A whistleblower dates the USPS ballot portal build to 58 court-permitted days

The dates carry this story more efficiently than the adjectives do. Work opened on June 15 and shut ten days later under Judge Indira Talwani's first order, then restarted on July 29 when the administration appealed ^[3], so 10 of the first 44 days were permitted development, about 23% ^[2]. Take the three-month window the disclosure describes ^[2] and the stoppage eats 34 of roughly 92 days, leaving 58 ^[1], which against a normal year or more of build time is compression nearer six to one than the four to one the three-month framing implies ^[5].

Batch processing is a choice about where correlated failure sits, and the whistleblower puts it at the batch, where one bad scan in a lot of tens of thousands of envelopes stops all of them ^[4]. Set that against the exposure the design is meant to reduce: Brookings counted about four cases of mail voting fraud per 10 million ballots ^[9], one per 2.5 million ^[3], so a single failed lot of 20,000 withholds five thousand envelopes for every fraud case found across that 10 million ^[4].

The term I would flag first in a review, or rather the more interesting version of it, sits in the August 21 final rule, which conditions ballot delivery on voter lists submitted through a portal the rule itself treats as not yet active ^[5]. The portal exists to receive state lists of verified voters ^[8], so a state's compliance runs through a system no state controls, and non-delivery still arrives as the state's failure.

The counter-thesis belongs in the same paragraph as the thesis. The technical account comes from one unidentified whistleblower ^[16] routed through a senator ^[1] whose party is a plaintiff in the case ^[10], USPS declined to comment and the White House did not respond ^[12], and the White House's standing position is that the mail provisions are commonsense measures against fraud ^[13]. Blumenthal's line about a system designed to disenfranchise millions of Americans ^[14] is a claim about intent, not a finding about code. The framing is loose in smaller ways too, since the same account leans on one-third of voters while the 2024 figure given is more than 29% ^[1][15], a spread of about four points ^[6].

If the appellate track clears implementation again, as the Supreme Court briefly did last week without touching legality ^[10], the portal goes live partly built and failures land unevenly by state. If Talwani's 14-day order ^[11] hardens into something longer, or the merits go against the directive ^[17], the build becomes stranded cost that never meets a ballot. The third branch is the one I would weight, and this is probably wrong: the software mostly works, and the binding constraint turns out to be the approval queue, with North Carolina mailing on Friday ^[6] and states not certified in time. What would falsify that is a live portal returning rejects at envelope level, plus a visible run of state envelope approvals cleared inside a week.