Styg

Withdraw 11,787,063 ada for the OpenZeppelin Stack administered by Intersect — Voting Rationale

Governance Voting Rationale
GAID	gov_action1gxx...lkl2l8
Title	Withdraw 11,787,063 ada for the OpenZeppelin Stack administered by Intersect
Type of GA	Treasury Withdrawal
Amount	₳11,787,063 (₳11,443,750 delivery, stated as USD 1,831,000 at a reference rate of USD 0.16 · ₳343,313 Intersect administration fee)
Date submitted	8 September 2026 (epoch 654)
Expiration Date	11 October 2026 (end of epoch 660)

---

Contents

• 1.0 Introduction
  • 1.1 Summary
  • 1.2 Description of Governance Action
• 2.0 Discussion
  • 2.1 Method
  • 2.2 What this proposal gets right, stated properly
  • 2.3 A price agreed before the thing priced
  • 2.4 The gap is real, and it is not empty
  • 2.5 The part that forks and the part that is bought
  • 2.6 Who checks the checker
  • 2.7 A blueprint is a default
  • 2.8 The reference rate, the buffer and the ceiling
  • 2.9 What the commons retains
  • 2.10 Why this is a No and not an abstention
  • 2.11 Fiduciary verification (external instrument)
  • 2.12 What would make this a Yes
  • 2.13 What the vote does not reach
• 3.0 Conclusion
• Sources

---

1.0 Introduction

1.1 Summary

We are voting NO (remediable) on the OpenZeppelin Stack, and I want the shape of that No to be legible, because it is not the No that the loudest objections to this proposal would predict.
It is not a No on price. Read in the currency the work is contracted in, USD 1.83 million for twelve months of library engineering, three reference protocols, developer tooling and twenty-two researcher-weeks of security capacity is not an inflated number; if anything the breadth promised is wide for the money. It is not a No to an outside firm arriving with a treasury ask, either. A commons that funds only its incumbents has quietly stopped being one, and this DRep's method exists in part to stop a sector label or a recipient's origin from doing work that only the structure of a claim should do. And it is not a No to the problem, which is real and which Cardano's own engineering organisations describe in much the same terms the proposal uses.
The No rests on three findings, each a property of this action rather than of the people behind it. First, the price is fixed while the thing priced is not: the smart contract language is to be chosen after the vote, the component list is described as preliminary and indicative, no team is named, and the budget is a payment schedule of five equal instalments rather than a statement of costs. The discretion over what the treasury receives is disclosed, and it is not bounded. Second, what the commons would retain is the code, under a permissive licence, and that is genuinely to the proposal's credit — but the stewardship of that code, its maintenance past month twelve, its canonical home and the name that gives it its institutional value all stay with the vendor, and the continuation is pre-figured as a second ask. The part that forks is not the part being bought. Third, the assurance that makes an audited library worth more than an unaudited one runs through the party being paid to build it: the security work is the author auditing the author, the final milestone's audit reports fall outside the last payment gate, and the acceptance criteria measure whether code compiles and is covered rather than whether anyone independent has found it sound.
The external fiduciary instrument this DRep uses for the capital-allocation axis reaches No on the same structure from an unrelated foundation — budget clarity, continuity, independent verification and upfront exposure at a size where none of those is survivable. Two instruments built on different ground arriving at the same verdict is worth naming as what it is.
None of the three findings needs a different team or a different ambition. Each names its own repair, and the proposal's own text commits to returning with a revised submission if this one does not pass. That commitment is the right one, and this rationale is written to be useful to it.

1.2 Description of Governance Action

This Treasury Withdrawal requests ₳11,787,063 to fund a twelve-month engagement by OpenZeppelin, administered by Intersect, with Cardano Development Holdings as the contracting counterparty. The proposal states that it is not sponsored or endorsed by Intersect, which is prepared to act as administrator should it pass, and that this is OpenZeppelin's first request to the Cardano Treasury, with no Project Catalyst funding received.
The work is presented as three coupled workstreams. The first is three end-to-end reference implementations — a liquid staking protocol issuing a transferable yield-bearing token for staked ada, a self-repaying loans protocol, and a tokenized money market fund — each with working code, architecture documentation, a demonstration front end and a threat model. The second is a contracts library for Cardano in the role OpenZeppelin Contracts plays for Solidity, with a preliminary component list running from access control, cryptographic utilities and DeFi mathematics through vaults, timelocks, staking and delegation utilities, governance, credentials and claims, account abstraction and multi-signature, vesting, a standardised messaging gateway and a proposed Cardano Contract ABI, together with a Contracts Wizard, a UI builder, AI-assisted development tools and documentation. The third is a security retainer of twenty-two researcher-weeks covering audits of the library and full-stack reviews and penetration tests of each reference implementation, scoped to OpenZeppelin-produced code, with findings published and with OpenZeppelin funding bug-bounty payouts at the levels of its standard program. A developer enablement and co-marketing program is described as included at no additional cost. Library code is to be released under the MIT licence and OpenZeppelin tooling under AGPL 3.0.
The delivery budget is USD 1,831,000. The on-chain request converts that at a reference rate of USD 0.16 per ada to ₳11,443,750 and adds a three per cent administration fee of ₳343,313; the arithmetic is correct to the ada. On enactment the delivery portion is to be converted to a USD-denominated stablecoin at contract signature. Twenty per cent, USD 366,200, is released as a kick-off payment on signature, subject only to legal sign-off, and the remaining eighty per cent in four equal quarterly payments, each released after OpenZeppelin submits a milestone acceptance form with public evidence and Intersect's delivery assurance verifies it. Funds remaining in the contract at expiry sweep back to the treasury. Custody runs through the Sundae Labs treasury contract framework, with a six-entity oversight committee and multi-party authorisation for funding, disbursement, pause and sweep, and the funds are held undelegated to a stake pool and delegated to the predefined abstain option as the Constitution requires.
The proposal states compliance with the 500 million ada Net Change Limit covering epochs 613 to 713. On the public count available during the voting window, roughly ₳457.4 million of that limit was already committed, leaving about ₳42.6 million of headroom for the nine months remaining in the period; this request is a little over a quarter of that remainder.

2.0 Discussion

2.1 Method

This action earns a deep reading, and on grounds that are easy to mistake, so they are stated plainly. It is not the size of the ask, and it is not that the proposer comes from outside the ecosystem. A twelve-month engagement that sunsets and returns unspent funds changes nothing structural on its face. What deepens the reading is what happens if the work succeeds. A standard library is a dependency by design: it is valuable exactly to the extent that many independent builders import it and stop writing their own. A dependency of that kind is easy to decline today and progressively harder to decline at each renewal, without any further decision being taken, and when this DRep read a maintenance grant for an existing wallet earlier this year the distinction that mattered was whether the action serviced a dependency the ecosystem already had or manufactured a new one. This action would manufacture one. That is not an argument against it — shared infrastructure is supposed to be depended on — but it is the reason the terms on which the dependency is created deserve to be read before it exists rather than after.

2.2 What this proposal gets right, stated properly

A finding is worth less if the affirmation around it is grudging, so this section is written at the length the criticisms get.
The problem is correctly identified. Cardano does not have a standardised, audited contracts library that a builder can import with confidence, and the cost of that absence is paid in exactly the way the proposal says: each team rebuilds foundational components, audit effort is spent repeatedly on the same primitives, and an institution evaluating the chain has nothing canonical to evaluate. Input Output's own repository for a reusable contracts library opens by observing that the EVM ecosystem matured in part because OpenZeppelin gave it reusable contracts and that Cardano lacks an equivalent. The diagnosis is shared by the people best placed to dispute it.
The financial construction is better than most of what crosses this desk, and better than the general critique of treasury procurement would lead a reader to expect. The contract is denominated in dollars, so the vendor carries no incentive to pad for volatility in the price of the work. Conversion happens once, at signature, rather than leaving a year of milestone payments exposed to the ada price. Four of the five payments are gated on public evidence verified by a party other than the vendor. What is left in the contract at expiry returns to the treasury automatically, at the contract level, rather than at anyone's discretion. Custody is on-chain, multi-party and auditable by anyone. That is a design by people who expect to be checked.
The public-asset commitment is real. Library code under the MIT licence is as unencumbered as code gets; security findings and audit reports are to be published alongside releases rather than held; and OpenZeppelin commits its own funds to bug-bounty payouts at the level of its standard program, which is a genuine sharing of risk and one of the few places in any recent proposal where a vendor puts its own money beside the treasury's. The disclosures are clean: a first-time recipient that says so, an administrator that is explicit about not sponsoring the ask, and a statement that a failed vote will be followed by engagement on the concerns raised and a revised submission. That last sentence is the posture of a counterparty that expects correction, and it is the reason this rationale ends in a specification rather than a refusal.
And the capability is not in doubt. OpenZeppelin has carried a contracts library into several ecosystems that are not the EVM, and the claim that its presence makes a chain easier for institutions to evaluate is true as a description of how those institutions behave.

2.3 A price agreed before the thing priced

The first finding is about order, and it is the one that makes this a No rather than a Yes with reservations.
The proposal says that the specific smart contract language targeted will be determined during the initial evaluation phase, in collaboration with ecosystem stakeholders. That evaluation phase sits inside the first milestone, after the vote, after the contract is signed, and after the kick-off payment has been released. The choice is not a detail. Whether a Cardano library is written for Aiken, for Plinth, or for something else decides who can be hired to write it, what toolchain it is tested against, which existing codebases it can absorb or must duplicate, who in the ecosystem can review and later maintain it, and how much of the ecosystem will ever import it. It is also, on Cardano, a decision with a constituency: most application code written in the last two years sits on one side of that choice. A standard library whose language is unknown is not yet a specified deliverable, and delegated stake is being asked to fix its price.
The same openness runs through the rest of the scope. The component list is described in the proposal's own words as preliminary and indicative. Additional standards are to be determined with ecosystem stakeholders. The reference implementations can include institutional patterns such as credential-based access and multi-signature controls. Penetration testing applies where applicable. The architecture documents that anchor the first milestone are to be reviewed by pre-agreed counterparties who are not named. None of this is unusual language for a statement of work between two firms, where a contracting officer on the buyer's side holds the pen. Here the buyer is a treasury, and the pen is held after the vote by the vendor and by stakeholders the proposal does not identify.
Against that, the budget offers no purchase at all. The Constitution asks a treasury withdrawal to state the relevant costs and expenses of the proposed activities. What this proposal provides is a payment schedule: five instalments of exactly twenty per cent each, identical whether the quarter contains a design phase or two audited protocols and a developer toolchain, with the columns for security and developer enablement reading "included." There are no roles, no head-count, no rates, no allocation between the three workstreams, and no named delivery team with eUTXO experience. A reader cannot tell what share of the dollar figure buys security capacity and what share buys engineering, and so cannot tell what would be lost if any one component were cut or what should be saved if the language choice made a third of the list unnecessary.
Stated in this DRep's own terms: the discretion over what the commons receives for a fixed sum is real, it is disclosed, and it is not limited. Disclosing an open discretion answers the question of whether it is known and leaves untouched the question of whether it is bounded. The repair is the one the failure names — scope the discretion — and it is available in either of two forms. Fix the language, the component list, the reviewing counterparties and the cost allocation before the vote. Or, if the evaluation genuinely has to come first, ask for the evaluation: a small first action that funds the language decision, the overlap analysis described in the next section and the architecture documents, and returns with a priced specification. The proposal's own milestone structure already isolates that work. What it does not do is isolate the commitment.

2.4 The gap is real, and it is not empty

The proposal's problem statement says there is no standardised, audited contract library and that audited code is scattered across Cardano Foundation repositories, Intersect resources and individual project codebases. The first half of that sentence is true. The second half is the place where an inventory should be, and is not.
The space is occupied, unevenly and incompletely, by work this proposal neither names nor positions itself against. Input Output maintains a public contracts library with an Aiken on-chain layer and off-chain builders, early in its life — a vesting contract in progress and a longer list triaged — and explicitly aimed at the role this proposal claims. Design-pattern libraries for both Plutarch and Aiken were built by Anastasia Labs with community funding, and an Aiken open-source contract library by MeshJS and TrustLevel was funded the same way. The programmable-token standard that the proposal lists as a library component and as a milestone-four contribution was merged into the CIP repository while this action was live, with reference work around it already underway in the ecosystem. None of these is the finished, audited, canonical thing. All of them are the material that a consolidation would have to consolidate.
This matters to the proposal's own argument more than to any objection about duplication. The value claimed is consolidation — trusted, consolidated building blocks in place of fragments. A new library that arrives without stating what it absorbs, what it supersedes, what it leaves alone and whose maintainers it has spoken to is, on its face, one more fragment with a better-known name. It may well be intended otherwise; the proposal says OpenZeppelin will collaborate with Input Output, the Cardano Foundation, Intersect and other stakeholders. But a commitment to collaborate after the vote is not an account of the landscape before it, and the people asked to decide were given a problem statement in which the existing work appears only as scatter.
I want to be precise about what this finding is not. It is not a claim that the native efforts are sufficient, and it is not a claim that the treasury should prefer them because they are native. It is that the decision is being taken against an incomplete picture, by an electorate with limited time, and that the picture is cheap to complete. A public gap-and-overlap analysis, written with the maintainers of the existing work rather than about them, converts the central uncertainty of this proposal into its strongest section.

2.5 The part that forks and the part that is bought

An open-source licence answers one question about a public asset — may anyone copy and change it — and it answers it here in the commons' favor. It does not answer who maintains the asset, where its canonical copy lives, who decides what is merged, or what the name on it is worth. For this proposal those are the questions that carry the value.
Consider what an institution is buying when it treats an OpenZeppelin library as a reason to build. It is not buying the source code, which it could read under any name. It is buying the assurance that a firm with a decade of reputation stands behind that code today: that it was audited by that firm, is maintained by that firm, and will be patched by that firm when the ledger changes underneath it. The proposal is candid that this is the point — production-ready paths with the enterprise-grade security patterns institutions require. That assurance is a relationship with OpenZeppelin, and a relationship cannot be forked. If the engagement ends, anyone may copy the repository, and what they will hold is a library that used to be maintained by OpenZeppelin.
So the question the proposal has to answer is what binds the maintainer, and on that it is silent. Nothing is committed past month twelve. The final milestone's acceptance criteria include a twelve-month scope review completed with a year-two continuation proposal shared with the Cardano Foundation — which is to say that the last payment is released partly on delivery of the next ask, and that the next ask is addressed to a foundation rather than to the body that would pay for it. The documentation lives in a Cardano section of OpenZeppelin's own documentation site, the network page on OpenZeppelin's website, the Contracts Wizard in OpenZeppelin's tooling under a copyleft licence. There is no stated home for the repositories that survives the vendor's interest, no co-maintainer, no handover path, no contribution policy that gives the builders who will depend on the library any standing over it.
Cardano changes underneath its contracts. A hard fork is in preparation and more will follow, and each one is an occasion on which a contract library must be re-examined. An audited library that is no longer maintained is not a neutral object: it is trusted precisely because it was audited, and that trust outlives the audit's relevance. This is the sense in which the ecosystem would be easier to enter than to leave. Declining the dependency costs nothing today. Declining to fund year two, once protocols and their users sit on top of the library, is a different decision, taken against a party whose standing this action would have created.
Held against the standard this DRep applies wherever one party acquires the power to set terms that others build on — is that power matched by an obligation running the same direction — the reading is a partial inversion. The power to define default primitives for an ecosystem would pass to the vendor; no corresponding duty to maintain, to accept contribution, or to hand over would pass with it. The exit that exists on paper, the fork, is real and is credited. The exit that would matter in practice is not provided for. The repair is a correction path written into the design: stewardship terms that put the canonical repositories somewhere the commons can reach, at least one maintainer from the ecosystem with merge standing, a stated maintenance commitment across the next ledger era or a funded handover, and a continuation that is costed now, as a recurring service with a point at which it is re-competed, rather than discovered at month twelve.

2.6 Who checks the checker

The third finding concerns the one property that distinguishes this library from the ones that already exist: that it is audited.
The security retainer is scoped to OpenZeppelin-produced code and is performed by OpenZeppelin. That is how the firm works in other ecosystems and there is no suggestion here that its auditors are anything but separate from and rigorous toward its engineers. But the structure asks the treasury to pay one party to write the code, to examine the code, and to report what the examination found, and then releases payment on the publication of that report. The independent element in the acceptance criteria is a single line in the third milestone: at least one independent Cardano developer has reviewed and provided feedback on the library components, confirmed in writing. For three financial protocols and a foundational library intended to be imported across an ecosystem, one reviewer providing feedback is not an independent assurance function.
The scale of the security capacity deserves a moment's attention for the same reason. Twenty-two researcher-weeks is to cover line-by-line audits of some fourteen library components and full-stack reviews and penetration tests of a liquid staking protocol, a synthetic-debt lending protocol and a tokenized fund with credential verification and dividend distribution. For comparison, the engagement OpenZeppelin announced with the Stellar Development Foundation allotted forty auditor-weeks over two years within a program covering a library and its tooling, with no reference protocols in scope. I do not know OpenZeppelin's internal sizing and it may be sound; the reference implementations may be thinner than their descriptions suggest. But that is the difficulty. The proposal describes them as production-ready, the acceptance criteria describe them as demonstrable on testnet with a functional demo front end, and those are not the same claim. A blueprint that carries an auditor's name and is read as production-ready by a team without the expertise to know the difference is a hazard created by the gap between the two descriptions.
There is also a sequencing gap at the end. Audit reports for each milestone are published in the milestone that follows, with critical and high findings resolved — a sensible lag. But the fourth milestone has no successor. Its criteria require that critical and high findings be addressed and that a publication timeline be agreed, and the final twenty per cent is released on that basis. The audit of the tokenized fund, the most institution-facing deliverable in the program, is therefore published after the last payment gate has closed.
The acceptance criteria elsewhere are objective, which is to their credit, and mechanical, which is their limit: code compiles, continuous integration passes, test coverage exceeds ninety per cent, documents are published. Those measure that work was done. They cannot register that the work was unsound, unused or unwanted, and nothing in the milestone table could come out badly in a way that told the ecosystem something it needed to know. The proposal says its deliverables are designed to enable measurable on-chain impact; no measure of it appears.
The repair is modest. Fund an independent review — a second firm or a named panel of ecosystem auditors — for the reference implementations and the highest-risk library components, paid from the withdrawal and not chosen by the vendor. Hold a retention against publication of the final audit reports. And describe the reference implementations as what the criteria make them: audited testnet reference designs, with production readiness a claim to be earned by a specified standard.

2.7 A blueprint is a default

One of the three reference implementations needs separate attention, because it touches the layer this DRep is elected to care about.
Staking on Cardano is already liquid. Ada delegated to a pool never leaves its owner's control, is never locked, and carries with it the owner's own choice of governance representative. A liquid staking protocol of the kind described — deposit, delegation across a set of pools, issuance of a yield-bearing token — replaces that arrangement with a pooled one. The ada sits behind a script; the script, or whoever governs it, chooses the pools; and because the ledger currently requires a stake credential to delegate its voting power before rewards can be withdrawn, the script must also choose a governance delegation for everything it holds. Each depositor's individual decision about where their stake and their vote go becomes a single protocol-level setting.
The proposal's acceptance criteria for this implementation list deposit, delegation across a pool set, token issuance and accrual, and redemption. They do not mention the governance delegation of the pooled ada at all. The threat model is to document eUTXO-specific attack vectors; it is not asked to consider stake concentration or the accumulation of voting power as vectors. Yet a reference implementation is, in practice, a default. Whatever this blueprint does with the pooled vote — passes it through to token holders, parks it on abstain, or leaves it to an operator key — is what the first several forks of it will do, with an auditor's name attached. Other ecosystems learned what a dominant liquid staking protocol does to the distribution of both stake and governance weight only after it had happened.
This is not a reason to oppose a liquid staking reference design, and it is not a finding of harm; the Constitution itself contemplates designees holding ada and reserves to the owner the choice of whether they may vote it. It is a reason the design constraints should be stated before the treasury pays for the template. A revised proposal should make the handling of pooled voting power and the pool-selection rule first-class requirements of the liquid staking implementation, name stake and governance concentration in its threat model, and put the architecture out for public review — not review by pre-agreed counterparties — before it is built. I will be monitoring this closely in any resubmission, and separately: an instrument that pools delegation also further obscures what the delegation register is measuring, at a moment when the ecosystem is about to change the rule that shapes that register.

2.8 The reference rate, the buffer and the ceiling

The ada figure in the title is larger than the work costs, by design, and the design is mostly a good one. It nevertheless has three consequences the proposal does not discuss.
The dollar budget is converted at USD 0.16 per ada. Across the voting window ada traded between roughly USD 0.20 and USD 0.25. At about USD 0.235, the price on the day of this reading, the delivery portion of the withdrawal would convert to about USD 2.69 million against a contract of USD 1.83 million — a margin of some forty-seven per cent, or about ₳3.6 million more than the work requires at spot. That margin is not paid to OpenZeppelin, whose milestone payments are fixed in dollars, and on the proposal's terms what is left sweeps back at expiry. This is the right way to carry price risk between a vote and a signature, and the common complaint that treasury proposals hand vendors a volatility windfall does not describe this one.
The first consequence is that the margin is nonetheless withdrawn, and the ceiling counts what is withdrawn. With about ₳42.6 million of headroom left under the Net Change Limit for the nine months that remain in its period, this request occupies more than a quarter of it, and something like a third of that occupation is a buffer that nobody intends to spend. Under a ceiling this nearly exhausted, every dollar-denominated proposal that prudently buffers its conversion crowds out later proposals with money that will come back after the period has closed.
The second is that the return path is unspecified. The whole delivery portion is converted to a stablecoin at signature. What sweeps back at expiry is therefore, on its face, a stablecoin balance, and the treasury is an ada account. When the contract expires, whether the surplus is swept on conversion rather than held for the year, at what rate and at whose cost it is converted back, and what happens in the opposite case — ada below the reference rate at signature and the contract short — are questions the proposal's text does not answer. They may have settled answers in the administrator's procedures; a self-contained withdrawal should carry them.
The third is small in amount and worth recording for its shape. The administration fee is three per cent of the buffered ada figure and is not converted. An administrator's fee that scales with the size of the buffer rather than with the cost of the work administered is paid, at current prices, about half again what three per cent of the contract would be. I impute nothing by noting it. It is simply a place where the incentive and the service have come apart, and it is cheap to align: denominate the fee on the contract value.
None of the three would carry a No by itself. Together they are why the favorable reading of this proposal's financial design, which I hold, is not the whole of the reading.

2.9 What the commons retains

The treasury may fund the maintenance of commons relations; it may not fund the transfer of a private surplus. That is a test of the structure of a claim and not of the recipient, and a for-profit firm maintaining a non-rival good the ecosystem depends on passes it routinely.
There is a private surplus here and it should be named without color. A security and infrastructure firm whose business is audits, retainers and enterprise tooling would be paid by the treasury to establish itself in a new ecosystem as the author of that ecosystem's standard library, with a dedicated account manager for a foundation, a network page, co-marketing, and first position with every institution that arrives later needing an audit. That is ordinary commercial logic and nothing in it is improper. It is also exactly the circumstance in which the proposal carries the burden of showing, in terms of relations rather than quantities, what capacity the commons holds afterwards and how its degradation would be recognised.
Part of that burden is met, and met properly. Reusable primitives under an unencumbered licence, that any builder can import and any builder can fork, are coordination capacity in the plainest sense, and the proposal states them as such. Published audit findings are a public record. Those are relational claims and they are credited in full.
Part of it is offered in the wrong currency. The case for why this firm is made almost entirely in quantities — trillions of dollars transferred through its contracts, a share of the largest protocols, billions of transactions, weekly downloads. Every one of those numbers is true and every one of them describes another ecosystem. They establish that OpenZeppelin is very good at what it does elsewhere, which is not disputed, and they say nothing about what relation would be maintained here. The more impressive such figures are, the more care is owed in not letting them stand in for the answer; precision about a quantity produced somewhere else is not evidence about a capacity retained here.
And the remaining part is the finding of section 2.5: the capacity the commons would retain has no durable form beyond the source text. I do not read this as a dispute at the margin about whether a well-stated relation is genuinely served — the kind of honest two-frame disagreement on which this DRep would stand aside. The relation is stated; what is missing is the set of terms that would make it hold, and those terms are nameable and within the proposer's gift. That places the proposal on the side of the line where a No that specifies is owed.

2.10 Why this is a No and not an abstention

Four grounds for abstaining were live and each is foreclosed, so the reasoning is set out rather than asserted.
The first is that the ballot cannot carry what the analysis found. Some of it genuinely cannot: the tensions set out in section 2.13 are about how Cardano funds infrastructure at all, and no single vote reaches them. But the three findings that carry this token — scope left open under a fixed price, stewardship that stays with the vendor, and assurance that runs through the funded party — are properties of this action, stated on its face and fixable in it. Where the ballot reaches the finding, the ballot is owed the finding.
The second is that the subject is outside what this DRep's framework can read. It is technical, and I am not a smart contract engineer. That is not a ground. Not understanding a domain is either an obligation to study or an admission of insufficiency, and neither is a claim that the framework has nothing to say; here it has a good deal to say about discretion, stewardship and verification without needing an opinion on a single line of code.
The third is that the disagreement is a margin dispute, and it deserves the most care because this is the kind of proposal on which that reading is most tempting. The proposer is serious, the construction is careful, the track record is long. It would be easy to read all of that as good faith contested only at the edges and to withhold stake accordingly. But care in construction and reputation elsewhere are not evidence about the three structures in question, and withholding stake is not a neutral act: it lowers the bar an action must clear.
The fourth is interest. I hold ada and use the chain, which is the interest every participant shares and disqualifies no one. This DRep has no engagement, past or pending, with OpenZeppelin, with Intersect's administration service, or with any of the teams whose existing work is discussed in section 2.4. One adjacent interest is declared rather than left to be discovered: several of the repairs proposed here would be natural clauses in a common standard for treasury proposals, and this DRep works on standards of that kind. That is an interest in the remedy and not in the action, and readers should weigh the specification with it in view.
What pulled hardest the other way should be said too, because this vote was closer than its token suggests. The gap is real and has been real for years; the existing efforts have not closed it; and here is a capable party offering to close it on better financial terms than most. A treasury that answers that with an indefinite No has chosen the status quo, and the status quo has costs that fall on every team rebuilding an access-control module and on every user of what they build. That is why the No is remediable, why the specification is short, and why I would rather read a corrected version of this proposal in the next cycle than see the question go quiet. At the time of writing the action stands at roughly three per cent approval against a threshold of sixty-seven; this rationale is addressed less to its outcome than to its successor.

2.11 Fiduciary verification (external instrument)

This DRep's own framework reads the relational and structural question and by design does not read the capital-allocation one — price, instrument fit, size-calibrated discipline, upfront exposure, opportunity cost. Leaving that region unexamined would raise vigilance rather than lower it, so the proposal was run through the DRep Treasury Rule Book v17 as a subordinate gate, one that can lower a merits-Yes but cannot raise a merits-No.
By economic substance this is open-source infrastructure and shared tooling delivered by a commercial firm, which the rulebook reads on its infrastructure scorecard with the public asset, its operability and its continuity treated as the principal return. At ₳11,787,063 nominal it classifies very large — the band that requires exceptional public return, capped upfront exposure and an explicit opportunity-cost case, and where a passing score is necessary and nowhere near sufficient. It should be said that the band is set in ada and the contract is modest in dollars; the rulebook's instruction is to take the highest band any test triggers and not to move down for a good purpose, and that instruction is followed here.
The gate does not clear, and most of what stops it stops it before scoring. Budget clarity fails at this size: there is no itemisation, no rates, no allocation, and the rulebook's instruction where an applicant cannot separate a material bundled budget is unambiguous. Continuity fails: no maintenance, succession or handover is stated beyond the term, and the rulebook's warning that open source without operability and continuity is not yet a public asset is written for this case. Independent verification is partial: milestone evidence is public and is verified by the administrator, which is a real strength, but the security assurance is self-performed and the rulebook does not accept applicant self-certification at this band. Upfront exposure is twenty per cent on signature, subject only to legal sign-off, against a normal ceiling of fifteen per cent for very large requests and lower for a team new to the platform with its toolchain undecided. Dispute, jurisdiction and termination terms are deferred to a contract not yet written. And on instrument fit the rulebook names, as a warning sign, one large master plan before feasibility is shown — where here the feasibility work is literally the first milestone of the plan.
Against all that, several things score well and should be recorded: the public need is evidenced, custody and disbursement are exemplary, conversion and unspent-funds rules exist where most proposals have none, and prior funding is cleanly disclosed.
So this is convergence at the conclusion rather than corroboration at a soft spot. A rights-derived reading and a capital-stewardship rulebook reach No against the same structure from opposite directions, and they meet at two points in particular — scope before price, and continuity after term. Had the fiduciary read come back clean, the relational No would have stood unchanged; had the relational read passed and the fiduciary one failed, the allocation defect would have been dispositive. The gate's function here is to sharpen the specification.

2.12 What would make this a Yes

The specification is deliberately short, and every item is cheaper before the vote than after it.
Fix the scope or stage the ask. Either name the language, the component list, the reviewing counterparties and the delivery team before funds are approved, or bring a small first action that funds the evaluation phase alone and returns with a priced specification.
Publish a gap-and-overlap analysis, written with the maintainers of the existing libraries and the programmable-token work, that says what this library absorbs, supersedes and leaves alone.
State the costs. Roles, allocation, rates, and the split between the three workstreams, so that the Constitution's requirement is met in substance and a change of scope can be priced.
Write the stewardship into the proposal. A canonical home for the repositories that survives the vendor's interest, at least one ecosystem maintainer with merge standing, a maintenance commitment through the next ledger era or a funded handover, and a costed continuation presented now as a recurring service with a point of re-competition.
Put assurance outside the author. An independent review of the reference implementations and the highest-risk components, funded from the withdrawal and not selected by the vendor; a retention held against publication of the final audit reports; and the reference implementations described as what their acceptance criteria make them.
Constrain the liquid staking blueprint. Treat the governance delegation of pooled ada and the pool-selection rule as stated design requirements, name concentration in the threat model, and open the architecture to public review before build.
And on the allocation side: bring the kick-off payment inside the band's normal ceiling or tie part of it to the first verifiable output; say when the contract expires, in what asset the surplus returns and at whose cost, and what happens on a shortfall; and denominate the administration fee on the contract value.
Do those, and this becomes a proposal this DRep would expect to vote for.

2.13 What the vote does not reach

A vote is a snapshot; the log is the trajectory. This proposal is an unusually clear specimen of several tensions in how Cardano funds infrastructure, and converting those into this vote would misrepresent both the proposal and the tensions. They are recorded here, and they are held open rather than resolved, because I do not think any of them has a solution so much as a position the ecosystem occupies and can be watched moving along.
The first is between the instrument and the good. A treasury withdrawal is a discrete purchase: a scope, a sum, a term, an end. A standard library is a continuous service whose value is its maintenance. Forcing the second into the first produces exactly what is visible here — a twelve-month build whose real subject is years two through ten, with the continuation appearing as a line in the final milestone. The ecosystem has no recognised instrument for a recurring public service with service levels and periodic re-competition, and until it has one, every infrastructure proposal will be a first instalment that cannot say so.
The second is between prudence and the ceiling. This proposal handles price risk well and is penalised for it by a limit that counts ada withdrawn rather than value spent. The same mismatch sits underneath the several competing proposals to reform the Net Change Limit, each of which reaches in its own way for some reading of what was actually spent. Buffers that return after the period closes are a concrete case for that work, and I intend to carry this one into it.
The third is between the brand and the reading. Representatives with limited time read signals, and a famous name is a strong one; the worry that large outside firms will be waved through on reputation is a reasonable worry. This vote is evidence against its strong form. The electorate did not defer: at the time of writing, explicit opposition outweighs support by more than ten to one. Whether that reflects close reading or a different reflex — suspicion of the outsider — the tally cannot say, and the second would be no healthier than the first. What I will be watching is the reasons given, here and at resubmission, more than the totals.
The fourth is between openness and cultivation, and it is the one I would least like to see settled cheaply. A large allocation to a dominant outside firm for work that native teams have been doing on small grants tells those teams something about where the funding path leads, and capability that is not exercised atrophies. That is a real cost. So is its opposite: an ecosystem that treats nativeness as a qualification has built a guild, and the people it most reliably protects are not the ones doing the best work. I do not think the answer is a preference in either direction. I think it is a requirement that any proposal entering occupied ground say what is already there — which is why the overlap analysis is in the specification and why a standard for it belongs in a future proposal rather than in this vote.
The fifth is between standardisation and shared fate. The argument for a standard library is that everyone stops rebuilding the same components badly. The argument against a single one is that everyone then fails together. Both are correct. A healthy position probably contains a canonical library and at least one independent implementation of the critical parts, and nothing in the current funding landscape asks for the second.
Three smaller readings belong beside these. The engagement's accountability is addressed to a foundation — the account manager, the continuation proposal — while its payment comes from the commons; the foundation also sits on the oversight committee. That is a pattern in how institutions come to stand in for the community in the self-description of proposals, and it is worth watching across proposals rather than objecting to in one. The component list would give the standing of a standard to a particular set of primitives — pause, timelock, role-based access, credential-gated transfer, token-voting governance — and what is on the paved road shapes what gets built; the commons was not asked which primitives it wanted paved. And the questions this DRep's intake asks that no proposal template yet does — what reverses this if it is wrong in eighteen months, what reads through it, what would register its degradation — went unanswered here because they were never put. That is a reading of the ecosystem's vocabulary and not a defect in the submission.
Two of these are candidates for a future proposal rather than a future vote: a recognised instrument for recurring public-service funding, and a common standard for what a treasury proposal entering occupied ground must state — scope before price, an overlap account, stewardship terms, and how a conversion buffer is counted.

3.0 Conclusion

We are voting NO (remediable).
The problem this proposal names is real, the firm proposing to solve it is capable of solving it, and the financial construction is among the more careful this DRep has read. The refusal is not of the ambition, the price or the proposer. It is of three structures: a sum fixed before the language, the components, the team and the costs are; a public asset whose code the commons would hold while its stewardship, its maintenance and its name stay elsewhere, with the continuation arriving as a second ask; and an assurance claim verified by the party making it. Run separately, the external capital-allocation instrument reaches the same conclusion against the same structures.
Each is repairable by the proposer, and the proposal has already said it will return. Scope fixed or staged, an honest map of the ground already occupied, costs stated, stewardship written in, assurance placed outside the author, and the liquid staking template constrained where it touches the vote. That is the whole specification.
What I will carry forward beyond this action is the larger pattern it makes visible: a treasury that buys continuous infrastructure through discrete purchases, under a ceiling that counts buffers as spending, with no settled way of asking a newcomer what is already there. Those are not this proposer's doing, and they will outlast this vote. This DRep will continue to maintain a state of inquiry around them and will bring the two that are ripe into proposal form.

DRep ID: drep1yfaq8dsam...3nq50q DRep Profile: On DRepTalk

My thanks to the delegators who make this reading possible, and to the people at OpenZeppelin who brought a serious proposal to an open vote and said in advance that they would listen to the answer.

Stay in touch. X: https://x.com/styg50

Sources

External figures were read on 8 October 2026 and should be re-verified at submission.

• Governance action record, vote state and anchors: Cardano Cube; submission epoch and closing time: CryptoTicker; status at 2 October: Intersect weekly update #131.
• Net Change Limit usage (DRepTalk count, as reported): CryptoBenelux, 94% article and reform article.
• Ada price across the voting window: CoinGecko; CoinGape.
• Existing library work: input-output-hk/contracts-library; Anastasia Labs design patterns and its community funding record; MeshJS and TrustLevel Aiken library; CIP-113 merge: Cardano Forum digest, 30 September.
• Comparator engagement: Stellar Development Foundation and OpenZeppelin.
• Contracting counterparty: Intersect due diligence policy.
• A public rationale by another DRep reaching No on overlapping grounds: as syndicated.

Abstaining to help provide a clearer SPO signal.

Governance Incentives Framework — Voting Rationale

Governance Voting Rationale
GAID	gov_action174l...67pvcp
Title	Governance Incentives Framework
Type of GA	Treasury Withdrawal
Amount	₳4,207,967
Date submitted	Epoch 649 (Aug 14, 2026, 18:47 UTC)
Expiration Date	Epoch 656 (Sep 16, 2026)

---

Contents

• 1.0 Introduction
  • 1.1 Summary
  • 1.2 Description of Governance Action
• 2.0 Discussion
  • 2.1 Method
  • 2.2 What this proposal gets right, stated properly
  • 2.3 The number being studied is currently manufactured
  • 2.4 The pilot is scheduled across the change that moves its population
  • 2.5 A mechanism chosen before the method that was supposed to choose it
  • 2.6 Success criteria that cannot fail, and the missing quality term
  • 2.7 Who authors the measure
  • 2.8 What the commons retains
  • 2.9 Why this is a No and not an abstention — and my own interest, declared
  • 2.10 Fiduciary verification (external instrument)
  • 2.11 What would make this a Yes
  • 2.12 What the vote does not reach
• 3.0 Conclusion
• Appendix: Published Sort and Instrument Record

---

1.0 Introduction

1.1 Summary

We are voting NO (remediable) on the Governance Incentives Framework, and the shape of that No matters more than the token, because almost everything this proposal says about the problem is correct.

The incentive deficit is real. Active DRep numbers fell in every twelve-epoch period of the first year of on-chain governance; a hundred-plus proposals now arrive against a couple of hundred active voters, unpaid; a Constitutional Committee consortium retired for want of compensation. Refusing compensation in principle is not a neutral position, and this rationale does not take it. Sustained uncompensated labour in a relationship that participants already read as asymmetric does not stay a gift — it hardens into a role, expected rather than owed, and that entrenchment is its own harm and a quieter one than overspending. The people behind this proposal identified that problem early, wrote it up as a problem statement rather than jumping to a mechanism, and have done the unglamorous work of cataloguing forty-nine separate efforts so the ecosystem can stop duplicating itself. That is real contribution and this DRep wants it to continue.

The No rests on three findings, and all three are properties of this action rather than of the people proposing it. First, the central measurement the project is denominated in – delegated stake, and the population of registered DReps it is drawn from – is currently produced in part by a protocol condition rather than by anyone's governance intent, and the proposal neither says so nor names the pending change that would separate the two. Second, the controlled pilot is scheduled to launch in months five to six and be evaluated through month eleven, a window that on present expectations straddles that change; a measurement baselined on one population and read out on another is not hard to interpret, it is uninterpretable, and the project's most expensive deliverable would be unrecoverable. Third, the pilot's mechanism is pre-committed at milestone two, ahead of the modelling at milestone four that the proposal says is what selects mechanisms – which inverts the project's own stated ordering, and pays live governance actors under a rule chosen before the evidence, while the project's refinement discretion over that rule is disclosed rather than bounded.

Run independently, the external fiduciary instrument this DRep uses for the capital-allocation axis reaches No on the same structure from the opposite foundation – missing contracting entity, undisclosed disbursement profile, no licence on the assets, continuity promised only for as long as the team maintains the website, and activity counts standing in for outcomes at a request size where none of that is survivable. Two instruments built on unrelated grounds arriving at the same verdict is worth naming as what it is.

None of the three findings needs a different team, and none needs a different problem. Each names its own repair, and the repairs are cheap now and unrecoverable later, which is precisely why they belong in a No rather than in a note attached to a Yes.

1.2 Description of Governance Action

This Treasury Withdrawal requests ₳4,207,967 over twelve months to produce a Governance Incentives Framework: a structured, evidence-led approach to designing, testing and evaluating incentives for Cardano's governance actors. The stated architecture separates evidence from legitimacy — candidate mechanisms are to be researched, modelled and tested first, and only then put to governance for a decision on adoption.

The work runs across seven milestones. Discovery and mapping in months one and two; a preliminary framework, parameter landscape, anti-gaming framework, state-of-the-art evidence review and pilot testing plan in months three and four; data-collection infrastructure and the launch of a controlled pilot in months five and six; incentive modelling and simulation in months seven and eight; a public dashboard and model refinement in months nine and ten; final validation in month eleven; and in month twelve a final report, a submitted Governance Incentives Framework CIP, a governance integration pathway, and a governance action put to DReps seeking approval of the completed framework and its recommended incentive-model pathway.

Budget is presented at workstream level: conceptualization ₳445,000; community engagement ₳366,400; modelling ₳884,000; data engineering and monitoring ₳295,000; website and dashboard ₳253,333; administration and coordination ₳472,000; science and research ₳476,667; pilot ₳333,333; additional costs ₳118,667; contingency ₳437,328; independent audit and oversight ₳126,239. The arithmetic totals correctly to ₳4,207,967; the contingency is twelve per cent of the pre-contingency base and ten point four per cent of the whole, and the audit allocation is three per cent of the total including itself, as stated. Against a ₳350,000,000 net change limit the request is 1.20 per cent, as the proposal says. Unused contingency and pilot funds are to be returned to the treasury, and allocations rendered unnecessary by overlap with Input Output Research's already-funded work are to be returned proportionally.

The pilot is the part that spends real money on live governance behaviour. It applies a "middle-out" mechanism, drawn from a consultant's published process-first reward design, distributing eligible ADA rewards to DReps by script, with parameters, safeguards and eligibility rules refinable before launch and fixed thereafter except for predefined safety or termination conditions.

---

2.0 Discussion

2.1 Method

Every governance action gets read; not every action earns the same reading. This DRep runs each one first through an intake that assigns depth — how much analysis is owed and which instruments run — before any merits are read and before any standing policy applies. That intake returns a lane, never a verdict. Reading the merits first and then choosing the depth is how depth ends up chosen to fit a conclusion already reached, so the order is held. The token is selected last, under a discipline whose operative question is not whether a vote helps something pass but whether a warrant has been derived to deploy delegated stake against it. The full machinery of the run is published in the appendix, so a reader who disagrees can locate whether the disagreement is about how deeply this was read or about what the reading found — those are different arguments, and only one of them is about the proposal.

This action earned the deepest available reading, and it earned it on grounds worth stating plainly. It is not a wall: a twelve-month funded program that sunsets and returns unspent funds tunes nothing structural. But the failure mode is not reversible by simply declining to renew. Money paid to voters cannot be unpaid, and a measurement series baselined across a protocol change cannot be re-baselined afterwards. And what the project produces — the dataset, the dashboard, the metric definitions, the framework itself — is the lens through which the governance layer would afterwards evaluate every incentive proposal that comes to it. That is the kind of action where a mistake does not merely produce a wrong answer; it produces a wrong instrument, and the wrong instrument keeps working.

2.2 What this proposal gets right, stated properly

The proposal refuses the easiest and most damaging move in this whole space. It says, in its own alignment section, that higher activity or turnout will not be treated as sufficient evidence of success, and that mechanisms will also be assessed for participation quality, reward distribution, concentration, gaming risk and effects on delegators. That is the failure this genre almost always walks into, and this proposal walks around it deliberately. It commits to failure-mode testing against anti-gaming, anti-concentration and budget constraints, to publishing data sources and known limitations rather than only results, and to documenting material stakeholder feedback that it does not adopt along with the reason. It commits to reproducing and evaluating the formal reward-scheme research the ecosystem's own researchers produced, including its behavioural and cost assumptions, and to testing whether those assumptions transfer to a live governance environment — which is the right instinct, and more than most proposals in this category attempt.

It also handles two things that are usually handled badly. Overlap with Input Output Research's already-funded governance-incentives work is acknowledged rather than papered over, with a commitment to coordinate and to return the corresponding allocation where a work package becomes unnecessary. And unused restricted reserves are returned rather than retained, with proportional return on reduced scope. Those are the marks of people who expect to be checked and have decided to make checking easy.

The work behind it is real. The problem statement came before the mechanism, which is the correct order and an unusual discipline. The landscape map is a genuine public service; the ecosystem is materially better off for having forty-nine efforts catalogued in one place than it was for having them scattered. This DRep has read that map and has drawn on it, including in the analysis below.

2.3 The number being studied is currently manufactured

Here is the finding the rest of this rationale hangs from, and it is not a criticism of anyone's care.

Nearly every mechanism in the landscape — and nearly every metric this project proposes to collect, model and dashboard — is denominated either in delegated stake or in counts of registered and active DReps. Both of those numbers are currently produced, in unknown proportion, by something other than governance intent. Drawing staking rewards presently requires delegating voting power. So the register does not contain the population of people who wanted to participate in governance; it contains the population of people who wanted their rewards, with the governance-interested subset embedded inside it — and not separable by any refinement of the measurement, because a delegation made from conviction and a delegation made to unlock a withdrawal are byte-identical on chain.

That is not a noisy measurement. It is a measurement whose referent is unresolved, and everything downstream inherits it: delegation concentration, the Gini figure the proposal quotes, participation rates, active-DRep trend lines, and every reward formula denominated in any of them. The proposal opens by citing the DRep Gini rising from 0.92 to 0.94 and eleven to sixteen DReps holding half of voting power. Those figures are load-bearing in its motivation section, and they are measured over a register that is partly compelled.

The pending change that would unmix this is CIP-0181, which removes the delegation requirement for reward withdrawal. It does not appear in the landscape map. Neither does the reward-withdrawal condition itself, under any heading. This DRep does not read that as a failing of the people who compiled the map — CIP-0181 reads as a staking-rewards change rather than an incentives one, and falls outside the map's scope by construction. But the effect is that the map is thorough about mechanisms and silent about the one pending change that alters what its central variable means. A framework built on that map inherits the silence.

What is owed here is not a new opinion. It is a declaration. Before any of this is assessable, the project should state what population each of its measurements is denominated in, and say out loud that the population is currently produced in part by a ledger condition rather than by anyone's choice. That single paragraph, written into the framework at milestone two, changes what every later number means and costs nothing to write.

2.4 The pilot is scheduled across the change that moves its population

The sequencing consequence is sharper than the framing one, and it is the finding that makes this a No rather than a Yes with a note.

The proposal launches its controlled pilot in months five to six, applies a finalised specification, and evaluates it through final validation in month eleven, with findings published at month twelve. If the delegation requirement is removed during that window — and it presently looks likely to land in the next hard fork — then the population the pilot was baselined on and the population it is read out on are not the same population. Registered stake will fall, possibly sharply. Every metric time series breaks at the fork and no analysis spans it. A pilot straddling that break does not become harder to interpret; it becomes uninterpretable, because the thing that changed underneath the measurement is the measurement's own denominator.

This is a scheduling problem, and scheduling problems are cheap to solve in advance and impossible to solve afterwards. ₳333,333 of pilot funding and a good deal of the ₳884,000 modelling allocation would be spent on a result that cannot be defended against the obvious objection, and no later analysis recovers it. There is a second and worse version of the same problem: if a compensation pilot, a decoupling change and any concentration intervention all land in overlapping windows, the register moves for three reasons at once and the ecosystem permanently loses the ability to attribute any of the movement to any of the causes. That would cost more than this proposal's budget, and it would cost it in a currency the treasury cannot replace.

There is also a moment worth naming in advance, because it is the moment where the error becomes most attractive. When the register settles and registered stake falls, that fall will be the instrument beginning to work rather than a crisis — and it will arrive at exactly the point where a mechanism designed to push the number back up looks most compelling. A project holding both the dashboard and the pilot at that moment is the project best placed to read the fall correctly and also the one most exposed to reading it as a problem it is funded to solve. Stating the expectation in advance is what protects against it, and the proposal does not state it.

The repair is small. Sequence the pilot behind the sensing change, or, if the project judges the timing unknowable, pre-register in advance how a break in the series will be handled and accept publicly that no series spans it. Either is a paragraph in the milestone-two pilot plan. Neither is available once the money has moved.

2.5 A mechanism chosen before the method that was supposed to choose it

The proposal's central promise is an ordering: evidence first, governance second, mechanisms evaluated before any is adopted. That promise is what distinguishes it from the dozen proposals that simply advocate a scheme. The milestone structure does not keep it.

The pilot's basis is fixed at milestone two, where the acceptance criterion is a completed pilot testing plan "with the intended middle-out basis" documented. The pilot launches at milestone three, in months five to six. The incentive modelling and simulation that the proposal describes as the process by which candidate mechanisms are designed and tested happens at milestone four, in months seven and eight. So the one mechanism that will actually pay live governance actors real ADA is selected two milestones before the machinery that exists to select mechanisms runs. The proposal is careful to say the pilot does not prejudge the final model selection, and this DRep accepts that entirely as a statement of intent. But the ordering is the structure, and the structure says the first mechanism tested in production was chosen by something other than the project's own published method.

The mechanism chosen is the one published by a consultant and research contributor to this proposal. That is not a wrongdoing finding and this rationale makes none; it is ordinary and often good that the person who designed something is the person brought in to test it. It is, however, an interest that should be disclosed on the face of the proposal and is not, and it interacts with the ordering above in a way that neither would produce alone: the project's own published criteria are not what put that mechanism into production, and the project's own team authored it.

The discretion around it compounds this. Before launch, "parameters, safeguards, eligibility rules, and implementation details may be refined," with material changes documented publicly. Publishing a change is not the same as being bounded by a limit. A rule that pays voters, whose parameters the funded party may revise up until the moment it goes live, with disclosure as the only constraint, is an open discretion that has been disclosed rather than a scoped one — and disclosure of an open discretion answers the question "is it known?" while leaving "is it limited?" untouched. The repair is to fix the specification and the eligibility rules before funds are approved, or to bind refinement to a published change-control with someone outside the project able to object and be answered.

2.6 Success criteria that cannot fail, and the missing quality term

Several of the proposal's stated success measures cannot register a failure, and at this request size that matters.

"Parameter validation: 100% of parameters are available" is satisfied by the parameters existing. "Developer Feedback: positive feasibility assessment" specifies its own required outcome, which means it is not a test. "Simulation Scenarios: 50+", "Public Workshops held: 10+", "Amount of meaningful Dashboards created: 5+", "Page Load Speed: <3 seconds" and "12 monthly public project updates" are counts of activity performed. They are all honest numbers and most of them are useful project-management targets, but they are not evidence about whether governance incentives work, and none of them could come out badly in a way that would tell the ecosystem something it needed to know.

Behind this sits the deeper absence. Nothing in the metric set can register a governance-quality failure. That is not a local oversight; it is the hard constraint this whole field runs into, and the proposal deserves credit for not pretending otherwise. The chain cannot verify, after the fact, whether anyone judged well. So every scheme must pay for a proxy, and the only real design question is which proxy and who is able to produce it. But a framework with no quality term at all has a determinable optimum, and the way to find it is to ask what maximises its metrics at minimum effort. If the answer is a farm, the framework is incomplete however rigorous the rest of it is. This is the failure that has already caught the process-first critique from the other side: correctly refusing a reified binary outcome measure and then optimising three activity variables with no quality term, under which a couple of hundred credentials auto-abstaining is the global optimum.

What is owed is a named falsifier, fixed in advance, in a variable the intervention does not deform — and, alongside it, an answer to the question the proposal never puts to itself: what result would cause this project to recommend against compensation, or against the mechanism it pilots? A program that cannot generate disconfirming evidence about its own subject cannot later be argued out of the ecosystem, and a knob that cannot be turned back is a wall.

The related absence is easier to fix and almost as consequential. Every one of the forty-nine catalogued efforts is a payment scheme. If the problem is that DReps are burning out reading a hundred proposals, then subsidising the burden and reducing it are both answers, and only one of them is represented anywhere in the landscape. A framework claiming to determine which governance problems are actually incentive-related ought to carry a category for the answer that costs the treasury least.

2.7 Who authors the measure

Two of the six rights this framework reads are engaged here in a way worth stating without the machinery.

The first concerns the honesty of the measuring instrument. This project would not merely observe governance; it would define the parameters, the categories of valued contribution, the reliability standards and the dashboard through which governance health is subsequently read — and it says so, accurately, as its central deliverable. That is a proxy being installed as an operative definition, and once a payment is attached to a proxy the proxy stops merely measuring value and starts constituting it. The proposal disclaims the drift honestly at the level of turnout. It binds itself to nothing that would detect the drift, and there is no route by which a participant whose standing the new measure will shape can contest, appeal or reverse it short of the single up-or-down governance action at month twelve, after the framework is built and the pilot has run.

The second concerns who is answerable to whom. The parties proposing to build the framework are the working group whose mapping document establishes the need, the authors of the problem statement cited as motivation, and — for the pilot — the author of the mechanism selected. Each of those is legitimate on its own; taken together they describe a body that authors the yardstick, applies it, evaluates the result, and reports the evaluation, with feedback incorporated "or addressed with rationale" at its own discretion. The financial audit allocation is real and welcome, but it audits money rather than method, and the ₳126,239 buys no independent verification of the pilot's findings or of the metric definitions.

Held against the standard this DRep applies to every field that exercises authority over participants — is power matched by accountability running the same direction — the reading is a partial inversion on both counts, and it fails on the same prong in both cases. The discretion is real, it is disclosed, and it is not limited. That failure names its own cure and the cure is not expensive: scope the discretion, and put a correction path in the design. An external evaluator of the pilot's findings, a published change-control with standing to object, and a defined remit for the parameter work would answer it.

2.8 What the commons retains

The treasury may fund the maintenance of commons relations; it may not fund the transfer of a private surplus. That is a test of claim structure rather than of who the recipients are, and this proposal does substantially better on it than the size of the ask might lead a reader to expect. There is no performance fee, no captured customer relationship, no token, no retained exclusivity. What is being bought is labour, at workstream rates, for a research program whose outputs are intended to be public — which is service compensation, and service compensation is the legitimate category.

The relational half of the value claim is genuinely present and should be credited: reusable datasets, pipelines, metric definitions and dashboard components that lower the setup cost of every later piece of governance research; a shared map that reduces duplicated effort across forty-nine initiatives; and decision-uncertainty reduced before permanent mechanisms are adopted. Those are coordination capacity, stated as coordination capacity.

What is missing is the part that makes the retention real. No licence is named anywhere — not for the dataset, not for the pipelines, not for the dashboard, not for the pilot script. Archival is promised "while website is maintained by us," which is the opposite of a commons asset: it is a service the ecosystem would depend on and does not hold. There is no handover path, no second maintainer, no repository or registry that survives the team. At this request size, a public good that exists at the pleasure of its builder has not yet been transferred to the public, and the framework's own question — what capacity does the commons retain, and how would its degradation be recognised — does not currently have an answer beyond the team's continued goodwill. That goodwill is not in doubt. Its durability is not a thing the treasury should have to rely on.

2.9 Why this is a No and not an abstention

Three of this framework's grounds for abstention were live and each is foreclosed, so the reasoning is set out rather than asserted.

The first is that the ballot cannot carry the finding. Much of the deepest analysis in this space genuinely cannot be reached by this vote: the formal negative result on proportional reward sharing, the impossibility result on saturation caps, the argument that pricing an unpriced trust signal destroys the property that made it worth using, the criterion that sorts funding sources by direction and live exit. Those are findings about mechanisms, not about this action, and they are recorded below as trajectory rather than converted into the token. But the three findings that carry this No — the undeclared denominator, the pilot's scheduling, and the mechanism chosen ahead of the method — are properties of this action, stated on its own face, and fixable in it. Where the ballot reaches the finding, the ballot is owed the finding.

The second is that the disagreement is a margin dispute about whether a relation is genuinely served. It is not, and this framework carries a specific warning against mistaking rigour for that margin. This proposal is careful, well-organised and methodologically serious, and it would be easy to read that seriousness as good faith contested only at the edges. Care about the method is not evidence about the denominator. The findings here are structural and each names its own repair, which puts them on the other side of the line.

The third is that this DRep should stand aside on grounds of interest, and this one deserves the most care because the interest is real. I am a small DRep. A pilot that pays DReps could pay me. A framework that governs DRep compensation would govern mine. That is not the general interest every participant holds by owning ada or using the chain; it is membership of the class the action pays. Two things nevertheless make standing aside the wrong answer. The interest is universal across the entire electorate — every DRep voting on DRep compensation is inside the paid class, so abstention does not produce a less conflicted decider, it only lowers the bar while the conflict remains. And the direction of my arguments runs against my own financial interest: the positions set out here and in my published work on compensation point away from the schemes that would pay the largest DReps most, and away from paying anyone sooner. I would rather that alignment be declared than discovered, and I would rather it be checked by someone who assumes the interest and reads for it.

There is a fourth consideration that is not an abstention ground but bears on the direction of the token. Withholding stake here would lower the passage bar for a program whose central defect is that it spends before the sequence is right. When the mechanical effect of withholding would assist the very ordering a finding objects to, withholding is not neutral, and neutrality was the only thing abstention was ever for.

The honest token is a No that specifies. And the finding that pulled hardest the other way — the one that came closest to changing this vote — is the one this rationale opened with: non-payment in a relationship participants read as asymmetric is not a safe default, and every month this stays unresolved, the volunteer framing hardens further into a role. That is why the No is remediable, why the specification below is short, and why this DRep would rather see a corrected version of this proposal in the next cycle than see the question go quiet.

2.10 Fiduciary verification (external instrument)

This DRep's own framework reads the relational and structural question and by design does not read the capital-allocation one — price, instrument fit, size-calibrated discipline, upfront exposure, opportunity cost. Leaving that region unexamined would raise vigilance rather than lower it, so the proposal was run through the DRep Treasury Rule Book v17 as a subordinate gate, one that can lower a merits-Yes but cannot raise a merits-No.

By economic substance this is public-good and civic-service work — independent research, governance and standards development, and community coordination — with a subordinate data-engineering component and one workstream, the pilot, that deploys a live payment mechanism rather than studying one. At ₳4,207,967 nominal it classifies very large, which is the band that requires exceptional public return, capped upfront exposure, stress testing and an explicit opportunity-cost case, and the band where a passing score is necessary and nowhere near sufficient.

The gate does not clear, and it fails before the scoring rather than inside it. The proposal does not name a contracting entity, a jurisdiction or a dispute path, which the rulebook requires at this size. It does not state a disbursement profile, tranche logic or upfront exposure, and unstated exposure at very large is treated as material opacity rather than as an open question. No administrator is named for a request of this size. No licence or rights position is stated for any of the assets, and continuity is promised only for as long as the team maintains the website, which fails the continuity and succession test outright. Milestone verification runs substantially through the project's own reporting; the independent allocation is a financial audit, not a verifier of method or findings. And several headline measures are activity counts of exactly the kind the metric-integrity rule exists to refuse. Independently of all that, the additionality question bites hardest: with the formal reward-scheme research already published, a live already-funded research program covering overlapping ground, and forty-nine catalogued efforts, a much smaller staged instrument — a denominator study, a map amendment, and a pre-registered baseline taken after the sensing change — would buy most of the decision-relevant learning at a fraction of the exposure, and the rulebook's own guidance warns specifically against one large master plan where the shared need is still uncertain.

So this is convergence at the conclusion rather than corroboration at a soft spot: two instruments built on unrelated foundations — a rights derivation and a capital-stewardship rulebook — reach No against the same structure, from opposite directions. Had the fiduciary read come back clean, the relational No would have stood unchanged; the gate's function here is to sharpen the resubmission specification, not to supply the verdict.

2.11 What would make this a Yes

The specification is deliberately short, and every item is cheap now.

Declare the denominator. State in the framework which population each measurement is drawn from, and that the register is currently produced in part by the reward-withdrawal condition rather than by governance intent. Add CIP-0181 and that condition to the landscape map, which is the map's own stated purpose.

Sequence the pilot behind the sensing change, or pre-register how a break in the series will be handled and say publicly that no series spans it. Say in advance that registered stake is expected to fall and that the fall is the instrument working.

Scope the discretion. Fix the pilot specification, safeguards and eligibility before funds are approved, or bind refinement to a published change-control with an external party able to object and be answered. Disclosure is not a limit.

Select the piloted mechanism by the project's own published criteria after the modelling milestone rather than before it, and disclose on the face of the proposal the authorship interest in the mechanism currently pre-selected.

Fix falsifiers in advance, in variables the intervention does not deform, and state what finding would cause the project to recommend against compensation or against its own piloted mechanism. Replace the success measures that cannot fail. Carry a quality term, or say plainly that none exists and what follows from that.

Add a cost-reduction category to the framework, so that reducing the deliberative burden sits alongside subsidising it.

And on the allocation side: name the contracting entity, jurisdiction and dispute path; state the disbursement tranching and cap upfront exposure; name a licence for the dataset, pipelines, dashboard and pilot script, with a handover path that does not depend on the team maintaining a website; fund independent verification of method and findings rather than only of money; disclose prior treasury receipts and the conflicts named above; and stage the ask, with a smaller first tranche covering the denominator work and the pre-registered baseline before the full program is committed.

Do those, and this becomes a proposal this DRep would vote for.

2.12 What the vote does not reach

A vote is a snapshot; the log is the trajectory. Several of the strongest findings this reading surfaced are about the mechanisms in the landscape rather than about this action, and converting them into this vote would misrepresent both.

The formal result on proportional reward sharing belongs here first (IOR research paper), because it bears on the mechanism family most likely to be recommended. Modelling DReps who choose how much effort to invest, where effort both raises the chance of a correct vote and attracts delegation, the analysis of the rule that pays each DRep a share of the budget proportional to their share of total delegation finds effort per DRep scaling as one over their number, and the probability of the community selecting the correct outcome tending to one half as the number grows — worse, not better, under concave costs. The authors report being surprised, since proportional sharing is the natural rule and performs well elsewhere; it fails here because a fixed prize pool with many entrants produces under-investment by everyone. The historical precedent usually cited in favour of pay-by-delegated-stake is the same rule. Their preferred alternative, a threshold paying a fixed amount to any DRep attracting at least a defined share, performs far better on effort and is structurally hostile to credential splitting, which is the opposite property to a saturation curve. This proposal already commits to reproducing that work, which is to its credit; the trajectory item is whether the ecosystem's discussion absorbs the result.

Two structural readings sit alongside it. Delegated stake resists gaming largely because nothing pays for it; attaching income to it is the pricing event, after which it becomes an asset with a calculable return and is rational to acquire up to the point where acquisition cost meets expected return — through marketing, delegation incentives or revenue sharing, none of it fraud. The property that makes the metric attractive is the property that paying for it removes. And a saturation cap does not repair this: the impossibility result on sublinear voting power shows that whenever a wallet of any size yields nonzero power, splitting recovers power growing at least linearly in holdings, whatever the splitting, voting or setup costs — so the curve fails to bind on anyone large enough to split and binds only on those too small to bother, which is a regressive incidence rather than a neutral failure. Registering a DRep credential is cheap and carries no pledge, so the discipline stake pools have against splitting does not exist here.

The observed precedent belongs in the log too: in Uniswap's delegate reward program, paid delegates participated strongly while quorum fragility and voting-power concentration remained unresolved. Participation among the compensated population rose; the governance health the participation was supposed to indicate did not follow. That is what it looks like when a payment restores a number without restoring what the number stood in for, and the landscape map contains it already. Anyone familiar with Cardano’s Project Catalyst saw the same thing happen with Community Review: the goal was crowdsourced deliberative capacity to inform the voting, but incentives actually reduced deliberative capacity and introduced gaming and extraction.

One open question in the landscape is higher-leverage than anything else in it, and it is open rather than settled: whether concentration indices become the primary success criteria for an incentive pilot. This proposal, in treating concentration as one of several evaluation dimensions, does better than most; the trajectory risk is that the question gets answered by default in the course of building a dashboard rather than deliberately.

The reading also records something that cuts the other way and should not be buried because it is inconvenient to the token. Compensation functions as a regime-keeping device: uncompensated labour in a relationship perceived as asymmetric converts contributors into role-holders whose giving becomes expected rather than owed, and neither of the ecosystem's two problem statements says so. That is a finding in favour of solving this problem soon, and this DRep carries it as an obligation rather than as a concession.

Finally, two things to watch. The direction-and-exit criterion — whether payment flows from the party the DRep serves, and whether that party holds a live exit — sorts the four candidate funding sources quickly and without requiring agreement on compensation in principle, and this DRep intends to keep arguing for the criterion rather than for a preferred answer. And the sharpest single sensor available: when registered stake falls after decoupling, does the ecosystem read it as the instrument beginning to work, or as a crisis requiring a mechanism to restore the number? That reading will say more about whether participation-as-a-target has been dislodged than any argument made before it.

Two of these are candidates for a future proposal rather than a future vote: the missing cost-reduction category, and a standard requiring any incentive proposal to declare its denominator before it is assessed. This DRep will be monitoring the first, and intends to draft the second.

---

3.0 Conclusion

We are voting NO (remediable).

The problem this proposal names is real, the people behind it have done real work, and the refusal here is not a refusal of governance compensation, of research into it, or of this team. It is a refusal of this sequence. The project would measure a register that is currently manufactured without saying so; it would run its most expensive experiment across the change that unmakes that register; and it would pay live governance actors under a mechanism selected two milestones before the method meant to select mechanisms, under a discretion that is disclosed rather than bounded. Run separately, the external capital-allocation instrument reaches the same conclusion against the same structure, on missing entity and disbursement terms, an unlicensed public asset, continuity contingent on a website, and success measures that cannot fail.

Each of those is repairable, and the repairs are paragraphs rather than redesigns. A declared denominator, a pilot sequenced behind the sensing change, a mechanism selected by the project's own criteria, falsifiers fixed in advance, a licence and a handover, and a staged first tranche. That is the whole specification, and it is cheap now and unrecoverable after the money moves.

What holds this DRep's posture together is easy to state and this rationale will keep to it: support the problem, contest the sequence, supply the instruments. The problem is real and the need is not in doubt. What is contested is order and proxy, and the constructive half of that is owed rather than optional, which is why the denominator decomposition, the delegator-authored signal family that is buildable today from data already in the ledger, and the missing cost-reduction category are all offered here rather than held back. This DRep would rather help this proposal succeed on resubmission than watch the question go quiet, and will say so publicly to the proposers as well as here.

DRep ID: drep1yfaq8dsam...3nq50q 

DRep Profile: https://dreptalk.com/dreps/styg-nq50q/

My thanks to the delegators who make this reading possible, and to the people behind this proposal, whose work on the problem preceded almost everyone else's.

Stay in touch. X: https://x.com/styg50