Corrections

Every correction we have ever made, kept forever — including the embarrassing ones. Especially those.

When we get something wrong, we fix it, mark the change at the top of the affected article, and log it here permanently. We do not quietly edit. We do not remove corrections from this log. We do not blame "the AI" — errors are attributed to Claridas as an organization, and the process that caused them is fixed.

• Colorado had Monday's second-highest expected offense of 20 lineups — and scored the fewest runs of all 20: none

  De-scope after Alden cross-LLM HOLD (6/10). The core spine reproduced independently and is unchanged: Colorado's team xwOBA .387 ranked second of the 20 slate lineups, and the Rockies were shut out 9-0. Three framing errors were corrected at source: (1) team xwOBA was relabeled from a contact-quality / 'hit the ball second-hardest' ranking to what it is — an expected-offense rate on the on-base scale, built from quality of contact and from non-contact outcomes (walks, strikeouts); (2) xwOBA and runs are no longer compared as commensurate units — the .025 rate gap and the 9-run gap are stated as distinct measures on distinct axes; (3) the causal claim that sequencing/timing 'explained' the margin was removed — no run-expectancy model supports it, so cause is left [speculative]. Notable was rebuilt with external reporting; data endpoints moved to Sources; Statcast row count corrected to 3,001 data rows. All figures re-verified 2026-08-15 against Baseball Savant and the MLB Stats API. No prior public version; this is a pre-publication de-scope. 2026-08-15 — EDITORIAL-VOICE-v2 pass applied (v2 voice pass, no facts changed).

• You'll remember Caminero's home run — but the play that turned the Athletics' 96% game was a soft single two batters later

  2026-08-15 — De-scope of win-probability claims after Alden cross-LLM review (HOLD 4/10), re-verified at source against the MLB Stats API winProbability/playByPlay/linescore feeds (gamePk 824969) and the ESPN summary feed (event 401816477). Four corrections: (1) the ~97% peak was reached late in the seventh after Jacob Wilson's two-out double while the Athletics were still batting, not with six outs to go; the six-outs-to-go value was 96.4% (MLB) / 96.5% (ESPN). (2) The 93.8% (MLB) / 93.4% (ESPN) figure is the value after Yandy Díaz's leadoff single in the eighth, not the pre-inning baseline. (3) Removed the false claim that every plate appearance moved the number more than the one before; replaced with the reproduced per-at-bat deltas, which are not monotonic. (4) Dropped unsupported assertions of model independence, an observed loss-rate calibration, and specific model inputs; win-probability figures are now framed as one model feed's published output and tagged [modeled] throughout. Headline changed "97%" to "96%." The box-score facts and rally sequence were re-confirmed unchanged. 2026-08-15 — EDITORIAL-VOICE-v2 pass applied (voice only; no facts, numbers, dates, or confidence tags changed).

• We read all 231 Senate roll calls of 2026 at once — nomination business drew 94, bill passage 6

  Corrected 2026-08-15. The first draft classified nomination votes from the top-level PN field only. That parse missed four en-bloc roll calls stored under nested en_bloc/matter records (votes 124, 125, 221, 225) and wrongly counted vote 66, a motion to proceed, as a confirmation. The nomination universe was rebuilt from the nested records. Nomination-related roll calls: 90 → 94. Nomination matters confirmed: 47 → 170 (47 individual + 123 en-bloc). Nomination share: 39.0% → 40.7%. Cloture: 68 → 70 (54 agreed → 56 agreed; 14 rejected unchanged). The old "1.91 votes per confirmed nominee" ratio was an artifact of the miscount and is removed. The nominations-vs-bills contrast survives and widens. 2026-08-15 — EDITORIAL-VOICE-v2 pass applied (v2 voice pass, no facts changed). 2026-08-15 — Cross-LLM HOLD remediation (no verified figure changed). Headline and deck rewritten to keep three units distinct: nomination business (94 of 231), final confirmation votes (49), nomination matters confirmed (170). The 170 count is stated as distinct PN matters (nomination records), not a headcount of persons; the file lists PN matters and one nominee can hold more than one. The 12 final legislative-type dispositions are itemized (6 bills + 2 simple + 2 concurrent + 2 joint resolutions). The objection-mechanism and "overwhelming" wording were scoped to what the tally shows.

• We read all 68 reporters: the low-income trade-tax median (15.3%) is 16x the high-income median (0.95%); US 46th of 68

  2026-08-15 (de-scope; pre-publication, non-live). Cross-LLM review (Alden, HOLD 4/10) held the draft for over-reach around a reproducing 2024 core. Fixes: (1) corrected the closer's rank fraction from 'three governments in four' (75%) to '45 of 68 — about two in three' (66.2%); (2) replaced the narrowed definition with the World Bank's exact indicator and denominator definition; (3) scoped every ranking, median and comparison to the 68-reporter panel and removed world/every-country generalization; (4) dropped the unsupported claim about why data is missing and the claim that the true global distribution is 'wider,' replaced with 'coverage prevents inference'; (5) separated the 2024 observation from later tariff policy and removed the 'now raising tariffs' bridge; (6) demoted SACU, EU own-resources and small-open-economy mechanisms from asserted causes to [modeled] hypotheses and removed unverified fiscal/enforcement/leverage language ('load-bearing wall,' 'survival,' 'existential,' 'cheapest to collect,' 'hardest to enforce'); (7) disclosed all 68 reporting ISO3 codes and the 11 removed aggregates in the methodology for auditability. Central 2024 arithmetic unchanged and re-verified at source. 2026-08-15 — EDITORIAL-VOICE-v2 pass applied (voice only; no facts, numbers, dates, or confidence tags changed).

• Eight of the 34 League Cup first-round ties went to penalties — every one was level at 1-1 after 90 minutes

  2026-08-15: Rebuilt on the complete first round. The original piece scoped the round to the 8–9 August subset (30 ties) and called it the whole round. The complete first round ran 7–10 August and comprised 34 ties (three on 7 August: Wolverhampton Wanderers 3-0 Port Vale, Wycombe Wanderers 1-2 Stevenage, Middlesbrough 1-0 Wrexham; one on 9 August: Mansfield Town 0-3 Sheffield United; one on 10 August: Plymouth Argyle 2-0 Exeter City). Whole-round figures were re-derived on all 34 ties: 84 goals (was 75); 8 of 34 shootouts, 23.5% (was "8 of 30, more than a quarter"); 19–15 home/away advances, 55.9% (was 16–14, 53.3%). The claim that Premier League clubs "enter in later rounds" was corrected: Wolves and West Ham played the first round. The all-eight-shootouts-were-1-1 finding was re-verified on the complete round and holds (all 8 shootout ties were 1-1 after 90 minutes). Derby County's regulation loss corrected to 1-2 (was "2-1"). Source: ESPN Carabao Cup scoreboard/summary APIs, dates 7–10 August 2026, cross-checked against the 2026–27 EFL Cup Wikipedia first-round results. 2026-08-15 — EDITORIAL-VOICE-v2 pass applied (v2 voice pass, no facts changed).

• The US inflation index has no October 2025 — in our sample, the three survivors line up with BLS's non-survey series

  Pre-publish de-scope (candidate not live; live URL 404 verified — no Correction Protocol event). Editorial remediation after Alden cross-LLM HOLD 6/10 (review artifact: cross-llm/cpi-october-2025-hole-selective-survivors-2026-alden-live-review.md). Removed the universal 'a weighted total goes silent the moment one input does / all-or-nothing / cannot be computed' aggregation mechanism from the headline deck, analysis, disagreement, view-from and closer; replaced it with BLS's own sourced reason (survey data for October 2025 could not be collected during the lapse, so BLS did not publish all-items or core estimates, while a few non-survey-built indexes could still be published) per https://www.bls.gov/bls/bls-handling-of-missing-data.htm [verified]. Scoped 'three survivors' to 'three of the narrower indexes we queried' throughout; scoped 'every major group' to 'six of eight major groups we queried' with the two unqueried groups named. Re-tagged the per-series survey/non-survey collection-mode causal chain as [modeled]. Softened volatility/weighting/representativeness language to the four series actually compared, with each Jun-to-Jun figure sourced. Corrected the 12-month-comparison wording to October-as-endpoint. Split the four grouped BLS API source entries into one re-runnable endpoint per named series ID, and added the BLS missing-data notice as the load-bearing methodology source. All BLS values re-pulled and confirmed at source on 2026-08-15. gradingScore/benchVerdict/crossLlmVerdict left unchanged for Forge to re-gate. 2026-08-15 — EDITORIAL-VOICE-v2 voice pass, no facts changed. 2026-08-15 — cross-LLM HOLD (Alden 7/10) remediation, no verified figure changed: (1) corrected the analysis 'smaller irony' comparison — the prior 'two of the three survivors moved more than the two core anchors' was false (used cars -1.8% moved less than both rent +2.8% and medical +2.0%); rewritten to state only what the four figures support (gasoline +26.7% is the largest of the four, used cars -1.8% the smallest) and dropped the unsupported 'leans toward the more volatile end' generalization; (2) scoped headline, subhed and opening to the queried sample (not 'every series') and re-marked the survivor→non-survey mapping as [modeled] inference rather than asserting it as fact in the headline/lede; (3) added the primary lapse/revised-dates source https://www.bls.gov/bls/2025-lapse-revised-release-dates.htm to notable[] and sources[] (October 2025 CPI withheld; November 2025 release omitted the 1-month changes anchored on the missing month).

• A plane pushes back late on the last row of its scheduled sequence ~3.4x as often as the first — hold the hour fixed and it is a morning thing

  2026-08-15 editorial de-scope (Alden HOLD 5/10 remediation): (1) Removed causal-decomposition framing. The scheduled-hour split is now presented as a descriptive hour-conditional association, not a decomposition of aircraft-history vs time-of-day cause. Headline, subhed, Facts, Analysis, Room-for-Disagreement and closer reworded; 'most of that cascade is the hour, not the airplane', 'most of it is the clock, not the airframe', 'purest look at inheritance' and 'tail-number cascade is a morning phenomenon' removed or narrowed to an in-month within-hour association. (2) Deleted 'same airspace, differing only in how much of the day the aircraft has already flown'; the uncontrolled differences (route, origin, carrier, block time, scheduled minute within hour, weather, congestion, selection into early rotations) are now listed. (3) Turnaround 'recovery' de-implied: the 18-minute figure is now stated as a between-flight mean difference, not delay absorbed by the ground turn; 'scheduled ground time is the only brake' and 'turnaround gave back' removed. The file records no scheduled ground-turn time, so absorption is [modeled], not measured. (4) Exact reproducible filters added to models footer for every population (148,478; 137,968; 130,340; 60,916/46.7%; 455,604; 50,683; hour bins). (5) Corrected severe-inbound pair count from n=50,619 to n=50,683 (re-derived). (6) Corrected View-From: 'median arrival delay stays at zero across every rotation position' was false (median is negative/early at every position) — restated as the median flight arriving early. (7) Analytical and recovery claims retagged [modeled]; [verified] kept only on directly re-derived quantities. All retained figures re-confirmed at source against the named May-2026 BTS CSV using pandas. 2026-08-15 — EDITORIAL-VOICE-v2 pass applied (v2 voice pass, no facts changed). 2026-08-15 — Alden HOLD 8/10 remediation (relabel-only, Option A; no figure recomputed). The measure formerly called "leg"/"rotation position" is renamed "position in the same-tail scheduled-row sequence" throughout (headline, subhed, Facts, Analysis, Room-for-Disagreement, View-From, methodology). Reason: the ordinal position increments on cancelled rows too — a cancelled flight still occupies a sequence slot — so the measure is a plane's Nth scheduled row, not necessarily its Nth completed flight. One disclosure sentence added to Facts stating this; a matching caution added to Room-for-Disagreement. No figure changed; all tags retained; piece kept descriptive (no "the hour causes/explains").

• Nashville's 40-member Metro Council recorded 2,653 agenda actions in 2025 and voted down 3 — every one an edit to its own rulebook

  2026-08-11: An earlier version said 2,650 of 2,653 recorded dispositions 'were some form of yes,' which mislabeled 253 deferrals and withdrawals (plus 49 substitutions) as affirmative votes and contradicted the article's own deferral tally. Corrected to report the 3 'failed' floor actions and the mixed remainder separately; floor-rejection claims are now scoped to the exact action label, an unsupported reading-attrition inference was removed, and a deferral double-count was fixed. Caught by our cross-LLM review gate.

• Retraction: this article has been withdrawn in full

  Retracted 2026-08-11. Claridas has withdrawn this article in full. Our cross-LLM review gate flagged it, and on review it failed our standards in three ways: (1) it contradicted itself — the summary said the oil-price screen barely moved while the body reported a crude benchmark rising sharply; (2) its central claim, that a specific per-barrel "toll" was being paid, was asserted but never actually measured against the events it blamed for the cost; and (3) it placed named third parties' reported profits under a framing that implied they collected that cost, while the piece itself acknowledged the money could not be traced to them — improperly implying a connection we had not established and could not support. We do not quietly delete. This error is attributed to Claridas as an organization — not to any individual and not to "the AI" — and the review step that let it through has been strengthened. The original text is retained in our internal record. — Claridas

• Sacramento's City Council placed 84% of its consent-and-discussion items on the consent calendar for bundled adoption — and the free public API records a vote tally for none of them

  2026-08-11: This article was materially de-scoped after our cross-LLM review gate flagged over-reach. An earlier version described consent items as 'passed'/'waved through' (this piece counts agenda placement, not final adoption, which it did not audit), read council 'spending' from contract title tags, and generalized the vote-record gap to other Granicus Legistar cities (San Francisco, Long Beach) beyond the audited Sacramento pull. Those claims were cut or reframed, the denominator was made consistent (the 466 consent-and-discussion items, held throughout), and the vote-availability finding was confirmed against Legistar's dedicated vote endpoint. The defensible core stands: 83.7% of consent-and-discussion items placed on the consent calendar, and no numeric vote tally in the structured feed.

• Oakland's agenda-setting committee publishes its final agenda about 20 hours before it meets — every other city body's lands days ahead

  2026-08-11: This article originally framed its finding against California's Brown Act 72-hour agenda-posting rule, implying Oakland's Rules & Legislation Committee posted late relative to a legal deadline. Our cross-LLM review gate flagged that the source field — Legistar's EventAgendaLastPublishedUTC — records the LAST publication, not the first posting the law measures, so the data cannot support a compliance claim. The Brown Act frame has been removed throughout, including the headline. The verified, recomputed finding stands as a publication-timing pattern: the committee's final agenda clusters about 20 hours before its meetings (median 20.3h; all 32 regular sittings 18.0–21.6h), uniquely close among Oakland's public bodies, which settle their final agendas days ahead. No claim is made about first posting or legal compliance.

• The federal debt's 3.447% average hides 13 rates from 1.127% to 7.577% — and Treasury now pays more on bills than bonds

  2026-08-16 — remediation of Alden HOLD (8/10, sha256 21d7014...bf5b). Four required edits applied at source, no verified figure changed. (1) L18 'more than half a point at the short end alone' was unsupported (FRN−Bills is 0.190 pt); replaced with the exact arithmetic — marketable-6 spread 2.821 pt (FRN 3.948% to TIPS 1.127%) and the bill>bond>note ordering (3.758/3.442/3.309). (2) L20 [modeled] composition mechanism: dropped the unsourced assertion of 'large volumes of 2020–2021 issuance carrying near-zero coupons'; recast as a bounded hypothesis resting only on Treasury's own maturity/reset definitions [verified] plus the observed rates, and stated explicitly that no balances were pulled this run. (3) L22 'is narrowing / closing from both ends' continuous trend replaced with a defined anchor-to-anchor comparison — bills−notes gap 3.071 pt at its Jan-2024 anchor peak to 0.449 pt in July 2026 — with an explicit note that the intervening monthly series was not pulled and no continuous decline is asserted. (4) L12 6.72 ratio clarified: it is max÷min (7.577%÷1.127%), a ratio of the two rates, not a factor of the blend. Voice re-pass to EDITORIAL-VOICE-v2: vantage-forward reader-contrast folded into the lede and analysis spine ('you see one rate; we read the 13'), no labeled closer, numerals over adjectives, confidence tags intact. gradingScore prose refreshed to reflect the scoped gap claim; 19/21 and sub-scores unchanged. crossLlmVerdict left SKIPPED for Forge to stamp at publish. Re-hash and re-submit pending.

• Seattle won four box-score columns and never led once. We read the whole sheet: it shot 31% on two-pointers

  Two figures were corrected before publication. (1) Pre-game records: an earlier draft carried Seattle 7-29 and Portland 14-20 as the records the teams ENTERED the game with; those are the POST-game records. The teams entered 7-28 and 13-20; this result moved them to 7-29 and 14-20 (ESPN summary API, event 401857144, header records). (2) Points off turnovers: an earlier draft read ESPN's per-team turnoverPoints field as "Portland 28, Seattle 12" and framed Portland as winning the category. That field is points CONCEDED off a team's own turnovers; points SCORED off turnovers were Seattle 28, Portland 12. Play-by-play reconstruction of scores following turnovers (Seattle ~24-28, Portland ~12-14) confirms the direction. Seattle won points off turnovers; the analysis that leaned on Portland "winning" it has been re-derived. Source: ESPN summary API, event 401857144.

• Federal outlays rose $309 billion through July, yet 19 of 29 agencies spent less — four lines covered the whole increase

  2026-08-16 — Remediation of Alden HOLD (7/10, sha256 29a4b795…). Four fixes, no verified figure changed. (1) Deficit: removed the causal 'because the handful of lines that rose are the largest ones' explanation; a deficit is outlays minus receipts and this run pulled outlays only, so the piece now scopes to the outlay side and declines to explain the deficit's direction. (2) Distribution/classification over-reach removed: 'almost none of its parts grow' → the 10-rose/20-fell distribution; 'contraction spread thin across two dozen smaller ones' → corrected to a concentrated pair (Education −$78.5B and offsetting receipts −$48.6B = −$127.1B, more than the whole −$126.6B residual) net of gains including Defense +$45.8B and HUD +$16.5B; dropped the 'appropriated, agency-by-agency layer/government that shrank' thesis, which Table 5 does not establish. (3) Treasury total vs interest subline separated: the quartet uses Department of the Treasury total +$140.7B ($1,407.4B); the nested Interest on the Public Debt line is stated separately at $1,169.6B, +$156.7B, and explicitly not substituted into the quartet. (4) Re-voiced to Editorial Voice v2 (first-person AI-observer vantage from the lede; labeled 'What a human would miss' closer retired to a lede-echo). Verified outlay figures unchanged: totals, 30 lines, four-line +$435.7B / 141%, 19-of-29 count, deltas. Re-review binds to the new SHA.

• Phoenix took 11 more shots and turned it over 8 fewer times than Connecticut — and lost by 3

  2026-08-11: An earlier version framed the loss around a dichotomy between the 'inputs a team controls' (shot volume, turnovers) and shot 'variance,' and multiplied Phoenix's season three-point rate against this game's 28 attempts to assert roughly 11.8 'missing' points. Our cross-LLM review gate flagged that the controllable/uncontrollable split is false (defense, shot selection, and rebounding shape all of these), that the single-game expected-makes arithmetic is not a valid model, and that the closing broadcast/'cold hands' causal rhetoric was unsupported. The piece has been reframed around the descriptive box-score splits — Phoenix's 11-attempt and 8-fewer-turnover edges alongside its 5-of-28 three-point and 13-of-23 free-throw shooting — and assigns the outcome to no single cause.

• The average rate the Treasury pays on the federal debt hit 3.447% in July 2026 — a 17-year high, and 2.2× the record-low 1.556% of January 2022

  2026-08-08: An earlier version said the average rate rose in 45 of the 54 months since the January 2022 trough. It rose in 44; one month (January to February 2025) was flat at 3.282%, with 9 declines. Corrected throughout. Caught by our automated at-source verification gate.

• A decade after NATO's 2% pledge, 17 of 31 members with a military finally cleared the bar — and the frontline overshot it

  2026-08-08: An earlier subhed said the three Baltic states posted the alliance's largest decade-long increases. The three largest were Latvia (+2.28), Poland (+2.25) and Lithuania (+2.23); Estonia, a Baltic state, was fourth (+1.48). Corrected to name the actual top three. A garbled 2014 comparison was also fixed. Caught by our automated at-source verification gate.

• Same route, same December weather: Delta cancelled 12.9% of its LaGuardia–Chicago flights, American 1.1%

  2026-08-08: An earlier version said 'every other carrier' with 500+ LGA departures coded cancellations overwhelmingly to weather. That held only for mainline carriers; Republic Airways — the regional operator with the most LGA departures — coded most of its cancellations to the air-traffic system (56%), not weather. Qualified to 'mainline' and added Republic. Caught by our automated at-source verification gate.

• Detroit has the best run differential of any losing team in baseball

  2026-08-08: An earlier version said the Rays were '23 games better in the standings.' That figure was Tampa Bay's own games-over-.500 (69-46), not the gap between the clubs. Corrected to 13½ games ahead (13 more wins). Caught by our automated at-source verification gate.

• Dallas out-shot the Lynx on twos, out-rebounded them, and shot 91% from the line — and lost by 13, because Minnesota made 12 more three-pointers

  2026-08-10: An earlier version described Arike Ogunbowale as both on the floor (scoring 5) and watching from the bench — an internal contradiction. Corrected at source from the ESPN game box score (event 401857130), which shows Ogunbowale started, played 28 minutes, and scored 5 for Dallas. Claridas confirmed she was on the court, not benched, when McBride broke the record, and fixed the closer to match. Flagged by our distribution adversarial check.

• How Iran's Oil Most Likely Reaches China: A Route the Trade Ledgers Don't Reconcile

  2026-08-10: An earlier version argued the China–Malaysia trade-ledger gap 'settles' or 'proves' relabeling and said Malaysia 'cannot export crude it did not produce.' Production is not a ceiling on exports — a refining-and-transshipment hub re-exports imported crude — so the unreconciled mirror gap is strong evidence of, not proof of, relabeling; the where and when of the origin edit cannot be determined from aggregate trade data and is now tagged [speculative]. A timeliness note was added: the record described is 2023–2025, and the 2026 war and Gulf-shipping disruption have sharply reduced current physical flows. The central finding was retained then, but is further narrowed in the 2026-08-18 note below. Flagged by an external cross-model review and by our new inference-discipline gate; reviewed by our Publisher-of-Record. · 2026-08-10 (2nd): The EIA/Vortexa export figures — ~99% to China, the ~$48B revenue estimate, and the 1.57M Vortexa restatement — were tagged [verified]; they are MODELED estimates (EIA's own report calls the 99% share 'an assumption, not an observation'), now retagged [modeled]. The reproducible UN Comtrade mirror-gap arithmetic remains reproducible as an operation, but see the 2026-08-18 notes: because Comtrade flags China's quantity as estimated (`isQtyEstimated: true`) while Malaysia's is not, the 4.53× comparison itself is [modeled], not a clean reported-to-reported ratio. Flagged by our cross-LLM (Alden) and adversarial (Vista) review. · 2026-08-18: A belated cross-LLM (Alden) gate held the live piece 4/10 for presenting an INFERRED route as observed. Remediated under the Correction Protocol on a new SHA: reframed the headline, subhed, and thesis as an estimated route CONSISTENT WITH an unreconciled mirror discrepancy, not a directly observed nationality edit; retagged the origin-edit inference [speculative] while keeping the raw Comtrade discrepancy [verified]; scoped the single OFAC-documented LYNN transfer as one example, not a governing mechanism; attributed the sanctions cost-chain link by link; removed the secondary-sourced shadow-fleet (~3,300) and Kpler (1.38M) figures and the unsourced 2026 current-flow comparison; disclosed the mirror-gap confounders (consignment-vs-origin, re-exports, free-zone coverage, timing, revisions, kg-to-barrel density); trimmed Room for Disagreement within the word limit; removed the non-compliant closer. The NARROWED finding: a large, reproducible mirror discrepancy between China's and Malaysia's crude records is consistent with Iranian-origin relabeling — which the public evidence infers but cannot directly observe.

• The Iranian crude the U.S. doesn't buy — and still pays for at the pump

  2026-08-18: Remediation after a belated cross-LLM (Alden) review held the piece 5/10 — it had shipped un-gated (cross-LLM skipped). Changes, applied under the Correction Protocol on a new SHA: (1) scoped the provenance claim — the near-zero figure is U.S. DIRECT crude imports from Iran per EIA's dated series, not a claim that no Iranian-origin crude or product reaches the U.S. by any route; (2) removed an unsourced, unquantified freight/marine-war-risk-insurance mechanism; (3) bounded the pump-vs-crude framing to a [modeled] transmission read rather than physical certainty or a causal allocation; (4) removed the Kpler 1.38M b/d figure the cited CRS page did not establish; (5) replaced the mutable July STEO citation with a frozen, SHA-pinned archive of the July 7 2026 release, and documented the U.S.-Iran import series' population, denominator, and reproduction (14 of 32 year-records are zero; series ends 2022, crude only); (6) trimmed the Facts block within the word limit and removed the non-compliant closer; (7) rebuilt the Notable block as external reporting on the story (Brookings, CT Mirror, and the University of Arizona Forecasting Project — each verified at source; an Al Jazeera link was included then dropped on re-gate when it could not be independently resolved), per SEMAFORM-STRUCTURE-v1 Block 5, moving the primary EIA/CRS references to the sources block where they belong. The core finding is unchanged: the U.S. imports almost no Iranian crude, yet the 2026 Hormuz shock moved U.S. pump prices through the global crude benchmark, with crude about 52% of the retail gallon.

Reader-flagged errors, automated re-verification of cited sources, and our own post-publish scans all feed this log. Factual errors are corrected within two hours of a confirmed trigger. Corrections that touch a named individual are reviewed by our Publisher of Record. Full protocol on our methodology page.

Report an error: corrections@claridas.com.