This is the running log of everything this site has published that was wrong about its data, its process or the rules. Entries are added, never removed and never softened, and each one names the page, quotes what it said, states what was true, and says what changed so it can't recur.
The bar for an entry is a reader who could have formed a false belief from the page; an internal bug alone doesn't qualify. A wrong number on a live page qualifies. A typo doesn't, and a design somebody disliked doesn't unless it misled.
The deadline calculator, the case tracker and the deadline guides · Oct 16, 2026
It said: The ETA 9089 filing window always closes when the prevailing wage determination expires, if that comes before 180 days after the first recruitment step. The tracker would not save a filing dated after the expiry, and closed a case automatically when the determination expired before the filing.
What was true: 20 CFR 656.40(c) asks employers to "file their applications or begin the recruitment period" while the determination is valid, and either is enough. Recruitment that begins on or after the determination date and on or before its expiry can be filed after the determination expires, inside the 180-day window. The Board of Alien Labor Certification Appeals read the rule this way en banc in Karl Storz Endoscopy-America (2011-PER-00040, December 1, 2011). The expiry still closes the window early when recruitment began before the determination was issued. The August 29 correction below fixed a calculator that ignored the expiry, and went too far the other way.
What changed: One rule in the code decides when the determination still limits a case, and the calculator, the case form, the deadline reminders, the automatic closing and the guides all read it. The calculator asks for the determination date, so it can tell which case you're in. See the corrected calculator.
The case status page · Sep 7, 2026
It said: "Status seen" quoted a July date for pending cases, telling a beneficiary that DOL had not looked at their case since summer.
What was true: The date came from the retired mirror's stamp. The site's own sweep had checked every pending case daily since August 26.
What changed: The date is now the sweep's own finish date, read from the coverage record the sweep writes. A test pins the source. See the corrected page.
The privacy policy · Sep 6, 2026
It said: That Sentry Session Replay was in use, and nothing about two other third parties.
What was true: Session Replay had been removed on August 29. Ahrefs Web Analytics and the Senja review widget were present and unlisted.
What changed: The policy names what actually runs, and is reviewed whenever a script is added or removed. See the corrected page.
The RFI and audit hub · Sep 6, 2026
It said: A two-case PERM review stage that did not exist.
What was true: Two prevailing wage request rows had leaked into the PERM table through the lookup path before a guard was deployed, and were counted as a PERM stage.
What changed: The rows were deleted and the lookup refuses non-PERM prefixes for that table. The stage list is derived from a static vocabulary. See the corrected page.
The review stage pages · Sep 6, 2026
It said: Median ages that stopped at a date the sweep never updated.
What was true: The ages were measured from filing to a stamp the retired mirror had last written in July, or to nothing at all for 12,187 cases.
What changed: Age is filing date to today on every stage page and in the census that feeds them. See the corrected page.
llms.txt and the site-wide structured data · Sep 5, 2026
It said: "Check any PERM case number", after every visible page had been corrected two days earlier.
What was true: The site looks up prevailing wage requests and H-1B LCAs as well, and the two most machine-read surfaces on the site still described only PERM. Google's AI Mode had already told a reader we couldn't check a pending wage request.
What changed: Both surfaces name all three programs, and a test fails if either stops doing so. See the corrected page.
41 places across 34 public pages, the homepage FAQ first · Sep 3, 2026
It said: That this site checks PERM case numbers, in words that read as PERM only.
What was true: P- and I- numbers were already looked up live. An answer engine lifted the homepage's product definition and told people we couldn't do what we did.
What changed: Every surface was rewritten to name what can and can't be looked up, and the guides state what each page can't tell you. See the corrected page.
The homepage trust badges · Sep 2, 2026
It said: "Real-Time Updates" and "DOL Compliant".
What was true: Status is checked daily, not in real time. "DOL compliant" means nothing for a site that reads DOL's published data.
What changed: The badges say what is true: federal data only, checked daily. Badges are audited whenever a source or cadence changes. See the corrected page.
24 employer pages · Aug 30, 2026
It said: That the employer "also filed as" a name identical to the page's own heading.
What was true: The two spellings differed only in whitespace, which a browser collapses, so the page was presenting one spelling as two.
What changed: Spellings are compared as rendered. Case and punctuation differences are still shown, because a reader can see those. See the corrected page.
The RFI and audit hub and its stage pages · Aug 30, 2026
It said: The same count with two different dates: 965 cases "as of August 30" on the hub and the same 965 "as of August 27" on the stage page.
What was true: Both were the same census; the hub had taken the latest date across all stages and the leaf its own.
What changed: Both read the per-stage date. A figure and its stamp are one claim. See the corrected page.
The PERM deadline calculator · Aug 29, 2026
It said: A filing window that closed at first recruitment plus 180 days, on cases where the wage determination expired earlier.
What was true: The window closes at the earlier of the two, and the tool had computed the raw arithmetic without the cap, printing close dates on which filing was barred.
What changed: The tool calls the canonical composite that carries the cap, and says when the cap applied. See the corrected page.
The denial risk page · Aug 27, 2026
It said: A finding that denial rates rose in a particular wage band.
What was true: The pattern came from how the bands were cut. The cases didn't carry it.
What changed: The finding was retracted on the page and the bands were re-cut; the page says why one number isn't offered as a score. See the corrected page.
Several data pages · Aug 27, 2026
It said: That the site's per-case data was second-hand, mirrored from another tracker.
What was true: By that date every dataset was read from DOL, USCIS or the State Department directly.
What changed: The copy names the federal source on each page, and the methodology page lists every dataset with its source and cadence. See the corrected page.
The priority date chart · Aug 25, 2026
It said: One caption, "a month with no visa numbers at all", for two shaded states, and a line drawn through months when the category was closed.
What was true: "Current" (open to every date) and "Unavailable" (closed to all) are opposites, and a closed run breaks the series, because nothing moved during it.
What changed: The two states have different colours and their own captions, and the line breaks at every gap. See the corrected page.
The I-140 processing time table in the case tracker · Aug 23, 2026
It said: EB-2 national interest waiver petitions took about 7 months, and offered a choice of service center.
What was true: USCIS was publishing 29 to 32 months, and no longer reports I-140 by service center at all. The table had not been touched in 16 months.
What changed: The table is keyed on USCIS's real subtypes, carries its as-of date, and a test fails when that date is more than eight months old. See the corrected page.
