRjqTgFvD7ayPMpA7ZfaAX34wrE4tDZYJssktysATfCb1yf6, 19 September 2026 at 16:39:25 UTC, sole signer, instruction BatchWithdraw on the Pacifica program, with the token balance delta confirming exactly 999 USDC out of 72R843Xw... into a user's token account. It had signed at 16:26 and 16:32 in the preceding thirteen minutes, and again at 17:15. The remaining three slots hold keys whose accounts do not exist on chain at all, so their roles are unresolved.
Put the audit next to the chain and the picture is complete. BlockSec's note 2.3.1 says privileged roles govern system-wide operation and that accounts with withdraw_authority permission can withdraw user assets from the vault unlimitedly, with key loss or compromise leading to user losses. On chain, the key that signs withdrawals is an ordinary keypair operating continuously in production. The withdraw_authority risk is strictly worse than the upgrade-key risk, because the upgrade key has not been touched since 8 April 2026 and sits mostly idle, while this one is hot by construction. Against that, note 2.3.2 flags a front-running risk in initialize() that is now moot, since the program was initialised long ago. The Fund Security page has not changed a word: it still says the upgrade authority is distributed across multi-sig participants, still says Pacifica is transitioning to that architecture, and still publishes no multisig address, threshold, signer set, timelock or spending limit. The $5 million daily hot-wallet limit on that page is offered as an example, not a committed parameter.
Oracle. No change from August, and re-read from the docs this run. Pacifica computes its own oracle every three seconds as a weighted average of USDT-denominated prices from Binance (weight 2), OKX, Bybit and Hyperliquid (weight 1 each), and marks positions at the median of that oracle, the median of its own best bid, best ask and last trade, and perpetual prices from major exchanges. The construction is sensible and resistant to a single venue print. The trust question is unchanged: the party holding your collateral also produces, publishes and applies the number that sets your liquidation price, with no third-party feed and no on-chain attestation. For our two shorts this is exposure we accept rather than exposure we can verify.
Liquidity and exit. Still two different things with two different answers. Market liquidity is fine and slightly broader than last month: 77 markets against 75, about $466.5M of 24 hour volume, about $96.1M of open interest. Closing 69 HYPE or 41.12 SOL is not a liquidity event. Getting money off the venue is still the structural problem, and the buffer behind it got thinner. Hot USDC in 72R843Xw... fell from $1,357,510.70 on 3 August to about $456.8k today, while the cold vault sat at exactly 24,000,000.00 USDC, unchanged to the cent, which reads as a manually managed reserve rather than a live sweep. That works out to 1.87% hot and 98.13% cold.
Two corrections to last month's read belong here. First, the hot and cold story is a USDC story only. The published SOL cold vault 8nFeyzTF... holds 0.001 SOL, and all 4,577.4 SOL, about $513k, sits in the hot bridge. On the SOL side there is no cold buffer at all. Second, the withdrawal caps are $500,000 per account per 24 hours, not the $250k both the draft and our August report printed. Deposits are capped at $500,000 of account equity, spot assets carry a $500,000 per-user daily withdrawal cap and a $50,000 daily deposit cap, and an exchange-wide cap applies whose value is not published. Neither binds at our size. The dependency does, and there is still no documented permissionless path that lets a user exit without the matching engine.
Yield: real versus emission. Nothing here is protocol yield, and that is fine, because the point of the position is the funding. Both legs print the same hourly rate, +0.0000125, which annualises to about +10.95% simple with longs paying our shorts on HYPE and on SOL. In August SOL was negative and we were paying; now both sides earn. Read that with care: both markets sitting on exactly the same figure looks like a baseline rather than a premium-driven rate, so treat +10.95% as the current print and not a forecast. Auto-lend remains on by default, gated on an auto_lend_disabled flag and a 1,000 USDC lendable-capacity floor, and it was economically nil in August at 0.109% pool utilisation; we could not pull money-market state this run, so we are not restating that. The points programme launched on 4 September 2025 and distributes 10,000,000 points weekly, so it has now run for roughly twelve and a half months with no token, and the only TGE chatter we found on X is people asking when it is and quoting OTC prices for points.
Holder concentration. Nothing to measure. DefiLlama carries no gecko_id, no cmcId and a symbol of - for Pacifica, and neither the official account nor the wider X scan between 1 August and 19 September produced a token launch, an airdrop or a dated TGE. Third-party sites still publish a PACIFICA price page with nothing behind it, so anything trading under that ticker is unverified. This dimension becomes real on the day the points convert, and the questions then are the insider allocation and the conversion ratio.
Track record. Fifteen and a half months of mainnet with no platform-level failure we can find. DefiLlama's hack registry holds 1,273 entries and none of them is Pacifica. An X scan of the official account and of the wider conversation from 1 August to 19 September turned up no outage, no withdrawal halt, no delayed withdrawal complaints, no ADL event, no bad debt and no insolvency claim; the account's own posts in that window are product releases such as OCO orders, trailing stops and referrals, between 15 and 18 September. Our own incidents table has recorded nothing since the last report.
One qualifier is firmer than last month rather than merely aged. Closed Beta is not a leftover guide page. The live Deposits and Withdrawals page conditions its current limits on it in so many words, and the official X bio reads "Closed Beta, now live". Fifteen months into mainnet, with $25M of deposits, the venue is still operating under a beta label by its own description. The Overview page separately still claims "over thirty-five perpetual markets" while the API returns 77, and quotes 3x to 50x leverage while SOL and HYPE both cap at 20x in the API. The January 2026 ATH of $46.77M is now 46.6% above where TVL sits. The one disclosed incident remains the Mithril API-key database breach of 21 January 2026, a third-party trading-bot builder, stated by Pacifica not to have affected its users. Watch for name collisions when scanning: most "Pacifica outage" hits are a city in California.
Worst case
The shape of the worst case changed this month. In August the headline scenario was a compromised upgrade key redeploying the program. That risk is still live and still reaches the hot balance, because the hot bridge's authority is a PDA of the program itself, but it now looks like the slower of the two paths. The faster one needs no redeploy at all. A single ordinary keypair signs BatchWithdraw in production every few minutes, and BlockSec's own note says that role can withdraw user assets from the vault without limit. Whoever holds FHFvAvN1... does not have to be clever. They have to run the instruction the venue already runs all day.
What that reaches for certain is the hot bridge, and the hot bridge is bigger than the USDC number suggests: about $456.8k of USDC plus roughly $513k of SOL, because the SOL cold vault is empty. Call it just under $1M rather than the $456k a USDC-only read would give you. Whether it reaches the 24,000,000 USDC cold balance depends on which program derives 5kwCMKjE3Krvs7cHfcQ9kBkGyPQphd3oJ4KnsXcpMoVc, and that is still not readable from account state, because an uninitialised off-curve PDA is system-owned and dataless whether Squads derives it or Pacifica does. Pacifica says the cold vault cannot send funds to arbitrary addresses and that this is technically impossible. We cannot check it, and the gap between those two sentences is roughly twenty-five times the hot balance.
The likelier scenario still does not need an attacker. The engine goes down, or slows, during the hour SOL or HYPE actually moves. Both showcase shorts are on this one venue and both hedges are elsewhere, jitoSOL on Solana and PT-kHYPE on Pendle. If Pacifica stops matching you cannot close, cannot add margin and cannot withdraw, while the long legs keep repricing. Nothing in the last six weeks suggests this is imminent. Nothing in the architecture prevents it either.
The short-specific version is unchanged and worth restating because the market moved against these legs. Pacifica's third liquidation tier auto-deleverages profitable opposing positions when a counterparty falls below zero equity and the prior tiers had insufficient liquidity. In a sharp drop the profitable side is our two shorts, so ADL closes the hedge at the exact moment it is working. There is a Pacifica backstop liquidator that absorbs positions at tier two, below two-thirds of maintenance margin, but no insurance fund is described anywhere in the documentation, and the backstop explicitly declines a list of experimental markets. On sizing: HYPE marks at $92.89 against $53.07 in August and SOL at $112.06 against $72.49, so both shorts are well underwater on price and correspondingly further from their liquidation levels, while the hedge legs have appreciated. Roughly $11.02k of short notional on one venue, and it is still 100% of the short side. A single Pacifica failure un-hedges both pairs simultaneously.
Bottom line
Avoid, and the reasoning is narrower than last month rather than broader. Give Pacifica full credit for what is true: the audit is real, it is hosted by the firm rather than by the project, it covers exactly the program that holds the money, we read every finding in it, and the code it describes has been on chain untouched since five weeks after the report was signed. But be honest about the timeline. That report has been public since 30 April 2026. Our August red on this dimension was a research failure, not a Pacifica failure, and correcting it is not the same as Pacifica fixing something. If the audit had been the only red, we would have been writing caution six weeks ago and we were wrong not to.
It was not the only red. The gate is worst-flaw-wins on the surface that can lose the money, and two flaws still sit on it. Custody is deposit to venue with no documented unilateral exit, so a dead or hostile engine means no way out, and on the SOL leg there is not even a cold buffer behind it. And the multisig claim still does not reproduce when you run the check the Fund Security page invites you to run, except that now we can be more specific about why it matters: the deploy key and the program admin role are the same ordinary keypair, and a second ordinary keypair signs unlimited withdrawals in production, a power the auditor documented in writing and nobody has constrained since. Verifiability did not improve on the dimension that matters most. It got easier to describe exactly how it fails.
Three things would move this to caution, and they are the same three as August with one made concrete. Move the program upgrade authority to a Squads vault so it reads off curve like the vaults do. Publish the multisig address, threshold, timelock and the derivation of the cold-vault PDAs so the Fund Security claims can be checked. And put withdraw_authority behind something better than a single hot keypair, or publish the spending limit that bounds it, because right now the auditor's word for its reach is "unlimitedly." A fourth would help: move the SOL balance into the cold vault that already exists for it and currently holds 0.001 SOL. Until then the operational advice is unchanged and is about correlation, not about Pacifica's odds of failing: both shorts on one venue is the thing that breaks two hedges at once, and splitting them across a second venue costs a little funding and removes a single point of failure. Auto-lend is still on by default and can be turned off with auto_lend_disabled.
Data appendix
- TVL: about $25.0M.
api.llama.fi/tvl/pacificareturned 24,972,288.94 earlier on 2026-09-19 and 24,957,459.93 on our own call. Reconciles on chain to 24,969,714.84, a 0.05% gap: 24,000,000.000000 USDC cold inFvGqozihHY4Mmh4Nmx2rkXnMDcumiXEbjxA9oRu3mUeJ, 456,757.733094 USDC hot in72R843XwZxqWhsJceARQQTTbYtWy6Zw9et2YV4FpRHTa, and 4,577.404381005 SOL in9sSr35zwnFTuv2kZ86i55sR9dqQLTG663homexrLYgYuat a $112.06 mark. History from the DefiLlama series (378 points): ATH $46,765,513 on 2026-01-18 (now 46.63% below), $26,414,872 on 2026-08-21 (30d, -5.52%), $27,544,326 on 2026-06-22 (90d, -9.39%). Our own facts scanner reads 30d at -1.2%; the two disagree and we have not reconciled the snapshot times, so treat the DefiLlama series as the one verified here. - Hot / cold split: Hot USDC fell from $1,357,510.70 (2026-08-03) to 456,757.733094 (this call; it moves every few minutes as withdrawals settle). Cold USDC is exactly 24,000,000.000000, unchanged to the cent since August, which reads as a manually managed reserve rather than a live sweep. USDC is 1.87% hot, 98.13% cold. SOL is a different story: the published SOL cold vault
8nFeyzTFhUXp11raJkSSvZWn9GXDjcrGq8LuzP9aKtg8holds 0.001 SOL, so 100% of the 4,577.40 SOL sits hot, up from 4,033.55 in August. The hot USDC token account's SPL authority is9Gdmhq4G...(the Pacifica-ownedcentral_state); the cold USDC token account's authority is5kwCMKjE..., which is an off-curve, system-owned, zero-data account. All four published addresses still match the Fund Security page verbatim and all four are off curve. Which program derives the cold PDAs: refresh, still not resolvable from account state alone. - Audits: Two, both in BlockSec's official GitHub repository. (1)
blocksec_pacifica_fi_perp_contracts_signed_20250424.pdf, HTTP 200, 1,713,273 bytes, client Pacifica Fi, target Perp Contracts, v1.0, 2025-04-24, scopeperp-contracts/programs/pacifica-mainnet/src/lib.rs, commitsb4c48da19e17efd015bc5c57a91e13731ee199e8(Version 1) thene3631ce1efe06308e838a8785c0ac71742346b4d(Version 2). Findings: 2 Low (potential DoS inbatch_withdraw(), missing minimum value check indeposit()), 4 recommendations, 2 notes. The table numbers eight rows; IDs 1 to 6 carry status Confirmed, which BlockSec defines as recognised by the client but not fixed, and IDs 7 and 8 (the notes) carry no status. The PDF's own digital signature block is dated 2026-04-30 13:49:25+0000, a year after the report date. Added toblocksecteam/audit-reportson 2026-04-30 (commitc40b6238fd, "Add missing public reports"). Now linked correctly fromdocs.pacifica.fi/other/audits.md; the olddocs.pacifica.fi/images/tjfRMLcTkAHqN7rfxcos.pdfpath still returns 404, and a February 2026 archive snapshot shows the page then pointed at a GitBook-hosted copy. (2)blocksec_pacifica_bridge_contract_signed_20260401.pdf, 1,683,016 bytes, target Bridge Contract (programs/sol-bridge,programs/spl-bridge), 2026-04-01, added to the repo 2026-04-07; 3 non-critical issues, no criticals. Only the bridge report is an indexed entry onblocksec.com/audit-report; the "Perp Contracts" entries on that index belong to HertzFlow. Repospacifica-fi/perp-contractsandpacifica-fi/bridge-contractboth return 404, so no verified build ties audited commits to deployed bytecode: refresh. Whether the 2 Low findings were fixed in the Version 2 commit: refresh. Audit of the matching engine, oracle, liquidation engine, money market or vaults: none found. A Bug Bounty Program page exists in the docs, with Critical to Low tiers paid in USDC to the reporter's Pacifica account. - Admin / governance: Program
PCFA5iYgmqK6MqPhWNKg7Yv7auX7VZ4Cx7T1eJyrAMH, BPF upgradeable loader, program dataFC5zcmPmqo5188fLVga2HN5QZ1T1ojhNneVkFdu6eF1E(tag 3, space 323,797), upgrade authority present and unchanged atBzZgjpBFHb19vWjBRoM5eVy6qnUjbGH5tMJrnN7cA47a(on curve, verified against off-curve controls this run). Program data account carries exactly two signatures in its whole history, slots 343778047 and 343778781, both 2025-05-31 at 23:01:26 and 23:06:14 UTC, so the program has never been upgraded in 15.6 months; last August's "roughly 14 months, exact date refresh" is now dated exactly.central_stateat9Gdmhq4Gv1LnNMp7aiS1HSVd7pNnXNMsbuXALCQRmGjY(185 bytes, owned by the Pacifica program, discriminatorc93123e704a4cd5b) holds five non-zero pubkeys: offset 8BzZgjpBF...(on curve, same as the upgrade authority), offset 40FHFvAvN1F6wzBGVwnAFM4kJ26BmM8rq1f7zhpTewnQ8r(on curve), offset 72CBXdrgyEp8QeTHYDCsifHaT7VJZwHB29ErpVM5uzDHC4(off curve), offset 104PvqRTZDnCaNLp1FHQX1a3e2JcwwYaAhViPX2Dg3SW2j(on curve), offset 136G5F6jvZYkUs7hAEFz5NyNAmtENRFufVRYpjUgqwQ9KEg(off curve). The accounts at offsets 72, 104 and 136 do not exist on chain (getAccountInfo returns null), and their field names are refresh.FHFvAvN1...was observed as the sole signer ofBatchWithdrawin tx5DTLfRXYFNjtZmaK8Jp2MKLFw4MS9DNEvshQdKuGjRjqTgFvD7ayPMpA7ZfaAX34wrE4tDZYJssktysATfCb1yf6(2026-09-19 16:39:25 UTC, slot 448464088), with the token balance delta confirming 999.0 USDC out of72R843Xw...; further signatures at 16:26:25, 16:32:55 and 17:15:25.BzZgjpBF...last signed 2026-04-08. BlockSec note 2.3.1, verbatim: privileged roles "are set to govern and regulate the system-wide operation. Additionally, accounts with withdraw_authority permission can withdraw user assets from the vault unlimitedly. If the private keys of them are lost or maliciously exploited, it could potentially lead to losses for users." Docs still claim distributed multi-sig and still say "transitioning to"; the $5M daily hot limit there is framed as an example. Multisig address, threshold, signers, timelock, spending-limit value: refresh. - Oracle: Unchanged. Pacifica-computed, 3-second updates, weighted average of USDT-denominated CEX prices (Binance 2, OKX 1, Bybit 1, Hyperliquid 1). Mark price is the median of that oracle, the median of Pacifica's own best bid, best ask and last trade, and perpetual prices from major exchanges. No third-party oracle, no on-chain attestation.
- Market size: 77 markets (75 in August), $466,468,378 of 24h volume and about $96.09M of open interest, computed from
api.pacifica.fi/api/v1/info/pricesthis run; both are rolling figures. Docs Overview still says "over thirty-five perpetual markets", stale against the API for a second consecutive report, and quotes "3x to 50x" leverage while the API caps SOL and HYPE at 20x. - Funding (this run, hourly): HYPE
0.0000125, SOL0.0000125, both with the same next print, so longs pay our shorts on both legs at about +10.95% annualised simple (11.57% compounded). Both markets on the identical figure suggests a baseline rather than a premium-driven rate. In August SOL was negative and our short was paying. - Liquidations: Three tiers, confirmed from the docs this run. Market liquidation below maintenance margin, with a chunked IOC unwind and a liquidation fee of
max(0.75%, maintenance_margin_ratio * 0.4); backstop liquidation below two-thirds of maintenance margin, where positions and remaining collateral transfer to a Pacifica backstop liquidator that declines a named list of experimental markets (SOL and HYPE are not on it); then auto-deleveraging of profitable opposing positions if equity falls below zero. Maintenance margin is half the initial requirement,1 / max_leverage. A separate spot-insolvency deleveraging path triggers at 95% money-market utilisation. Insurance fund still not described anywhere in the docs: refresh. - Withdrawals: Engine processed from the hot account. $500,000 per account per 24 hours for USDC and $500,000 per asset per day for spot, with deposits capped at $500,000 of account equity and spot deposits at $50,000 per asset per day, all stated as applying "during Closed Beta", plus a time-based exchange-wide cap across USDC and spot combined whose value is not published. Minimum withdrawal $1, fee about $1. Last month's report and the draft both printed $250k; the current docs contain no such figure. No permissionless escape hatch documented: refresh.
- Owned position (orchestrator input, marks verified this run): short 69 HYPE at a $92.886944 mark ($6,409.20) and short 41.12 SOL at a $112.062878 mark ($4,608.03), $11,017.22 of short notional, 100% of the portfolio's short side on one venue. Live account equity, margin used and liquidation prices were not pulled this run: refresh.
- Holder concentration: No native token. DefiLlama returns
gecko_id: None,cmcId: None,symbol: "-". X scan through 2026-09-19 found only TGE speculation and OTC point pricing. refresh at launch. - Recent news / incidents: Zero Pacifica entries across 1,273 rows of
api.llama.fi/hacks. X scan of@pacifica_fifrom 2026-08-01 to 2026-09-19 found no outage, withdrawal halt, socialised loss, ADL event, exploit, TGE, beta exit or multisig announcement; the account's posts in that window are feature releases (OCO orders, trailing stops, referrals) on 15 to 18 September, and its bio reads "Closed Beta, now live". A broad X scan of the wider conversation found no user reports of stuck or delayed withdrawals, downtime, liquidation disputes, bad debt or insolvency. Our incidents table has recorded nothing on Pacifica since 2026-08-03. The docs still carry a Close Beta Guide page whose entire content is a Canva edit link. The 2026-01-21 Mithril third-party API-key breach remains the only disclosure on record. Name-collision warning still applies: Pacifica is also a city in California and Pacific Life is an insurer.
Maintained monthly. Methodology: DeFi Research Instruction v2.