The decision frame The thread has represented this as a custody problem. Should the treasury move to the Foundation, stay with the DAO, or move in tranches? A representation of that kind settles in advance which alternatives are visible, and this one has hidden a whole class of them. Before asking which option is best, I would ask how the problem came to be framed this way. Take the temp check's own stated goal, "a DAO that votes rarely and matters every time it does." That sentence describes an allocation of attention. Voting rarely carries a cost the sentence conceals: a delegate asked to decide a consequential and largely irreversible question every eighteen months arrives at it cold, with no accumulated representation of the problem, no rehearsed alternatives, and no practised judgement of the operator. Rarity rations the very input that scrutiny is made of. Where control actually lives In any organization, the person who decides is seldom the person who determines the decision. Whoever sets the agenda, supplies the factual premises, and defines which alternatives count as available has shaped the outcome well before the vote is called. Administrative behaviour is governed at the level of premises, and formal authority follows along behind. This is why the custody safeguards on offer reassure me less than they appear to reassure others. Keep the keys onchain, cap withdrawals at five percent per year, tranche release against milestones: under every one of those variants the empowered Foundation still supplies the premises. It writes the reports, frames the options, and decides which questions reach a vote at all. A DAO holding custody without independent search capacity has kept the authority to ratify while giving away the capacity to choose. Let me be precise about the scope of that claim. It carries no accusation of bad faith and does not depend on one. The same structural result would follow with an exemplary board, because a single organized searcher produces a single well-argued alternative. That is what organized search does. Three amendments Each is cheap, reversible, and binds identically whether the DAO transfers custody, retains it, or tranches. That last property is the important one: they can be adopted now, without waiting for the custody question to settle. 1. A decision premise register Every Foundation request that reaches a DAO vote arrives in a fixed format. It states the decision to be made and the level at which it belongs, the alternatives considered and rejected with the reason for each rejection, every material factual premise with its provenance, and the evidence that would change the recommendation. The rejected alternatives carry the weight here. A recommendation arriving with no visible discards tells you either that the search was never run or that its results were withheld, and the register makes that difference legible. This bears on a pathology already visible in the debate. Treasury figures ranging from roughly $130M to "half a billion" circulate interchangeably, depending on whether the speaker means the Endowment, the full treasury, or treasury plus locked token supply. Under a premise register, a figure lacking a stated scope and source stops being a contested claim to argue over and becomes an incomplete filing, returned before the vote. 2. Funded counter-search Allocate a small standing retainer, one to two percent of the operating envelope at whatever size it eventually takes, to a rotating analyst charged with producing the strongest available case against each material Foundation request and publishing it alongside. The mechanism confers no oversight powers and no veto. It supplies a second searcher. Delegate scrutiny fails for want of attention and preparation far more often than for want of motivation or correct incentives, and paying for the counter-case is the cheapest way to hand delegates an alternative they did not have to generate themselves. 3. Ex-ante aspiration levels The temp check leaves underperformance to be defined after the fact, which in practice means it will never be defined at all, since every outcome admits some post-hoc account under which it was satisfactory. Before empowerment, name three to five observable conditions that would constitute adequate performance at twelve and twenty-four months, and fix a calendar date on which the arrangement reopens by default. A recall mechanism would demand a mobilized majority, which is precisely what this structure is designed to avoid needing. A scheduled reopening demands only a date. Reversibility contingent on someone organizing an insurgency will never be exercised. The limits of this argument None of these measures settles custody, and they should not serve as a way to avoid settling it. A DAO that adopts all three and transfers the whole treasury outright has still made a largely irreversible commitment under uncertainty, and I would still prefer the staged alternatives others in this thread have described. My claim here is narrower. Without a premise register and a funded second searcher, every custody arrangement on the table converges on similar behaviour, because each of them leaves the input side of the decision in one pair of hands. What would change my mind Show me that the Foundation charter already requires published rejected alternatives and sourced premises, and amendments one and three become redundant, at which point I withdraw them. Show me that delegate non-participation runs on apathy rather than preparation cost, and amendment two buys nothing. That second possibility is why the retainer deserves two cycles of funding and a review, with no presumption of renewal. The learning loop Adopt all three for four quarters. The evidence to collect is narrow and cheap: how many requests come back as incomplete filings, whether the counter-analyst's case ever changes a vote outcome or the contents of a request, and whether delegate participation on Foundation items rises against the current baseline. If none of the three moves, they were procedural decoration and should be repealed. I would rather run that experiment and learn I was wrong than argue the point for another eighteen months.