LIVE ADA-- MCAP-- TVL-- STAKE-- EPOCH-- Cardano & Midnight 総合情報ポータル
SIPO
𝕏 🔍
HOME › Weekly Brief EN › Weekly Brief #25 | Cardano’s k Poll: Three Pools Outweigh Ten by Delegated Stake
Weekly Brief EN

Weekly Brief #25 | Cardano’s k Poll: Three Pools Outweigh Ten by Delegated Stake

2026-09-26SIPO

The SPO poll asking whether stakePoolTargetNum (k) should be raised from 500 to 1000 received its first explicit no votes this week. Ten pools have voted yes; three have voted no.

By headcount, yes leads by more than three to one. But the criterion the proposal itself declares is weighted by stake, and on that measure yes holds 53,185,274 ada against no’s 95,984,098 ada, a gap of 1.80 times.

The third no vote came from the person who submitted the poll. At 04:42:19 JST on September 26, the pool STR8, run by the Chris named as the proposal’s author, voted no and published a rationale.

No new governance action was submitted on chain during the week. The same three actions remain pending, and only the votes moved. On the OpenZeppelin withdrawal, SIPO’s DRep voted no at 10:38:08 JST on September 25. The previous issue said SIPO would decide by October 9; the answer arrived two weeks early.

Executive Summary

  • No new governance actions were submitted: since epoch 657 began (06:44:51 JST on September 22), nothing has been put on chain. Three actions remain pending, and the proposal deposits held on chain are back to 300,000 ada = 100,000 ada × 3.
  • The k poll drew its first no votes: explicit yes went from 2 pools to 10, explicit no from 0 to 3. Yes stake rose from 2,114,251 ada to 53,185,274 ada.
  • Headcount and stake point in opposite directions: the ten yes pools hold 53,185,274 ada; the three no pools hold 95,984,098 ada. The declared criterion is stake-weighted, so reading the vote count gets the conclusion backwards.
  • The proposer voted against his own poll: as an SPO at 04:42:19 JST on September 26, and as a DRep 23 minutes later. His rationale says “we have not exhausted the current k=500 setting,” and that after submitting he read the case against and agrees with many of its arguments.
  • The 2023 poll had the same shape: in the tally the proposal cites, the option containing k=1000 took 69.64% by pool count and about 41.1% by participating stake.
  • Clearing 50% needs roughly 4.7 billion ada: the denominator is yes plus non-voting stake, 9,400,862,467 ada. Yes currently stands at 0.57%, and SIPO’s three pools have not voted.
  • The k=1000 saturation point is about 36.77 million ada: a mechanical figure, circulating supply divided by k. SIPO’s first pool (53,862,342 ada) sits at 1.465 times that line.
  • SIPO’s DRep voted no on the OpenZeppelin action: 10:38:08 JST on September 25. The rationale is recorded on chain in English and Japanese and rests on four points: overlap, an undecided language, conversion surplus, and bundling. Opposing stake is now 9.76 times the stake in favour.
  • minPoolCost reversed direction: the previous issue noted the last five votes were all no. This week’s 23 votes break down as 14 yes, 4 no, 4 abstain, 1 committee, and by stake yes now holds 535,228,477 ada against 490,675,179 ada of cast no votes.
  • The committee’s second vote came from Tingvard, on three actions at once: at 02:26:29 JST on September 22 — yes on minPoolCost, yes on OpenZeppelin, abstain on the k poll. A committee yes is a finding of constitutionality, not support for the proposal.
  • Midnight’s node received a security patch covering mainnet: node-1.0.300 (21:29:49 JST on September 22). The release notes say “security patch” in as many words, and every validator must move to this binary first or fall off the chain.
  • All four assets were up: ADA +13.19%, NIGHT +3.48%, BTC +3.32%, ETH +2.11%. Most of the move came on a single day, September 22.

1. Market Pulse

Values from September 19 to September 26.

AssetSep 19Sep 26Week
ADA$0.225627$0.255395+13.19%
BTC$81,238$83,933+3.32%
ETH$2,630.63$2,686.10+2.11%
NIGHT$0.02392995$0.02476157+3.48%

All four are positive, the same shape as the previous issue, for a second week. The composition differs. Almost all of ADA’s gain landed on a single day, September 22, up 8.36% against the prior day. It then gave back 5.65% on September 24 and recovered 4.83% on September 25. On the morning of September 26, ADA was up 3.36% day over day, with SOL at +4.72% and BTC at −0.12% at the same moment.

ADA taken exactly at the epoch boundary moved further. The epoch 656 boundary (morning of September 17) was $0.19436574 and the epoch 657 boundary (morning of September 22) was $0.24542512, or +26.27%. In yen that is ¥30.36 to ¥38.61, with USD/JPY moving from 156.22 to 157.30. Boundary values and the morning readings are taken at different times, so they are not mixed into the same table.

Network figures, for the record. The staking rate went from 56.45% to 56.34%. Delegator count fell by 958 to 1,336,450, a wider drop than the previous week’s 584. Pools that produced a block rose from 935 to 937. The treasury stands at 1,363,711,219.21 ada (+3,631,451.92 ada from the previous issue), and epoch 656 carried 21,215 blocks and 127,132 transactions, more than 40% above epoch 655’s 89,353.

2. Not a Single New Action Was Submitted

Since epoch 657 began, no governance action has been submitted on chain. The same three from the previous issue remain pending.

ActionTypeSubmittedExpires (JST)
Withdraw 11,787,063 ada for the OpenZeppelin Stack administered by IntersectTreasury withdrawalEpoch 65406:44:51, October 12
Reduce minPoolCost to 75 adaParameter changeEpoch 65406:44:51, October 12
Should stakePoolTargetNum (k) be raised from 500 to 1000? (SPO poll)Info ActionEpoch 65606:44:51, October 22

The deposits held on chain show the same thing. Each action carries a 100,000 ada deposit, and as of the previous issue 400,000 ada was held. This week it is 300,000 ada. Deposit handling for the two actions that expired at the epoch 656 boundary — Reimburse Ikigai Info Governance Action Deposit. and Governance Incentives Framework 2026 — completed in epoch 657, leaving only the three pending actions.

No constitutional amendment action appeared on chain either. The stage before submission did move. Per Intersect’s weekly update #129, the constitutional amendment portal currently holds seven proposals, none yet at the “ready” stage, with four of them active in the past three weeks. The most discussed is CAP-3 (evolve the Net Change Limit into a strategic spending plan), with sixteen comments.

One of those proposals connects directly to an action now live on chain.

“CAP-10, Require SPO Approval for Changes to Stake Pool Economic Parameters, came out of SPO discussion after the last minPoolCost action and is directly relevant to the one now live.”

CAP-10 is also what the k poll’s own text references. The next review session is 14:00 UTC on October 1.

In a week when no new entrance opened, what moved inside the three that were already there. That is the record below.

3. The k Poll Drew Its First No Votes

Should stakePoolTargetNum (k) be raised from 500 to 1000? (SPO poll) entered its ninth day since submission. At the time of the previous issue, explicit yes stood at two pools and explicit no at zero.

CategoryPrevious issue (Sep 19)This issue (Sep 26)
SPO explicit yes2 pools (0.02%, 2,114,251 ada)10 pools (0.57%, 53,185,274 ada)
SPO explicit no03 pools, 95,984,098 ada
Non-voting pool stake9,604,088,251 ada9,251,693,095 ada
DRep yes4 votes (0.06%, 3,061,763 ada)18 votes (2.18%, 101,317,767 ada)
DRep cast no3 votes, 171,938,311 ada14 votes, 212,318,456 ada
DRep abstain1 vote, 445,777,028 ada8 votes, 542,686,821 ada
Committee0 votes1 abstain

Eleven SPO votes landed during the period: eight yes and three no. The three no pools are pool1u6kpt8h… (02:25:06 on September 22), pool1853ffwh… (02:27:52 the same day) and STR8 (04:42:19 on September 26). Three of the yes votes were recorded in the same second, 10:03:06 on September 22 (pool149d5lzg…, pool12rs385d…, pool1wyfau3r…). Whether they sit in the same operator’s hands is not something the ledger discloses comprehensively. What can be written is that they were recorded in the same second.

The DRep side moved too, from 4 yes votes to 18 and from 3 cast no votes to 14. But the proposal text declares that DRep votes are not included in determining this poll’s outcome. The numbers accumulate; the calculation that decides whether a mandate exists does not use them.

One committee vote arrived: an abstain at 02:26:29 JST on September 22. That member voted on two other actions in the same transaction (chapter 7).

4. The Proposer Voted Against His Own Poll

The third no vote came from the person who submitted this poll.

The proposal metadata names Chris as its author, and his pool’s ticker is STR8. That pool voted no at 04:42:19 JST on September 26, and 23 minutes later, at 05:05:23, the same person voted no as a DRep as well. Both carry rationales; the DRep one is a single line pointing to the SPO rationale.

The rationale begins with why he filed the poll in the first place.

“I submitted this Info Action because I wanted a clear assessment of SPO sentiment using the established CIP-1694 SPO voting semantics.”

It then sets out what he did not have at the time.

“When I created the Info Action, the draft material available to me linked the advocacy paper arguing for k=1000, but I did not have access to the corresponding detailed paper arguing against it. To preserve neutrality, I could not reasonably reproduce an advocacy section containing only the pro paper, so I omitted that section and instead tried to present the substantive arguments from both perspectives as neutrally as possible.”

“I only became aware of Neil Davies’ detailed case against raising k after submitting this Info Action. Having now read it in full, I greatly appreciate the work and analysis that went into it and agree with many of its arguments.”

On the substance, he argues that lowering the saturation point does not directly help small pools that sit below saturation; any benefit depends on displaced stake actually redelegating to independent pools that need it.

“Large custodians and multi-pool operators can simply redistribute stake among additional pools under their own control. That increases the pool count without necessarily increasing the number of independent entities controlling consensus.”

“We have not exhausted the current k=500 setting. k represents an idealized equilibrium target for saturation-scale pools, yet Cardano is still far from an equilibrium of 500 saturated pools. With roughly 21.4B ADA of active stake, only about 285 pools could currently be fully saturated at the k=500 saturation point, while the network already has around 1,600 active pools.”

He also records his own position.

“I would personally be pushed toward operating a second pool. I have spent years building trust and organically attracting delegation to my pool. If k is doubled, my pool would become oversaturated and I would effectively be forced to establish another pool to avoid losing the result of years of work.”

The alternatives he names are sustainable pool economics (a minimum margin of 5–10%), effective pledge mechanisms such as CIP-50, pool alliances, and measures addressing inactive and undelegated stake.

What can be written here is what the vote and the rationale say. As to why the proposer did not hold this position when he filed, the rationale answers for itself: he learned of the detailed case against after submitting. Beyond that, the record says nothing.

5. Counting Heads, Counting Weight

This is the centre of the week.

By vote count, ten pools in favour against three opposed is a majority of more than three to one. The criterion the proposal declares for itself reads differently.

“Sufficient support is established only if explicit SPO YES votes represent more than 50% of the applicable active SPO voting stake.” “An explicit ABSTAIN vote is excluded from active voting stake, while registered stake that does not vote behaves like a NO vote.”

Stake-weighted, by declaration. Applying that measure flips the order.

MeasureYesNoWhich is the majority
Number of pools103Yes
Delegated stake53,185,274 ada95,984,098 adaNo (1.80×)

The bottom row is the one the criterion uses. Reading only the headline vote count gets the conclusion backwards.

There is a precedent, and the proposal text cites it itself: the 2023 SPO poll. Provisional results reported 797 pools voting, with the option containing k=1000 taking 69.64% by pool count and about 41.1% by participating stake, while the option containing k=500 took 21.08% and about 44.2%. Count heads and one side is the majority; count weight and the other is. The reason this poll repeats, in its own text, that voting power is stake-weighted and that yes must exceed 50% is that it comes after that experience. The same shape has now appeared in week one.

The denominator, for the record. Yes at 53,185,274 ada plus 9,347,677,193 ada of cast no and non-voting stake gives 9,400,862,467 ada of active voting stake. Yes stands at 0.57%. Clearing 50% would take roughly 4.7 billion ada of explicit yes. The 617 pools that have chosen “always abstain,” about 11.9 billion ada, sit outside this denominator. The deadline is 06:44:51 JST on October 22, and the proposal text asks operators who disagree to vote an explicit no and file a rationale. SIPO’s three pools have not voted.

6. The Saturation Line and SIPO’s First Pool

This touches SIPO’s own interests, so the figures come first. These are mechanical: the on-chain circulating supply of 36,767,374,068 ada on the morning of September 26, divided by k.

kSaturation point (supply ÷ k)SIPO’s three pools (epoch 656 boundary)
500 (current)about 73,534,748 adaall three below
1000 (proposed)about 36,767,374 adaSIPO 53,862,342 ada = over (1.465×) / SIPO2 25,046,488 ada / SIPO3 24,866,921 ada

In the same table last week, SIPO’s delegated stake was 57,052,790 ada. This week it is 53,862,342 ada, a fall of 3,190,448 ada across one epoch, with the delegator count unchanged at 1,113. Closer to the line than last week, but the k=1000 saturation point remains below it, and SIPO sits on the side that loses from the change.

Saturation caps rewards, not the right to produce blocks. That is unchanged from the previous issue. With that said, SIPO is a party to this poll, so its position goes on the record before its vote: k=1000 is an unfavourable change for SIPO’s first pool. The proposer’s line quoted above, about being pushed toward operating a second pool, describes the position SIPO is in as well. Anyone delegating can check today, without waiting for the result, which side of that line their pool sits on.

7. The Committee’s Second Vote Came From Tingvard, on Three Actions at Once

The previous issue described the new constitutional committee’s first vote landing on four actions simultaneously. The second vote arrived in the same shape.

Recorded (JST)Member (term)Votes cast
September 12, 17:10:17cc_hot1qgsl88a… (799)Four at once — GIF yes / Ikigai yes / minPoolCost yes / OpenZeppelin yes
September 22, 02:26:29Tingvard (726)Three at once — minPoolCost yes / OpenZeppelin yes / k poll abstain

Each of the three votes carries a rationale, and the member’s name comes from those rationales themselves.

The reading matters here. A committee yes does not mean the proposal is good.

“Tingvard judges the ‘Reduce minPoolCost from 170 ada to 75 ada’ Parameter Update governance action constitutional.”

The grounds are procedural: that PCP-006 was published on March 30, 2026 with the change not intended for submission before June 30, satisfying the required 90-day notice period; that the Intersect Technical Steering Committee ratified the reduction on July 9; and that the proposed value satisfies the minPoolCost guardrails (positive, and well below the 500,000,000 Lovelace ceiling).

The OpenZeppelin yes is the same in kind, grounded in the detailed budget, milestone-based disbursement, funds held in accounts that cannot be staked and are delegated to the auto-abstain DRep, and a public dashboard for community auditing. It is a finding that the action does not contravene the constitution, not a judgment that the request is sound.

On the k poll the vote was abstain, with its own reason.

“Tingvard does not consider an Info Action to be an appropriate subject for a Constitutional Committee determination. Its purpose is to gather information and signal stakeholder sentiment rather than enact a change to the Cardano ledger.” “Tingvard therefore abstains without expressing support for or opposition to increasing stakePoolTargetNum from 500 to 1000.”

That member was the only one to vote during the period. As a result, the committee votes on minPoolCost and on OpenZeppelin both stand at 2 yes and 5 not voted, an approval rate of 28.57%. That is double the previous issue’s 14.29%, and still three members short of the two thirds (66.67%) required. Both expire on October 12, with three epochs remaining.

8. SIPO Voted No on the OpenZeppelin Action

At 10:38:08 JST on September 25, SIPO’s DRep voted no on Withdraw 11,787,063 ada for the OpenZeppelin Stack administered by Intersect. The rationale is recorded on chain in English and Japanese.

The previous issue said: “SIPO will check the state of things around October 5 and decide by October 9.” The vote was cast on September 25, two weeks ahead of that. The rationale does not explain why it came earlier, so only the difference in dates is recorded here.

The rationale opens with what it concedes: this is a vote against this version of the request, not against OpenZeppelin or the work. An audited, standardised contracts library carries a level of trust that EVM developers already extend to OpenZeppelin, and that is not something Cardano can create on its own, it says. On accountability, it lists Intersect administration on the Sundae Labs treasury-contracts framework, an independent Oversight Committee, milestone payments released only against public evidence, and unspent funds returning to the treasury at contract expiry, calling the design sound and consistent with structures SIPO has approved before. It also records OpenZeppelin’s statement that its commitment to Cardano does not depend on this vote.

The grounds for the no vote are four.

  1. Overlap with work SIPO has already supported. In May 2026 SIPO voted yes on IO’s Developer Experience Initiative (about 3.6M ada), whose deliverables include an OpenZeppelin-style ContractsLibrary. The rationale at the time set an explicit expectation of avoiding parallel reinvention and fragmentation of the builder community. That library was updated in September 2026 and remains under active public development. This proposal asks for a further 11.8M ada for a library of comparable purpose, but does not reference that work, publish a gap analysis, or say which library builders should treat as the standard.
  2. A core technical choice is deferred. The smart contract language is to be determined during an initial evaluation phase, after funding is approved. Cost, interoperability with existing libraries, and the risk of splitting the builder base cannot be judged before that choice is made.
  3. Currency terms are incomplete. The delivery budget is USD 1,831,000, converted into the ada request at a fixed reference rate of USD 0.16, and converted to a USD stablecoin at contract signature. If ada trades above the reference rate at that point, the proceeds exceed the USD budget, and the proposal does not state what happens to the surplus. In addition, 20% of the delivery budget is released at signature, before any technical milestone is accepted.
  4. The scope is bundled and the Cardano adoption bar is low. Three reference implementations, a contracts library and a security retainer are combined into a single request that DReps cannot approve in part. The only Cardano-side usage criterion in the milestone table is that at least one independent Cardano developer has reviewed the library and provided written feedback.

The rationale also sets out what would earn a yes on resubmission: publish a gap analysis and a division of labour with IO’s ContractsLibrary; name the target language; put the library, security and standards work first, with reference implementations following on evidence of use; state that any conversion surplus above the USD budget returns to the treasury; reduce the upfront payment; and set adoption criteria measured by independent Cardano projects. Six conditions.

Disclosure. SIPO voted yes, in May 2026, to the very funding it now cites as overlapping. It is objecting to a later proposal on the strength of deliverables from money it helped approve. That is why the rationale frames the order of decision as a condition: evaluate what the funded programme delivers, then fund the gap that remains.

9. Opposing Stake Reached 9.76 Times the Stake in Favour

The full vote on the OpenZeppelin action, for the record.

CategoryPrevious issue (Sep 19)This issue (Sep 26)Threshold
DRep yes2.14% (7 votes, 111,152,249 ada)2.32% (9 votes, 119,111,932 ada)67%
DRep cast no29 votes, 930,769,896 ada41 votes, 1,162,262,851 ada—
DRep abstain1 vote, 40,205,215 ada7 votes, 43,200,391 adaoutside denominator
Committee1 yes, 6 not voted2 yes, 5 not voted2/3

Twenty DRep votes landed during the period: two yes, twelve no, six abstain. By stake, cast no votes now stand at 9.76 times the stake in favour. The previous issue described a gap of more than eight times, so it has widened. Abstentions rose from one vote to seven.

The treasury side is unchanged. Withdrawals enacted since epoch 613 total 34 actions and 457,395,629 ada against the 2026 cap of 500,000,000 ada, leaving 42,604,371 ada. If this action passes, the total reaches 469,182,692 ada, or 93.84% of the cap.

10. minPoolCost Turned Back

The previous issue wrote of this action: “the last five votes (22:07 on September 17 through 18:14 on September 18) were all no.” This week’s 23 votes have the opposite shape.

CategoryPrevious issue (Sep 19)This issue (Sep 26)Threshold
DRep yes6.26% (22 votes, 294,844,280 ada)11.77% (36 votes, 535,228,477 ada)67%
DRep cast no20 votes, 469,841,434 ada24 votes, 490,675,179 ada—
DRep abstain2 votes, 533,897,596 ada6 votes, 637,005,443 adaoutside denominator
SPO0 votes (not required)0 votes (not required)0%
Committee1 yes, 6 not voted2 yes, 5 not voted2/3

The period breaks down as 14 yes, 4 no, 4 abstain, plus the one committee yes. By stake, yes at 535,228,477 ada now exceeds cast no votes at 490,675,179 ada. The previous issue had 294,844,280 ada against 469,841,434 ada with no on top, so the two swapped places within a week.

The passing threshold remains far off. Yes stands at 11.77% against a threshold of 67%. As of September 23, DReps who had not voted held 74.38%. What changed was direction, not depth.

There was a public session as well. Cardano Governance Hour #11 on September 23 took this action as its whole subject. The speaker was Ryan Wiley (Cerkoryn), a member of the SPO Incentives Working Group and the proposal’s author, and the announcement summarised the case as arguing “that declining block rewards have increased the relative burden of the fixed cost on smaller stake pools.”

SIPO’s position is unchanged from the previous issue. It voted a constructive abstain on September 15, and the full rationale is published at SIPO DRep: abstaining on the standalone minPoolCost 75 ADA action. All three SIPO pools declare a fixed cost of 340 ada, so the change has no mechanical effect on them, and SIPO sits with the 340 incumbents.

11. The Ledger in Year Nine, and Node Diversity

September 23 marked nine years since Cardano’s first block. The Cardano Foundation’s figures read:

2,890 stake pools. 13,978,119 blocks. 123,982,111 transactions. A constitution, ratified on-chain. 36,322 votes on 158 governance actions.

The action count matches what the ledger returns today: 158. The pool count is in the same range; at the epoch 656 boundary it was 2,893.

The same ledger returns other cross-sections. Of the 2,890 pools, 2,671 hold delegation and 775 produced a block in the most recent epoch. At the epoch 656 boundary, 2,674 pools held delegation and 937 produced a block. These are taken at different moments, so lining them up does not permit subtraction. Either cross-section still shows a gap between the number registered and the number through which rewards and blocks actually flow.

What those pools run is changing too. Intersect’s weekly update #129 recorded a milestone.

“Dingo, a Cardano node implementation written in Go and developed by Blink Labs, validated blocks accepted by Cardano Mainnet this week, a significant milestone for node diversity efforts on Cardano.”

Dingo shipped five versions this week. Amaru, the Rust implementation, shipped two, with its latest target milestone recorded as “minimum viable block producer.” Per the public recap of September’s ambassador call, Amaru runs nominally on 2 GB to 4 GB of RAM (against the 16–24 GB and up the Haskell node requires), and its conformance with the Haskell node is checked across 282 validation scenarios, with a block-production alpha set for October / Q4. The Node Diversity Celebration Day is October 6, in Singapore and online.

The anniversary figures and the saturation point the k poll asks about are two columns of the same ledger. One counts the total accumulated over nine years; the other asks how many pools that total should be divided among.

12. Nodes and Tooling

cardano-node — no release this week, and 11.1.2 is the priority

No new version of the node itself appeared during the week. The most recent remains 11.1.2 from September 17. Intersect’s weekly update #129 makes it one of the week’s two deadlines.

“Cardano node 11.1.2 is out, and SPOs are asked to prioritize it. The release addresses an issue identified during testing, and it supersedes 11.1.1, so operators who have updated need to do so again.”

Neither the release notes nor #129 uses the word “security.” That said, what 11.1.2 carries over from 11.1.1 is substantial: removal of the V1 LedgerDB and the LMDB storage backend (LMDB users must switch backend), predictable Mithril-compatible ledger snapshots, faster transaction forwarding and Plutus validation, and initial network and consensus support for Peras, off by default. Removal of the old tracing system completed in 11.1.1, so monitoring setups may need work. Anyone running a Mithril signer should upgrade to 2630.1-hotfix (mithril-signer-1.1.9) at the same time as the node, per #129.

How far it has spread has also been observed: as of September 18, 20% of nodes reporting a version had completed the upgrade. The qualifier matters — the denominator is nodes that report, not all nodes. Cardano has no mechanism to push a new node to every machine at once. Each operator upgrades and restarts their own server, and only then does a release become part of the network. Decentralisation is usually discussed through pool counts and delegation concentration; what share has upgraded is another cross-section of the same thing.

Hydra — a bug that locked deposits has been fixed

This week’s report covers one user-reported case. A user who deposited two UTxOs into an open head through the Mesh SDK found neither reflected on layer 2 after three hours, both gone from their L1 wallet, sitting unspendable at the deposit script.

The cause sits in fee handling. When a deposit’s own value did not cover the minimum ADA of its output, the process topped the output up but left the inline datum, which records the deposited outputs, at its original value. The value-versus-datum check then rejected the deposit and the node never observed it. The fix validates before submission and fails fast with an error naming the shortfall, and adds a way to decode already-stuck deposits for manual recovery.

The report’s word is “bug,” and no security advisory has been issued. That said, from a user’s perspective it is the kind of bug where funds do not come back. Anyone running Hydra should treat this week as an upgrade week.

Leios — the database split in two

The same week’s consensus team update explains that the Leios prototype has split its database into two files: data a node can still roll back, and data that is settled. A certified Endorser Block moves from the first to the second in the background. Operators can put the two files on different disks and give the busier one the faster disk. Garbage collection now runs in two paced steps, one thread marking rows and another deleting them in batches, so deletion does not hold up the rest of the node.

Three prototypes shipped this week (September 21, 22 and 25). The Leios demo in the previous issue was a figure — 250 TxkB per second on a local cluster. This week’s changes are not about speed but about not falling over and not jamming.

Wallets and around them

Lace announced the contents of 2.3 this week, and the extension shipped 2.4.0 (September 23) and 2.4.1 (September 25); 2.4.1 adds support for version 8 of Ledger’s Cardano app. The Midnight-side changes are covered in the next chapter. cardano-js-sdk updated 14 packages at once, and Mithril had no stable release this week.

13. Midnight Watch

Node 1.0.300 — a security patch that covers mainnet

This is the item that calls for hands-on work this week. At 21:29:49 JST on September 22, node-1.0.300 was published. The opening line of the release notes states its nature outright.

“A security patch release on the 1.0.x line, for every environment including mainnet.“

Two things are fixed. One is a critical tar advisory (GHSA-23hp-3jrh-7fpw) reachable through the npm CLI bundled in the toolkit image. The other concerns the images themselves, and the wording is specific.

“Removes gdb from the node, toolkit and hardfork-test-upgrader images, which had been handing anyone with exec access ptrace-based memory inspection of a running validator, key material included.”

The order of operations is specified, not suggested.

“Upgrading is a coordinated binary-plus-runtime operation: every validator must be on this binary before set_code is enacted.” “A validator still running an older binary cannot instantiate the new runtime. It stops importing blocks at the set_code and falls off the chain.” “The ordering is not advisory.”

spec_version moves from 1_000_000 to 1_000_300 and system_version from 1 to 3. No state reset is required on mainnet, Preview or Preprod; only devnet’s genesis was regenerated. The open question from the previous issue, about how the 1.0.x and 2.1.0 version numbers relate, is settled by this release: the maintenance line follows 1.0.2 with 1.0.300. Operators tracking the development line should stay on 2.1.0-rc.2, the release notes specify.

Mainnet private contracts are “soon” to be enabled

At 14:36 JST on September 19, Sebastien Guillemot, CTO of the Midnight Foundation, posted that smart contract deployment would soon be enabled on Midnight mainnet.

“Smart contract deployment is soon going to be enabled on Midnight mainnet. Excited for private smart contracts! (incl. ability to mint private custom tokens) Thanks to all the builders who helped get here.”

One point deserves plain statement. No enablement date and no official release notes have appeared. The node release notes carry nothing on it, and node-1.0.300 lists “New features: None.” The official blog published nothing this week either. What can be confirmed is the content of the post.

The announcement when mainnet went live at the end of March said application deployment would proceed in stages rather than opening all at once. “Soon to be enabled” reads as one of those planned stages drawing near.

Improvement proposals — MIP-0018, MPS-0043, and an index that caught up

Two documents landed in the improvement-proposals repository on the morning of September 22.

MIP-0018 (on-chain token metadata) is an arrangement for a contract to declare a token’s name and symbol itself, as a public event. Midnight’s native tokens are distinguished by a “colour” derived from the issuing contract’s address and a domain separator, but there is no standard way to say what that colour represents. The consequence, the proposal writes, is that wallets display a 64-character hex string, explorers cannot label tokens, and applications hard-code their own lists. The mechanism is an event with a fixed 256-byte payload, and the design is stated to follow Ethereum’s EIP-7496. The point is that the declaring party is the contract that issued the token; no one can name someone else’s token on its behalf. Status is Proposed, the first stage.

MPS-0043 (selective disclosure to a chosen party) landed the same morning, raising the absence of a mechanism to reveal shielded data to chosen recipients.

The index mismatch noted in the previous issue is also resolved. Then, the document existed but the numbering index stopped at MIP-0016. This week’s index carries both MIP-0017 and MIP-0018, and the highest MPS number is MPS-0043.

Deployment requests — shipping with the treasury circuits removed

On September 23, Shadow Labz’s request to deploy three contracts to mainnet was approved. The set covers voting, feedback collection and DAO operation while preserving privacy, and none of them receive, hold or send native tokens.

What stands out is how ShadowDAO v1 was shipped. Two treasury-related circuits — accepting funds, and executing approved spending — are removed from the deployed build. Ship the build without the money-handling circuits first, and split the custody part into a separate request. The submission also records its own weakness: in a vote with very few participants, an observer correlating transaction timing could infer choices.

The repository’s review framework states plainly: “this review is not an audit.” Approval concerns whether a contract may be deployed, not a guarantee that its code is safe.

The same distinction surfaced elsewhere. The Cardano lending protocol FluidTokens announced on September 23 that its Midnight smart contracts are being audited. The post is one line, and neither the firm conducting the audit nor a completion date is stated. An audit and a deployment review are separate gates.

Privacy applied to a token sale

Midnight’s official account introduced Noctis Zone in the small hours of September 23, a token launchpad built by a Build Club participant. The first 24 hours of a sale run invisibly, at a single fixed price for everyone, and afterwards a zero-knowledge proof shows that no wallet exceeded its cap. With order and price both hidden, there is nothing to gain by front-running. The proof is written to Cardano’s L1. It currently runs on the pre-production test network.

Lace Midnight Preview stops on September 30

This one has a deadline. Lace announced on September 23 that the “Lace Midnight Preview” extension will end on September 30. The hosted service the extension depended on is being retired, and the extension stops working from that date.

Three steps, in order: install Lace proper if it is not already installed; re-import the wallet used in Preview using its recovery phrase; then uninstall the Preview extension. Delete Preview without the recovery phrase in hand and there is no way back.

Midnight development has already moved into Lace proper. The 2.3 release announced this week lets users reset the sync state for a Midnight account from its account settings if syncing gets stuck, and fixes DUST fee calculation. Extension 2.4.0 adds DUST generation from Cardano accounts holding cNIGHT.

A design that holds no keys also defines what cannot be undone

At 21:31 JST on September 24, the Midnight Foundation stated that for wallets whose private keys were exposed in SecondFi’s (formerly Yoroi) June incident, it cannot restore a Glacier Drop claim or reassign it to a different address. The reason is not an operational judgment but the design itself.

“The Glacier Drop redemptions process was deliberately designed so that only the holder of a claim and its associated private key can redeem that claim. There are no master keys, backdoors or administrative controls that would allow Midnight Foundation to move, redirect or otherwise tamper with a claim.”

The mechanism that protects and the mechanism that cannot undo are one and the same. On September 21 SecondFi itself had urged affected wallet holders not to attempt to redeem their NIGHT allocation from an affected wallet. This week the party that holds the mechanism reached the same conclusion in its own words. Separately, lost staking rewards were stated to be outside the current claims process.

Also

The stable 2.1.0 release has still not appeared. On events, “Enter Midnight City: Tokyo,” with Charles Hoskinson and the Midnight City team, takes place on October 10 from 15:00 to 19:00 at Happo-en in Shirokanedai, Tokyo. Attendance is free with host approval. NIGHT was up 3.48% on the week.

14. The Key Itself Leaked

On September 20, World Mobile Chain published a Security Notice: the SingularityNET bridge had been exploited and WMTx minted without authorisation on Ethereum. The blockchain analytics firm PeckShield reported that the same address minted 260 million AGIX and 53,838,000 WMTx. Fetch.ai and NuNet were hit by the same attacker on the same day.

What the three had in common was not a logical flaw in a contract but the leak of the key that authorises minting.

The two incidents covered in the previous two issues broke somewhere else. Hydra was an optimisation that skipped signature verification; Splash was a validator that never checked the sign of the tradable reserve. In both, one assumption behind correctness had been dropped. This time nothing was dropped. The mechanism worked as designed, and the authority to operate it went outside. The Midnight Foundation stating in the same week that it holds no master keys, and Midnight’s node removing a debugger that exposed key material inside its images, belong to the same layer.

For readers following the Cardano ecosystem, the shape is awkward. Nothing happened inside your wallet, yet the price of a token you hold moved because supply increased on a different chain. World Mobile Chain’s notice acknowledges that the unauthorised minting “led to recent price action across all exchanges that $WMTx is listed on.” No primary report of a newly occurring attack on the Cardano or Midnight chains appeared this week.

15. Several Doors Opened From Outside

A run of announcements this week connected Cardano to things outside it.

  • Cardano joined the official SDK for x402, the payments standard over HTTP. The first version of @x402/cardano was published at 08:19 JST on September 19, and the Cardano Foundation announced it on September 21. Install it and you can take API fees in ADA or native tokens.
  • Cardano native tokens are coming to Fireblocks, per the Cardano Foundation’s September 24 announcement. Its CEO, Frederik Gregaard, said he is “excited to finally align Cardano Native Tokens with $ADA, and bring unprecedented distribution to builders on Cardano.” ADA itself has been supported since 2021; what changes is the tokens on top. The standards covered (CIP-26 and CIP-68) and the timing (as a standard asset by March 2027) are as reported in the press.
  • Catalyst Pilot 2026 announced its 12 selected projects, from 117 applications. The payment design changed: 40% of funding is paid up front and the remaining 60% is tied to on-chain adoption.
  • Cardano PRIME published its ecosystem audit. Cardano DeFi scored 51.3 out of 100, against 96.3 for Ethereum and 72.3 for Sui on the same framework. Starting from that baseline, PRIME explained on September 24 that Phase 3 incentives will concentrate in two or three venues. Its call for projects closes September 30.
  • A public roundtable on treasury allocation reform was broadcast on September 21. The five panellists agreed that even if funds are divided into domain buckets, the permissionless right to propose should remain.
  • FC Barcelona launched “Barça Fan Lab” on Cardano (September 25), and Intersect’s board election closed at 21:00 JST on September 25 (eleven candidates for two seats, results published September 28).

One disclosure about SIPO itself: SIPO began taking part in RealFi’s SPO accelerator on September 24.

16. Risk Dimensions

  • Operations (the Midnight node). node-1.0.300 says “security patch” in its own release notes and covers mainnet. Unless every validator has moved before set_code is enacted, the ones that lag fall off the chain. The release notes state that the ordering is not advisory.
  • Governance (how you count). In the k poll, vote count and delegated stake point in opposite directions, and the criterion uses stake. Reading only the vote count leads to the wrong conclusion about where the action stands.
  • Governance (not voting). In the same poll, 9.2 billion ada of non-voting stake behaves like a no vote. The decision whether to express a position lands in the result either way. The deadline is 06:44:51 on October 22.
  • Governance (what a committee yes means). The committee yes votes on minPoolCost and OpenZeppelin are, per the rationales, findings that the actions do not contravene the constitution, not judgments that the requests are sound. Unless three of the remaining five move, neither reaches two thirds.
  • Pool economics (SIPO itself). The k=1000 saturation point is about 36.77 million ada, and SIPO’s first pool (53.86 million ada) sits above it. Its delegated stake fell by 3.19 million ada across this epoch, but it remains above the line. SIPO sits on the side that loses from this change.
  • Treasury (the cap). If the OpenZeppelin action passes, 2026 withdrawals reach 93.84% of the cap. With about 42.6 million ada left, the room for another withdrawal within the year is limited.
  • Operations (a Midnight deadline). Lace Midnight Preview stops working on September 30, and migration requires the recovery phrase. That is four days after this issue.
  • Operations (the Cardano node and Hydra). 11.1.2 supersedes 11.1.1, so operators who already moved to 11.1.1 need to upgrade again. Mithril signers should go to 1.1.9 at the same time. The Hydra deposit-lockout fix was also published this week. It is not a security fix, but it is the kind of bug where funds do not come back.

17. What to Watch Next Week

  1. The epoch 658 boundary — 06:44:51 JST on September 27. Whether minPoolCost keeps yes above cast no by stake will show in the votes across that boundary. Check the action’s page on gov.tools.
  2. Intersect board election results — published September 28. Eleven candidates for two seats. Check Intersect’s news page. Weekly update #130 had not appeared as of writing either.
  3. Lace Midnight Preview shutdown — September 30. If migration is not done, that is the date. Check Lace’s official account and the input-output-hk/lace releases page.
  4. Cardano PRIME’s call for projects closes — September 30. Check Cardano PRIME’s announcements.
  5. Enactment of Midnight runtime 1.0.300 — set_code runs once per network. Check the midnightntwrk/midnight-node releases page and the release’s own attachments for the proposal hash.
  6. The constitutional amendment portal’s review session — 14:00 UTC on October 1, where CAP-10 and CAP-3 will move or not.
  7. SPO votes and rationales on the k poll — how far explicit yes moves from 0.57%, and whether operators follow the proposer’s own no. The deadline is 06:44:51 on October 22; check the Info Action’s page on gov.tools.
  8. Committee votes on minPoolCost and OpenZeppelin — whether three of the remaining five move. Both expire at 06:44:51 on October 12.
  9. Node Diversity Celebration Day — October 6, Singapore and online, where Amaru’s block-production alpha timing and SPO incentives are due to be set out.
  10. Enter Midnight City: Tokyo — 15:00 on October 10, Happo-en, Shirokanedai, Tokyo. Registration is subject to host approval.

Conclusion

Not one new action was submitted on chain this week. Votes simply accumulated in the three columns that already existed.

Of those, the k poll took the most rereading. Ten pools in favour, three opposed: by count, yes leads more than three to one. Yet the criterion the proposal declares for itself is weighted by stake, and on that measure no holds 95,984,098 ada against yes’s 53,185,274 ada, a factor of 1.80. Both numbers come from the same ledger, both are correct, and only the conclusion is reversed.

The shape has a precedent, and the proposal text cites it. In the 2023 poll, the option containing k=1000 took 69.64% by pool count and about 41.1% by participating stake. This poll repeats that voting power is stake-weighted because it comes after that experience. And the same shape has appeared in week one.

The third no vote came from the person who filed the poll. His rationale says we have not exhausted the current k=500 setting, and that he learned of the detailed case against only after submitting and agrees with many of its arguments. The person who asked the question voted against it himself. A poll, it turns out, does not fix the opinion its author held on the day of filing.

The other two moved as well. minPoolCost turned from the previous issue’s “last five votes all no” to 14 yes against 4 no, with yes now ahead by stake too. OpenZeppelin went the other way, with cast no votes reaching 9.76 times the stake in favour. The committee’s second vote landed on September 22, leaving both at 2 yes and 5 not voted — though that yes, per the rationale, is a finding of constitutionality rather than support for the request.

SIPO voted no on that OpenZeppelin action, at 10:38:08 on September 25, two weeks ahead of its own stated deadline. The grounds are overlap, an undecided language, conversion surplus and bundling, and six conditions for a yes on resubmission are recorded alongside them. SIPO is also the party that voted yes, in May 2026, to the funding it now cites as overlapping. Its position on the k poll is disclosed the same way: k=1000 is an unfavourable change for SIPO’s first pool.

Last week’s reading was that an action stops if any one of four columns stays empty. What came into view this week sits one layer earlier. The same votes, counted by what. By the number of pools, or by the stake delegated to them. The anniversary figure counting 2,890 pools and the saturation point asking how many pools the total should be divided among are the same choice, in the same place.

References

Where the text links to SIPO’s own articles, the link is in place. Primary sources follow.

Transparency Notes

  • The “1.80 times” on the k poll: explicit no of 95,984,098 ada divided by explicit yes of 53,185,274 ada. Non-voting stake of 9,251,693,095 ada is not included. The proposal declares that non-voting stake behaves like a no vote, but a ratio including it is not used here.
  • The denominator behind 0.57%: yes of 53,185,274 ada plus 9,347,677,193 ada of explicit no and non-voting stake, giving 9,400,862,467 ada. Explicit abstentions and the 617 pools set to “always abstain” (about 11.9 billion ada) are excluded, as the proposal declares.
  • Why no “opposition is X%”: the opposition percentage returned by the public API includes non-voting stake and “always no confidence.” This text therefore states cast no votes only, as a count and an ada amount. Yes percentages are used as returned.
  • Identifying the proposer with STR8: the proposal metadata’s author name matches the ticker of the pool that voted no, and the rationale itself says “I submitted this Info Action.” The rationale’s own authors field was empty. Intent at the time of filing is written only as far as the rationale states it.
  • Three votes recorded in the same second: the text states that three yes votes on the k poll were recorded at 10:03:06 on September 22, but does not say whether they belong to the same operator. The ledger records pools; it does not comprehensively disclose which pools sit in the same hands.
  • Naming a committee member: the chain yields only key hashes and term epochs. The name Tingvard comes from the rationales attached to the votes, which identify themselves, and the text goes no further. For the member who voted on four actions on September 12, no rationale was confirmed, so the key prefix is used.
  • What a committee “yes” means: the rationales state that they are findings of constitutionality, and the text does not present them as support for the requests.
  • Why SIPO voted earlier than announced: the rationale does not say. Only the difference in dates is recorded; no motive is supplied.
  • SIPO’s interests: that SIPO’s first pool sits above the k=1000 saturation point is disclosed in the body; for minPoolCost, that all three pools declare a fixed cost of 340 ada and are mechanically unaffected and that SIPO sits with the 340 incumbents; for OpenZeppelin, that SIPO’s DRep cast a no vote and that SIPO voted yes in May 2026 to an action whose deliverables overlap. SIPO’s participation in RealFi since September 24 is stated as well.
  • The saturation calculation: about 73,534,748 ada at k=500 and about 36,767,374 ada at k=1000 are mechanical figures, the on-chain circulating supply of 36,767,374,068 ada on the morning of September 26 divided by k. They do not match the roughly 75 million / 37.5 million ada the proposal cites as of its own reference point, because the supply is taken at a different time. SIPO’s three pool stakes are values at the epoch 656 boundary.
  • When the market values were taken: the September 19 values are from the usual morning reading. The September 26 values were taken separately at 07:19:50 JST on September 26, because the morning reading had not run by the time this was written. The two days are therefore tens of minutes apart. The epoch boundary ADA figure (+26.27%) is taken exactly at the boundary and, because it is taken at a different time, does not reconcile with the weekly +13.19%.
  • Anniversary figures against ledger figures: the Cardano Foundation’s 2,890 pools, the 2,893 at the epoch 656 boundary, and the September 24 ledger reading of “2,671 with delegation, 775 producing a block in the most recent epoch” are all taken at different moments. No subtraction is performed between them. Only the 158 governance actions match the count the ledger returns today.
  • On the word “security”: there was no cardano-node release this week, and neither the 11.1.2 release notes nor Intersect’s weekly update uses the word, so it is not described as a security fix. The Hydra fix is called a bug in the weekly report and is treated the same way. By contrast, Midnight’s node-1.0.300 says “security patch” in its own release notes, and the text says so.
  • Fireblocks standards and timing: the Cardano Foundation CEO’s words are taken from the foundation’s own channel. The standards covered (CIP-26 / CIP-68) and the March 2027 timing are as reported in the press, and the text distinguishes the two.
  • Midnight’s mainnet contracts: the text does not say they have been enabled. It says the Foundation’s CTO posted that they will be enabled soon, and that no date and no official release notes have appeared. Midnight’s official blog published nothing new this week.
  • Prices against actions and incidents: no connection is drawn. ADA’s +8.36% on September 22 is not linked to any action or incident of the same week.

This is not investment advice. For information and research purposes only.

関連記事

Weekly Brief EN
Weekly Brief #26 | Midnight Mainnet Opens Contract Deployment Without Review
Weekly Brief EN
Weekly Brief #24 | Ikigai Cleared the Committee and Expired 1.81 Points Short
Weekly Brief EN
Weekly Brief #23 | minPoolCost Returns Without an SPO Vote