Bexar County revisited

Last night, I watched an interview with Dr. Walter Daugherity (@ZoomWalter on X) on Rumble. Dr. Daugherity is the person who initially sent me the Bexar County early voting check-in records for February 17 and 18. In the interview, Dr. Daugherity mentions that the only reason CD-21 candidate Weston Martinez was able to obtain the file and send it to him is that Texas has a law requiring distribution of the information no later than 11 A.M. of the day following each day’s voting.

The file contained exactly 4,110 fraudulent records for the second day of early voting, February 18. Those records were based on the full list of genuine voters who had voted that day. To make them, the software or algorithm needed the full day’s list of voters. Meaning, the fake records were created after close of voting at 6PM on February 18, then posted the next day befor 11 A.M.

Once the file was posted on the county website, it was downloaded and distributed by the Republican party chair to every candidate appearing on the ballot. This included Martinez, but also the thirteen other candidates he was running against, and candidates in other races, like John Cornyn and Ken Paxton, the current senator and attorney general of Texas, respectively.

Of this group of people, the only person to notice the irregularities in the file was Weston Martinez. Martinez then took the next step, and had the file examined by Lori Gallagher, Walter Daugherity, myself, and possibly others.

A few days after I received the file, Martinez sent me a new version of the file, downloaded on February 25, and with a February 25 creation date in the metadata, that didn’t contain any fraudulent entries. Or at least, not the fraudulent entries found earlier. They had been erased.

The county explained this as an unspecified glitch that occurred during export of the file. However, the explanation doesn’t come close to credibly explaining the complex logic used to generate the fake records based on real ones. They may believe the explanation due to lack of knowledge, but regardless, it is false.

These facts raise two questions worth examining carefully. The first: why were the fraudulent records still publicly visible when they were erased, and is the Texas disclosure law unique in a way that may have created a vulnerability for fraud software designed for other states? The second: why was Martinez apparently the only person to notice this?

---

Question One: Why were the fraudulent records publicly visible — and is the Texas law unique?

To answer the second part first, because it illuminates the first: Texas’s requirement appears to be unique. Research into election law across all fifty states found no equivalent provision. Texas Election Code Section 87.121, amended in 2019, requires that early voting check-in rosters be posted publicly on the county website no later than 11 a.m. on the day following each day of voting. No other state mandates proactive, automatic, next-morning online publication of check-in records during an active early voting period. Other states have general public records laws that allow access to voter registration files upon request, and most publish voter history after an election concludes. None require the check-in data to appear online the morning after each voting day, while the election is still in progress.

This matters because it is almost certainly the reason the fraudulent records were ever seen at all.

The injection required the complete list of genuine voters who checked in on February 18. The algorithm that generated the 4,110 fake records needed every real voter’s name, address, and alphabetical position in the full day’s roster before it could produce a single synthetic entry. This is not an inference — it is a mathematical necessity proven by the structure of the injected data itself, as documented in my forensic report. The injection could not have occurred until after polls closed at 6 p.m. on February 18. The fact that fraudulent records were added to a poll book after voting had concluded is itself evidence of criminal intent. There is no legitimate reason to add algorithmically generated voter records to an election system at all, and certainly not after polls have closed.

The file was then present in the system when the county posted it the following morning, as Texas law required, before 11 a.m. on February 19. The file metadata confirms it was still available on the county website at 5:13 p.m. that same day, when it was downloaded by Terri Richardson, the Republican precinct chair for Bexar County precinct 3042. The metadata lists Richardson as the file’s author and shows a save time of 5:46 p.m. on February 19 — consistent with the file being renamed to something human-readable shortly after download.

At some point after that, the file was replaced with a clean version. The replacement is confirmed by a striking exchange published in the San Antonio Current on February 20, 2026. In that article, in reference to allegations lodged by Martinez of fraud, Bexar County Elections Department Administrator Michele Carew stated: “After thorough review, we found no evidence to support these claims.” If that review was conducted on February 20, and it found nothing, there is only one explanation consistent with Carew being honest: the fraudulent file had already been replaced before she looked. The injected version was live on the county website on the afternoon of February 19. It was gone by the time the county reviewed it on February 20. The replacement occurred in a window of roughly twelve to eighteen hours.

Why was the replacement not executed before the file was ever posted — before 11 a.m. on February 19? This is where the history of KnowInk’s Poll Pad system becomes directly relevant.

KnowInk was founded in 2011 and had sold approximately 65,000 Poll Pads across dozens of states by the end of 2019. It was first certified for use in Texas on January 23, 2020 — the same year Section 87.121’s amended posting requirement took effect. Bexar County itself only adopted the KnowInk system in May 2025, nine months before this primary. None of the states where KnowInk had been operating for years before Texas had an equivalent daily disclosure law. Texas’s posting obligation falls on county clerks, not on KnowInk, meaning there was no automatic mechanism by which KnowInk’s engineers would have been alerted to it as a constraint affecting their data workflows.

The most straightforward reading of the evidence is that whatever governs the timing of file replacement in the ePulse system was calibrated for a world where no state forced same-day public disclosure of check-in data — because for most of the system’s operational history, no state did. The Texas law created an exposure window that had never existed anywhere else. The file that should have been quietly replaced overnight was instead publicly available on a county website, as required by law, before anyone had a chance to act on it.

This raises a further question that cannot be answered from the available evidence but deserves to be asked: how long has this system, or something like it, been operating in states without Texas’s transparency requirement? KnowInk’s Poll Pad ran in 29 states during the November 2024 presidential election, processing one in four in-person voter check-ins across the country, all through the same centralized ePulse platform. If the injection capability embedded in that system has been operational since before Texas came online in 2020, it has had years of exposure in states where no law would ever force the compromised file into public view before deletion. The Bexar capture was not the product of a well-designed oversight system. It was an accident, made possible by one state’s unusually aggressive transparency statute and the narrow window during which one person happened to look.

---

Question Two: Why was Martinez the only person to notice?

The answer has two parts, and together they are more illuminating than either one alone.

The first part is structural. The file reached Martinez through a specific chain of custody that differed from how every other campaign received it. Republican precinct chair Terri Richardson downloaded the file from the county website in her official capacity and distributed it through the party network to candidates. Most campaigns receiving it would have processed it operationally — piping it into voter file management software that automatically cross-references names against their contact lists, updates “already voted” flags, and routes canvassing resources for the remaining days of early voting. In that workflow, nobody opens the raw file. Nobody sees the State ID column. The software extracts what the campaign needs and the file disappears into a database. This is not negligence. It is exactly how these tools are designed to be used, and it is how every well-funded campaign on the distribution list almost certainly handled it.

The second part is about who Weston Martinez is. The Ballotpedia data on the CD-21 race is instructive. Martinez raised $19,907 for his congressional run — less than one percent of what frontrunner Mark Teixeira raised ($3.46 million), and a fraction of what mid-tier candidates spent. He received 2.1% of the vote, 1,758 ballots. By every conventional campaign metric, he was a long-shot candidate with no professional campaign infrastructure to absorb the file before it reached him.

What he had instead was orientation. Martinez has been running for Texas offices since 2016, consistently on platforms centered on government accountability and election integrity. In 2018 he received 24% of the vote statewide in a Republican primary for Texas Railroad Commissioner on essentially no money — a result that reflects name recognition built on a specific constituency of voters who share his preoccupations. When the check-in file reached him, he was not asking what it could tell him about voter contact optimization. He was asking whether it looked right.

It did not.

He gave the file to Lori Gallagher, who passed it to Dr. Walter Daugherity of Texas A&M, who brought it to me. The analysis is now public. The forensic window — brief, accidental, created by a transparency law that no other state has — was enough.

The lesson this carries for the broader question of detection is stark. The fraud was not invisible because it was cleverly concealed from human examination. It was invisible because the operational context in which check-in files are routinely used makes direct human examination essentially unnecessary. Texas’s transparency law created the one condition under which detection was possible: a raw file in the hands of a person with the motivation and presence of mind to actually read it.

That condition had apparently never existed before in any other state. It existed here because of a specific person, a specific orientation, and a state law that the operators of this system apparently failed to account for.

---

A note on the official response

The San Antonio Current article published on February 20, 2026 also quoted former Bexar County GOP Vice Chairman Kyle Sinclair, who was himself a candidate in the same race. Asked about Martinez’s allegations, Sinclair said: “Excuse my language, but he’s full of shit.”

Sinclair’s dismissal was published before any independent forensic analysis had been conducted or released. The mathematical proof that the 4,110 records were algorithmically generated — the exact integer result of span ÷ gap = 4,109.0000, the palindromic floating-point cycle, the alphabetical sort structure, the clone distribution constraint — none of that was public yet when he spoke. It is now. The proof is in the data, and the data is in the report.