The NexOSRadar Readout

This paper missed a week. No issue went out on October 2, so this one carries two numbers on its masthead and fourteen days in its sections, and four of the six grades in it are a week older than they should be. Nothing happened to the news. The paper was not here.

The grades come first, because they are what a ledger is for. Six calls settled between September 26 and this morning and four of them went against the paper. The Space Force made no award that fits the call. PJM's backstop auction did not open, because its regulator suspended the filing behind it the day before. Anthropic's prospectus reached the public through a leak and not through EDGAR. And the Atlantic produced a hurricane. War Department · FERC order, 9/29 · SEC EDGAR full-text search, read 10/9 · NHC Update, 10/7 On the other side, Congress obliged by not finishing the defence bill, and Brent settled above $100 on September 30 on one contract and below it on another, which this paper grades a hit and prints with the case against it at the same size. States Newsroom, 10/1 · FinanceFeeds, 10/1 The record is 18 hits, 9 misses and 20 on the clock. Two weeks ago it read 16 and 5.

The storm matters more than the grade. Isaias, the hurricane that ended this paper's zero-hurricane call at 10:30 p.m. Central on Wednesday, became a major hurricane at 7:20 this morning, at 120 miles an hour and about 240 miles south of Pensacola, and the Hurricane Center expects landfall tonight or early Saturday inside a warning area that runs from Ocean Springs, Mississippi to the Bay and Gulf county line in Florida. NHC Update, 10/9 · NHC Advisory 11A, 10/9 That is after this paper goes to press, and nothing in this issue reports it as done. Two weeks ago this paper marked that call "strengthened again". Issue 10 The Front has what it should have printed.

Around the storm, a fortnight of things that did not close. Washington and Tehran passed seven-day plans through Qatar and produced no ceasefire, and attacks on tankers in the Strait of Hormuz hit their highest weekly count of the war. Jerusalem Post, 9/30 · Baird Maritime, 10/8 September payrolls printed 29,000, and the ten-year Treasury closed at 5.31 percent the next session anyway. BLS, 10/2 · Treasury, October Ebola reached Nairobi on an international flight, with WHO's confirmed count in DR Congo at 8,728. WHO DON619, 10/8 Magdalena Andersson handed her mandate to form a Swedish government back to the Speaker and four days later was handed it again. Riksdagen, 10/2 And a draft prospectus put Anthropic's 2025 net loss at $42 billion and its infrastructure commitments at $518 billion, on Reuters's reading of a document the public still cannot open. Reuters via MarketScreener, 9/28

This is also the issue that prints the season scorecard, as the first issue said it would. Issue 1 Twenty-seven settled calls score 0.215 against a coin's 0.250. The calls this paper logged at sixty won eleven times in twelve, the calls it logged at a coin flip lost two times in three, and the audit found a call from July that was graded on the convenient half of its own sentence. It is all below, same font. So is the largest correction of the season: for two issues this paper told its readers that nobody outside OpenAI had compiled a proof that outsiders compiled, in public, on the day it was released. Lean Zulip, 9/8

Week Zero (logged 7/17, graded in Issue 1 with the answer sheet out): 2 hits, 1 miss. Exhibition. Doesn't count.

What moved this fortnight:

Calls #44 through #47 carried their seven days, and seven more that nobody asked them to, and enter the full table. Six new calls open below. The gridiron line, kept outside this record, went one for three and is tallied where it lives. The season scorecard this paper has promised since its first issue follows the weekly one.

Six grades, two for the paper and four against, and four of the six are a week late because this paper was not here on October 2 to print them.

Call #2, HIT. The call said no NDAA conference report would be filed and no bill signed before October 1, logged at 60 percent on July 17. None was. The Senate's bill has not moved since a motion to proceed on July 27, after cloture failed 50 to 46 on July 14; the House's bill, passed 216 to 212 on July 22, has sat in the Senate since September 14; and on the settle date itself a Capitol reporter wrote that "Lawmakers haven't brokered a compromise on the massive annual defense policy bill." govinfo, S. 4784 · govinfo, H.R. 8800 · States Newsroom, 10/1 This is the second row on the ledger and the one whose kill this paper had to rewrite in public in its first issue. Seventy-six days later it settled on the rewritten kill without needing a ruling, which was the point of rewriting it. A 60 percent bet that Congress would not finish something is not a hit to frame.

Call #35, HIT on the letter, argued, and the argument against is printed at the same size. The call said: "Brent crude settles above $100 per barrel on September 30, 2026." Logged at 60 percent on September 11. On September 30 there were two Brent settlements, because it was the last trading day of ICE's November contract. November settled at $103.50. December, which becomes the front month the next morning, settled at $98.03. EnergyNow, 10/1 · FinanceFeeds, 10/1 The physical barrel was at neither: the EIA's spot series printed $115.91. EIA, 10/7 And two prediction markets on the same question resolved in opposite directions: Kalshi NO, on an index print of 97.92 at 5 p.m. Eastern, and a Manifold market YES, under a rule that reads "Use the contract treated by ICE as front-month on that date." Kalshi · Manifold, 10/1

The case for the grade. "Settles" is a futures word, and on September 30 the nearest unexpired contract, the one the exchange's own convention treats as the front month through its last day, was November. It is also the contract this paper followed in print: the $106.60 it reported for September 24 was a November settle. And the call was logged against the EIA's forecast of the spot price, which finished the day fifteen dollars over the line.

The case against, which a reader is entitled to prefer. On September 25 the Associated Press's market wrap had Brent "below $98" on a day the November contract settled above $104, which means the wire was already quoting December. AP via La Nación, 9/25 The larger of the two markets, the one this paper itself held up in Issues 9 and 10 as the market's price on this very call, resolved NO. One trade wrap named December the front month that afternoon. BigGo Finance, 9/30 And a call that needs its author to choose the contract after the fact is a defective call, written by this paper, about a date this paper should have known was an expiry day. In Issue 3 this paper said most publications round in their own favour and call it editing. This grade rounds in the paper's favour. It is printed as a hit because the letter, read with the exchange's convention and the paper's own usage, says hit, and it is printed with the knife beside it because that is the only honest way to print it. Counted as a miss, the season's record is 17 and 10 and its Brier score moves from 0.215 to 0.222; both versions are in the Season Scorecard below.

Call #15, MISS. The call said the Space Force would award at least one more space-based sensing or missile-tracking contract over $500 million before September 30, logged at 65 percent on August 7. For two issues this paper could not read the Pentagon's daily contract announcements and said so. This fortnight all 38 were read, August 7 through September 30, and none carries such an award. War Department One entry is the argument for a hit and is printed so a reader can make it: on September 28, Space Systems Command awarded Amentum $572,358,987 for the "operation, maintenance, and sustainment" of the warning and attack-assessment systems at Cheyenne Mountain. War Department, 9/28 It is a Space Force contract, new, over the bar and inside the window, on a missile-warning system. It also buys no satellite, no payload and no sensor, the work is in a mountain in Colorado, and the words "space-based", "sensing" and "tracking" appear nowhere in it. The call said space-based. MISS, and the dry part is the pedigree: call #8, its parent, cleared in one afternoon in August, and this paper logged the sequel at 65 on the theory that money which moved once would move again on schedule.

Call #23, MISS, on one word of its own kill. The call said PJM's backstop reliability auction would clear at or above $400 per megawatt-day, logged at 55 percent on August 21; the kill included "the auction is delayed, cancelled, or restructured before results post." On September 29 FERC accepted PJM's filing, suspended it for five months and set its cost allocation, exit rules and collateral for hearing. FERC order, 9/29 The next day PJM wrote: "PJM will not proceed with the Reliability Backstop Procurement on Sept. 30 as originally planned." PJM Inside Lines, 9/30 Delayed. The price leg was never tested, and this paper does not get to claim it would have been right about a number no auction produced.

Call #25, MISS. The call said Anthropic would publicly file IPO paperwork by September 30, logged at 45 percent on August 21. It did not. For two issues this paper could not read the SEC's filing system and said so. On October 9 it read EDGAR's full-text search directly: every S-1 and S-1/A filed from September 1 to October 9 that contains the word "Anthropic" belongs to another company. SEC EDGAR full-text search, read 10/9 A draft leaked to two news organisations on September 28, which is in the Stack and is not a filing. The grade is clean. What is not clean is the label this paper kept on the row: "tracking YES" for four issues, "weakened" three times, and still YES five days out, on a call whose reported target had already moved to November. A label that can be weakened three times without changing was reporting the paper's attachment, not the call's odds.

Call #37, MISS. The call said the Atlantic season would end with zero hurricanes, logged at 45 percent on September 11; the kill was any Atlantic cyclone reaching hurricane strength by November 30. At 10:30 p.m. Central on October 7 the National Hurricane Center issued an update headed "ISAIAS BECOMES A HURRICANE". NHC Update, 10/7 Day 27 of 80. Eleven days earlier this paper had marked the row "strengthened again". The storm is in the Front, where it belongs, because it is forecast to reach the northern Gulf Coast tonight and the grade is the least important thing about it.

Tracking updates, nothing else settled. #5 (calibration headline): four frontier releases, none headlined calibration, and a factual-error rate is the closest any vendor came; the Threshold. #6 (control-plane cloud incident): a candidate is named for the first time, Azure incident 7Q30-010 of September 30, and the ruling waits for Microsoft's own review, due about October 14; the Stack. #7 (advanced packaging): tracking YES, TSMC reports October 15; the Stack. #9 (Gulf war-risk clock): not started, and the weekly transit count has crossed the call's own kill line with no ceasefire to count from; the Wire. #17 (PJM capacity price): $325 stands, results December 22. #26 (Venezuela): no date, 52 days left; the Wire. #29 (Sudan): El-Obeid struck again and not taken, 53 days left; the Wire. #30 (grid equipment rule): comments close today, leaning YES on nothing new; the Grid. #31 (Nevada): no calendar; the Grid. #32 (measles): CDC at 3,887, Pennsylvania past 1,000, no commission date; the Carry. #33 (Guinea-Bissau): December 6 stands, 58 days left; the Wire. #36 (Nvidia and Hugging Face): no Second Request found; the Stack. #39 (Fed): weakened, because the range is 3.75 to 4.00 and the call now needs December; the Tape. #40 (Swedish PM): a first vote is possible from October 14; the Border. #42 (Colorado River suit): no second suit; the Grid. #43 (Ceuta): no investigado in the docket the call names; the Border. And the four that carried from Issue 10 and enter the table: #44 (Australia and OpenAI), an apology and a review and no referral; #45 (Ratepayer Protection Act), cloture failed 57 to 43 and the Senate has left until after the deadline; #46 (Ebola), 8,728 against a line of 10,000 with WHO counting 487 a week; #47 (Lee Ferry), tracking YES with a problem underneath it that is this paper's own and is in the Grid.

Three of the four misses were the same bet in different uniforms: that an institution would do a scheduled thing on schedule. An award, an auction, a filing. This paper spent the summer writing that institutions arrive late to their own deadlines, and then priced three of them to arrive on time. The fourth miss was a bet against an ocean.

Six grades, two for this paper and four against, from twenty-seven calls settled to date. The record is 18 and 9. Last issue it read 16 and 5. It is the worst fortnight this ledger has had, and it lands in the issue that prints the season.

The first issue of this paper said: "Issue 12 prints the scorecard, hits and misses, same font." Issue 1 Every issue since has repeated it at the foot of the page. This is it. Its six parts and their order were fixed in writing on September 23, before eight of the calls it now counts had been graded, so that the paper could not choose its measures after seeing its results. It covers only what was printed. Every number below can be recomputed from the table at the bottom of this issue, and a reader who gets a different answer is invited to say so.

1. The record. Forty-seven calls logged across ten issues. Eighteen hits, nine misses, twenty on the clock. None struck, none unresolved. Of the twenty-seven settled, two in three went the paper's way. Week Zero, the three calls graded in Issue 1 with the answer sheet already out, stays excluded: two hits and a miss that do not count. The gridiron line, kept outside the record by design, is five and four.

The shape matters more than the total. At the end of August this ledger stood at ten and two. Since then it is eight and seven.

2. The Brier score. For each settled call, take the gap between the odds this paper logged and what happened (one for a hit, zero for a miss), square it, and average. Lower is better. A coin, saying 50 percent every time, scores 0.250.

The paper beat the coin. It barely beat the third row, which is the score of a forecaster who knows nothing about any single call except that this paper is right two times in three. Put plainly: knowing the paper's overall record was worth almost as much as reading the odds it attached to each call. The odds carried a little information. They did not carry much.

The second column the plan called for, market odds, exists only where the issue that logged a call also printed a market number for it. That happened once. Call #34 was logged at 60 percent, in its own words "below the 64 to 70 percent the market priced on September 10", and the Fed raised twelve to nothing. Issue 8 Paper: 0.160. Market: between 0.090 and 0.130. The market column has one row and the market won it. Three other market numbers sit in this paper's back issues and stay out, each for a stated reason. The same issue printed a Kalshi price of 51 to 52 cents on that hike, dated a week before the call; the call named the fresher number as its benchmark, and against the stale one the paper would have won. Issue 7 printed a market that had a September cut under 1 percent, a week after this paper logged call #28 at 50. Issue 7 Issue 9 printed call #35's contract at 49 cents, a week after this paper logged it at 60. Issue 9 The plan says no market numbers are added after the fact, so none are. A reader who adds them will find the market far ahead on #28 and, on the grade as it stands, behind on #35.

Five rulings bear on these numbers: one made last issue over this paper's objection, and four made at this issue's edit. Each is shown so that no ruling can hide in the average:

Under every reading on that table the paper beats a coin, and under none of them does it beat one by much.

3. Calibration. If the odds mean anything, calls logged at 60 should hit about six times in ten.

Every row is small. The highest bin is four calls, too few to read a rate from, and it prints because the plan says every row prints. The lowest is about where it should be. The middle is upside down. Calls this paper logged at a coin flip lost two times in three. Calls it logged at sixty won eleven times in twelve. The plain reading is that this paper has two prices, not four: sixty for things it believes, and fifty-five for things it wanted to say and could not defend. The ledger's advice is to trust the first kind more than its number admits and to stop writing the second kind.

One more cut, because it flatters the record and should be seen doing so. Ten of the eighteen hits were calls that needed one event inside a window and got it early: one award, one launch, one subpoena, one meeting. Those calls can be graded the day the event happens and cannot miss until the window closes. The fourteen calls that had to run to their date went eight and six.

4. Resolution discipline. Fourteen of the twenty-seven settled on the date written. Ten settled early because the single event the call required happened, under the rule printed in Issue 5. Issue 5 Three settled early because the kill fired: #27 when Congress funded the government four weeks ahead of its deadline, and #23 and #37 this fortnight. Five grades were late: #3 by a week in August, and #2, #15, #25 and #35 this issue, a week after the issue that should have carried them.

Clocks and terms changed after logging, each named. Call #2's kill was rewritten in the issue that first printed it, from "a clean floor vote inside three weeks" to a conference report or a signed bill, with the ruling in print. Issue 1 Call #17's "next posted auction result" was ruled in Issue 5 to mean the next regular auction and dated December 22 in Issue 9. Those two were done in daylight.

The third was not. Call #10 said Hormuz transits would "stay below half pre-war count through August," and its kill, as logged, was "two consecutive weeks above 80 transits." Issue 1 From Issue 3 on, this paper measured the call against 250 a week, half of a pre-war 500, and in Issue 7 graded it a hit on that line without mentioning the kill. Issue 3 · Issue 7 On the Lloyd's List counts this paper itself printed, the weeks of July 27 and August 3 came in at 95 and 91 after revision, and the last two weeks of August at about 98 and 84. Issue 4 · Issue 5 · Issue 7 The kill as logged fired, twice. The call's own sentence held on every baseline anyone has published. The two never agreed: the same issue that wrote "above 80" also put the traffic at 15 percent of pre-war, which makes half of pre-war about 175. So the kill was wrong the day it was written, the paper graded the sentence that suited it, and for ten issues nobody at this paper, including the grader, noticed. The row stays a hit, by a ruling at this issue's edit that the call's own sentence held; the table above shows it as a miss, and of everything this audit found it is the one that most resembles the thing this paper says it does not do.

Calls logged on top of facts older than the call: six of forty-seven. Call #3 was a bet on a record that had already fallen, on July 2, fifteen days before the call was logged; it was graded a hit in Issue 3 with an apology for lateness and no mention of the order of events; at this issue's edit it was left standing, on the letter. Issue 3 An award three days before call #15 was ruled not to count. An auction three weeks before call #17 was ruled not to count. Call #41 was logged the day its event happened and ruled a hit on the letter. Call #43 sits on a reported court step sixteen days older than it is, and that is not ruled. Call #47 sits on a federal draft three months older, which was ruled at this issue's edit not to settle it. Counted twice, excluded three times, pending once. A rule that has produced every available answer is not yet a rule.

Unresolved: none. Struck: none. One call may never resolve: #9 waits on a ceasefire that has not existed in eleven issues, and it has no expiry.

5. Corrections. Through Issue 10 this paper printed 47 entries under Corrections: 14 corrections of fact, 4 items resolved, 26 gaps or disagreements named, 2 figures suspended or downgraded, and one standing source changed. By issue: one each in the first six, then 2, 7, 14 and 18. This issue adds 27, the most yet, and among them the largest of the season: for two issues this paper told its readers that nobody outside OpenAI had compiled a proof that two people had compiled, and had announced compiling, on the day it was released. It is in the Threshold. One entry this issue corrects a correction: Issue 10 "fixed" an Italian date that Issue 9 had right.

Process failures, named as such. Issue 1 missed its own ship date. Issue 3 miscounted its own hits and graded a call a week late. Issues 9 and 10 each lost a first research pass to a search limit. Issue 10 also repeated a claim that this paper's own files had marked as wrong two days earlier. This issue: the paper missed a week outright, and its research ran into the same limit a third time. The entries rise with the size of the paper and with the reach of the checking, and this paper cannot tell from here which of the two is doing more of the work.

6. What changes in season two. Proposed, applied forward only, never to a call already on the table:

• Name the instrument. Every call names its truth source, and a price call names the contract month, the settlement source and the hour. Call #35 would have had one answer.
• The kill is the call, negated. Checked at logging by a second read. If the kill and the call can come apart, the call is not logged. Call #10.
• One rule for old facts. The check before logging runs to the start of the process the call is about, not fourteen days, and a call whose answer was already in print when it was logged is struck, not graded. Calls #3, #41 and #47.
• Conditional calls expire. Call #9.
• A number in place of a mood. "Tracking YES, weakened a third time" becomes the paper's current odds, printed beside the logged odds each issue. The grade still runs on the logged number.
• The market beside the call. Wherever a market trades the same question, its price is printed at logging, so the second Brier column has more than one row.
• No unexplained coin flips. A call logged at 50 to 55 carries the sentence that says why it is not a pass.
• The ledger ships when the paper does not. If an issue cannot be written, the table and the grades print alone that Friday. A grade should never again run a week late because the essays were.

This is a grade, not a verdict. Eighteen and nine, 0.215 against a coin's 0.250, a sixty that should have been bolder and a fifty-five that should not have been written, and one call in forty-seven graded on the convenient half of its own sentence. The paper said in its first issue that it keeps receipts. These are the receipts.

Washington and Tehran opened the fortnight passing seven-day plans through Qatar and ended it with no ceasefire, Iran's delegation ordered out of New York, and the heaviest week of tanker attacks since the war began.

The oddity is in the count. Lloyd's List Intelligence, silent since August 27, resumed its Hormuz briefs on September 30. Its October 8 brief reads "Total transits: 124 vs 122" for the weeks of September 21 to 27 and 14 to 20, the earlier week revised up from 117, and calls September's total of at least 477 "the highest number tracked since the outbreak of war at the end of February." Lloyd's List Intelligence, 9/30 · Lloyd's List Intelligence, 10/8 The kill on call #9 reads "transits back above 120/week inside a month of any ceasefire." Two straight weeks are above 120, and no ceasefire exists for them to be inside a month of. On the letter that settles nothing: the kill is conditional, the condition has not occurred, and a week above the line in wartime neither kills the call nor starts its clock.

What it does mean is that the line has stopped measuring a recovery. This paper wrote it on July 24, in the weekly series it was then using, as the mark traffic would not regain inside a month of a peace. Traffic regained it in the middle of the war, at about a quarter of the roughly 500 a week this paper took as pre-war when it graded call #10. If a ceasefire lands while that series holds above 120, the kill fires on the letter in the first week counted, even with war-risk cover still at thirty times peacetime. The line stays as written, because a kill moved after the number crosses it is not a kill. Which count gets read against it, Lloyd's List's weekly total or the Kpler average this paper adopted in Issue 9, has never been ruled, and that ruling is owed before a ceasefire, not after one. The thing to watch is whether the number holds: 124 describes the week before the surge in attacks, Kpler's daily count has since touched seven, and the next Lloyd's List brief is due October 22. Baird Maritime, 10/8

The insurance number finally moved, downward, on an estimate that predates the worst week for tankers. Straits.live puts Hormuz war-risk cover at thirty times pre-crisis rates, against the roughly forty printed last issue, with one VLCC transit still about $10 million; the site calls the multiple its own estimate, "set 15d ago," citing among its sources a September 25 Reuters report this paper could not find. Straits.live, 10/9 A multiple that fell by a quarter beside a dollar figure that did not move is an estimate being revised, not a market being reported, and it predates the surge that began September 28. Thirty times peacetime is not a recovery, which is all call #9 asks of this number for now.

Hormuz counts: the standing metric could not be read, and the source that came back counts something else. No Kpler ten-day average dated in the fortnight was found; the last this paper printed is about fifteen, as of September 22. What Kpler supplied, through one Reuters article, is daily: "Seven commodity vessels passed through the strait on Tuesday, for the lowest figure since July 23," ten on Wednesday, October 7, and more than twenty on each of the Sunday and Monday before, against about 125 a day pre-war. Baird Maritime, 10/8 Windward, the secondary, counted eighteen crossings on October 7, twelve in and six out. Windward, 10/8 IMF PortWatch, as relayed, has four vessels on Sunday, October 4, a day Kpler puts above twenty; not reconciled. Straits.live, 10/9 Alongside, and labelled as a different count: Lloyd's List's 124 and 122 are weekly totals of non-Iranian-linked cargo vessels over 10,000 dwt. Lloyd's List Intelligence, 10/8 Issue 9 promised a dual print if Lloyd's List resumed, and this is it. The standing count stays Kpler's ten-day average, which this paper could not print.

Sanctions: the China test came back split, companies yes, banks no. In six OFAC actions from September 29 to October 8, a Chinese national in Foshan was designated on the 29th, five days after the summit, three mainland companies in heavy equipment and steel on October 1, and two shipping companies with mainland addresses on October 8, beside Hong Kong firms in all three. OFAC, 9/29 · OFAC, 10/1 · OFAC, 10/8 No bank appears among the additions in any of the six. OFAC Recent Actions The nearest approach is five Hong Kong companies designated for receiving shadow-banking transfers, in a release that names no Chinese bank, and an October 5 alert known here by title only, "Notice to Foreign Financial Institutions Conducting Business with Iran." Treasury, 10/1 · OFAC, 10/5 Through nine weeks the carve-out meant no Chinese bank and no Chinese entity. The entity half lasted five days past the summit, which answers Issue 8's question, principle or scheduling, for companies: scheduling. The bank half is in its eleventh week and has no answer yet.

Russia: the sanctions law is three weeks old and this paper could find nothing done under it. No presidential action naming Russia, its oil buyers or tariffs appears on the White House list through October 7, read by title only; one list's silence is not proof. White House OFAC's six actions add no Russia-programme designations; the Russian touch of note is the A7 Network, designated October 1, which Treasury calls "a shadow banking network with ties to Russia." Treasury, 10/1 This paper could not verify anything on the Ukraine war or peace talks this fortnight.

The front that did move was Saudi Arabia's, and the last American forces left Iraq. By one tracker's account, Yemeni government forces "backed by Saudi air power" opened an offensive toward Mokha on October 4, with more than 100 Saudi aircraft in the first assault; the Houthis "significantly expanded the scale of their attacks on Saudi military and civilian infrastructure on October 5," fired two ballistic missiles at Riyadh's airport, one of which "hit the airport and sparked a fire," and urged Saudi oil workers to leave oil sites. Critical Threats, 10/5 · Critical Threats, 10/6 · Critical Threats, 10/8 The Hormuz bypass sits inside that threat: the Saudi energy minister said on October 6 the East-West pipeline was "back up to 5.8 million barrels" a day. Al Jazeera, 10/6 From the same tracker: "The United States also fully withdrew its forces from Iraq on September 30." Critical Threats, 10/1

Venezuela: no date, and the only calendar on offer runs past the call. Asked for an election date after the second round of talks with the opposition, National Assembly president Jorge Rodríguez said, "Obviamente no te voy a decir eso." NTN24, 9/26 As of October 1, by one small outlet, "todavía no existe una fecha definida para las próximas elecciones presidenciales," and it has the opposition's Dinorah Figuera targeting November for new Supreme Tribunal magistrates and pointing to December for renewing the electoral council. Radio America VE, 10/1 No gazette and no council rectors appear in anything found. Call #26 is OPEN, tracking NO, fifty-two days left: it needs a date set and gazetted by November 30, and the council's renewal is pencilled for the month after.

Sudan: El-Obeid was hit again and not taken, and a second source now disputes the date this paper corrected last issue. UNICEF says a drone strike on a displacement shelter in the city late on September 30 reportedly killed one boy and injured 12 children. UNICEF, 10/1 AFP has two strikes a day apart, a university dormitory with 5 killed and 47 wounded and a shelter with 5 killed and 13 injured; neither page establishes whether the shelters are the same, so both counts are printed. Arab News, 10/2 No ground breach or seizure was found, and nothing dated October 3 to 9. Call #29 is OPEN, tracking NO, fifty-three days left. On October 8 the Security Council voted unanimously to extend the Darfur sanctions two months, "until December 11, 2026," and the Panel of Experts "until January 11, 2027." Darfur24, 10/9 · The National, 10/8 What that replaced is disputed. Issue 10 corrected the prior expiry from October 9 to October 11 on the UN's press release; Security Council Report says Resolution 2828 "extended the sanctions regime until 9 October without changing the mandate." Security Council Report, 10/7 The UN page would not open on press day, so both dates are printed, unreconciled.

Guinea-Bissau: December 6 stands, with doubt now on the record. The only dated source found, one analysis piece, treats "l'élection présidentielle du 6 décembre" as the standing date and adds that "certains observateurs sont dubitatifs sur la crédibilité de cette date fatidique." allAfrica, 10/1 No postponement was announced in anything found, and no electoral-commission calendar, candidate list or ECOWAS statement either: an absence plus one article, not a confirmation. Call #33 is OPEN, tracking YES, fifty-eight days left.

Congress: the Senate left on September 30 with the defence bill untouched and the data-centre bill three votes short. No NDAA conference report was filed and no bill was signed before October 1: the Senate bill's last action is a motion to proceed on July 27, after cloture failed 50 to 46 on July 14, and the House bill, passed 216 to 212, has sat in the Senate since September 14. govinfo, 10/6 · govinfo, 10/3 Call #2 is graded HIT in the Scorecard. The Ratepayer Protection Act got one vote: cloture on the motion to proceed failed 57 to 43 on Wednesday, September 30, with sixty needed. US Senate, 9/30 Four Democrats crossed, Hassan, Klobuchar, Ossoff and Warnock, no Republican voted no, and Senator Schumer called the bill "toothless." Roll Call, 9/30 That week the Senate also passed the college-sports bill 77 to 22 and failed cloture on the stock-trading bill 53 to 47. US Senate, 9/30 It now "stands adjourned for Pro Forma sessions only with no business conducted," twelve of them from October 1 to November 5, and returns at 3:00 p.m. on Monday, November 9, six days after the call's deadline. Senate Democrats, 9/30 Call #45 enters the table OPEN, tracking NO hard, and is graded November 3. Its kill named three ways to fail, and the Senate has used two: cloture failed, and the chamber left.

Sweden's Left Party helped re-elect a Moderate Speaker, Magdalena Andersson handed back her mandate over it the same day, and four days later the Speaker handed it to her again.

The Speaker's verb is the thing to hold on to: a government the Riksdag can tolerate. A Swedish prime minister is elected unless a majority of the chamber votes against, so Andersson does not need the Centre and the Left to vote for her; she needs neither to vote against. Riksdagen, 10/2 On the 176 to 173 this paper printed, the other side cannot stop her alone, which hands each of those two parties its own veto. The push for an early vote came from Åkesson and the Christian Democrats' Ebba Busch, whose parties are not among her four, and this paper reads it as wanting the failure on the record. TV4, 9/30

So the vote pencilled in for Wednesday can fail without touching call #40. The call dies if no prime minister is elected by November 30 or if a snap election is triggered first, and the trigger is specific: "Vid fyra misslyckade statsministeromröstningar ska extraval utlysas." EFN, 10/7 One failure is a quarter of the way there. The Centre leader, holder of one veto, said on October 7 that "Ett extraval skulle ju vara ett stort misslyckande för alla partier." SVT, 10/7 That sentence and the two parties' positions cannot all survive four votes, and the call is a 60 percent bet that something gives before the fourth and before November 30. The precedent runs against the call, since 2018 took 134 days; the Speaker is reported to want a shorter process this time. Placera, 10/7 The date to watch is Monday October 12, when Andersson reports and may ask for more time; the 28-day clock Issue 10 printed, expiring around October 16, went back with the mandate. EFN, 10/2 SVT's live page had the four party leaders meeting for the first time on Friday October 9, press day, in most entries and on October 7 in one; it stopped updating on the evening of October 7 and this paper could not reach an outcome for either. SVT, 10/7 Call #40 is OPEN: Speaker elected, first prime-minister vote possible from October 14. Fifty-two days.

The Ceuta court heard its witnesses and opened a file on the former Government Delegate, and call #43 still has no investigado in the docket it names. Ceuta's president, Juan Jesús Vivas, testified before Judge María Tardón on Wednesday September 30, the date Issue 9 printed and Issue 10 carried as disputed; resolved, in Corrections. ESdiario, 9/30 · ESdiario, 10/1 His account, as reported: the central government "no parecía consciente de su magnitud," and his side tried Moncloa about ten times in the days before the entry and got through three; one outlet has Vivas placing the calls and another his chief of staff. The judge has since asked the city, "a la mayor urgencia," for its record of calls to Pedro Sánchez or his chief of staff from July 27 to 30. Artículo 14, 10/6 · El Faro de Ceuta, 10/6 The new object is an auto dated September 28 and reported by Europa Press on October 6, in which Tardón, finding that the facts "hacen presumir la posible existencia de una infracción penal," agreed to "abrir diligencias previas sobre el exdelegado del Gobierno" Miguel Ángel Pérez Triano and asked the Fiscalía for a report, on a querella from Abogados Cristianos alleging desórdenes públicos and prevaricación. The same accounts have Gonzalo Sanz, the Delegation's former chief of staff, testifying that he passed to the Delegate what CNI agents had told him; no page opened confirms that Local Police chief Sebastián Vega testified on October 1 as scheduled. Diario de León, 10/6 · ESdiario, 10/6 Those two accounts share their details and count as one source; a third, as read, has the judge asking the Fiscalía about his possible imputation. Artículo 14, 10/6 He is "exdelegado" in those accounts; ESdiario reported on the morning of the auto that he had told the Government he intended to leave the post, and the act ending his tenure was not found. ESdiario, 9/28 Call #43 asks for a person formally named investigado, or a charging decision, in Diligencias Previas 64/2026, and this is neither. Witnesses are witnesses, and the auto opens a file and asks the prosecutor a question. No account opened calls Pérez Triano investigado or imputado, gives the new diligencias a docket number, says they sit inside 64/2026, or prints the number 64/2026 at all. On the reasoning Issue 10 set out, a designation in a separate file is not in the docket the call names, so where Pérez Triano is concerned the call turns on two things this paper could not find: the Fiscalía's report, and the number on the file. Nothing opened this fortnight corroborates or contradicts the September 2 imputación printed last issue, and the Europa Press account of the new auto does not mention it. Call #43: OPEN, no designation in the docket the call names.

Spain and Italy: two dates after all, and the correction this paper printed last issue was the error. The European Commission's notification table lists Italy's controls on "all internal air and sea borders with Spain" as running 01/10/2026 to 16/10/2026, and Spain's on its borders with Italy as 23/09/2026 to 07/10/2026, then 08/10/2026 to 22/10/2026. European Commission Issue 9 printed Italy's checks as running to roughly October 1 and Spain's to October 8. Issue 10 corrected the Italian date to October 8 and said no source gave Italy a distinct one. The Commission's own table does: Italy's renewal starts on October 1, a week before Spain's. Issue 9 was right, the correction was wrong, and it goes back in Corrections with Spain's end date trimmed by a day. Behind the rows: Italy's Interior Minister Matteo Piantedosi announced a further 15 days on Wednesday September 30, and Spain's Orden INT/1033/2026, signed by Fernando Grande-Marlaska on October 1 (the only act of the minister's this paper opened this fortnight) and published in the BOE on October 6, runs to 24:00 on October 22; this paper read the order through ESdiario and did not open the BOE text. Euronews España, 9/30 · Il Post, 9/30 · ESdiario, 10/6 Issue 10 printed one more line the table overtakes: that no third Schengen state had announced controls citing Ceuta. Austria's entry for the period that began September 16 gives "recent incidents in Ceuta" among its reasons, and Sweden's for the period starting November 12 gives "recent events in the Spanish City of Ceuta." Neither borders Spain, and the table does not show when either was notified. France's rows, including a new period from November 1 that covers the Spanish border, cite jihadist threats and not Ceuta, while Il Post, as read, says France and Germany have invoked it; the two do not agree and both are printed. European Commission · Il Post, 9/30 Ceuta is now among the stated reasons for checks on Austria's border with Slovakia.

The death toll figure this paper suspended and then downgraded has the NGO's own page under it, and the page was there before this paper first said it was not. Caminando Fronteras, under the title "Dolor en la frontera de Ceuta": "Las cifras de víctimas confirmadas a fecha 11 de agosto de 2026 son devastadoras: 145 personas han perdido la vida." Caminando Fronteras, 8/3 The page adds that "decenas de personas más siguen desaparecidas," gives no breakdown by side of the border and nothing later than August 11. Its metadata shows publication on August 3 and a last modification on September 11, a week before Issue 9 suspended the figure for want of it and two weeks before Issue 10 ran it as attributed and untraced. The failure was this paper's search, not the NGO's silence. Corrections. The figure is printed for what it is: the NGO's confirmed count as of August 11, not a current toll. Beside it, the institutional counts, none updated inside this window: Interior's 72 on August 3, rising in official figures to 97 by August 20; the Ceuta city government's 88 on August 3; and 83 in Judge Tardón's order. El Español, 9/19 · Moncloa, 9/17 They are not all counting the same water. El Salto's account of the NGO's earlier provisional 141 split it into 78 deaths "reconocidas oficialmente por el Gobierno español" and 63 bodies recovered in Moroccan waters. El Salto, 8/5 A count that crosses the border and a count that stops at it answer different questions, and the page that carries the 145 does not say how it splits.

Ceuta's official number is now the count of people leaving, and it passed 11,000. Minister Ángel Víctor Torres, who holds the single command for the crisis, put total returns and expulsions at 11,026 on October 8, against 2,700 returns at the end of August, with 1,813 between one Sunday and Monday; 513 of the total, as this paper reads his sentence, are expulsions under court rulings. El Faro de Ceuta, 10/8 On September 22 the same minister had 6,791 voluntary returns and 318 expulsions. Maldita, 9/18 Registered minors went the other way, from 2,321 to 2,827, with 24 files sent to Morocco and 86 minors moved to the peninsula to rejoin family. El Faro de Ceuta, 10/8 What the account as read does not contain is the number Issue 10 printed beside it, the Prime Minister's 10,000 remaining: no count of migrants still in Ceuta dated inside this fortnight was found, and the Popular Party's charge that the official balance is misleading is known to this paper by headline only. El Faro de Ceuta, 10/8 A count of exits with no count of who remains is half a balance.

Morocco has a head of government, the first woman to hold the post, and ten days on no cabinet this paper could find. King Mohammed VI appointed PAM's Fatima Ezzahra El Mansouri on Tuesday September 29 and charged her with forming the executive; she succeeds Aziz Akhannouch, and with 97 seats, no majority and the PJD already ruling itself out, she has to build a coalition. APA News, 9/29 By headline only, on Hespress: the USFP leans toward joining and has referred the decision to its national council, the RNI will not press for a place but leaves the door open, and the King opens the legislative year on press day. Hespress, 10/7 · Hespress, 10/7 · Hespress, 10/8 Issue 10 said its reading of post-election volume on the Ceuta and Melilla claim started then. This paper could not take the reading: no statement by El Mansouri or her party on the two cities was found, nor any account of the meeting Albares planned with Bourita in New York, because the searches that would have found them were never run.

Bosnia voted on October 4: incumbents ahead in two presidency races on a partial count, turnout down ten points, and observers who found the competition real and the franchise still restricted by ethnicity. With about 80 percent of polling stations counted, the election commission's figures, carried from Anadolu by one outlet, had Denis Bećirović ahead of Bakir Izetbegović for the Bosniak seat on the presidency, 164,369 to 148,036; Željka Cvijanović ahead for the Serb seat, 213,812 to 102,877; HDZ BiH's Darijana Filipović leading for the Croat seat with 92,854; and SNSD's Savo Minić ahead of Branko Blanuša for the Republika Srpska presidency, 202,245 to 160,936. Turnout was 41.48 percent against 51.45 in 2022. Yeni Şafak, 10/4 That is one source and a partial count; certified results were not obtained. The observers' statement the next day: "There was genuine competition," and also "Suffrage rights continue to be subject to discriminatory restrictions based on ethnicity and residence," "serious shortcomings in ensuring secrecy of the vote," and governments in office handing out one-time payments "to entice voters." OSCE/ODIHR, 10/5 The commission's president, on the first countrywide use of ballot scanners: "the scanners did the job for which they were intended." Yeni Şafak, 10/4 They did, for the 41 percent who came.

Moldova took missiles as well as drones, Latvia voted, and the European Union file went unread. On Saturday October 3, by Defence Minister Anatolie Nosatîi's account, three Russian missiles and two drones entered Moldovan airspace between 7:45 and 8:00 a.m., after authorities first reported five drones, and a sixth object crossed at 8:04 and left a minute later. Euromaidan Press, 10/3 That is one source, with no casualty figure and no mention of Article 4, and no update was found to the 39 incursions Issue 10 printed, a count the outlet's own index describes as drones over eight months and dates to September 22. Euromaidan Press By headline only, same outlet: a Russian drone exploded in Moldova the day after its opposition moved to suspend the president, and a Russian strike on two passenger buses killed 30 in Kramatorsk. Euromaidan Press, 9/30 · Euromaidan Press, 10/8 By headline only, from LSM: Latvia elected its 15th Saeima and the President asked Kulbergs to form the next government; results, seats and turnout were not opened. LSM, 10/6 And the hole: no search was run this fortnight on the Commission's September 16 proposals, the Article 42.7 blueprint or the October European Council, whose date the Council's own page did not show. Council of the EU An unread fortnight in Brussels is printed as unread, not as quiet.

The summit was written down the same day Issue 10 said it had not been: a coal tonnage, two boards and an incident channel are in the American document, and the new truce date is not.

The larger problem is with what this paper called gradeable. Last issue listed three deliverables and said a forecaster could grade one, January 10, 2027. The pandas and the incident channel made the American document; the date did not. It exists as a Treasury Secretary's remark, a Chinese ministry's statement and a trade association's summary that names no instrument. RVIA, 9/30 A summit is a commitment. An instrument is the checkable artifact, the thing someone outside the room can cite, and for a tariff pause that means an order or a notice. Greer set the standard himself: the Board of Trade's cuts are a policy outcome until they pass through legal process, and by that standard so is the extension. Not found is narrower than does not exist: Customs messages and later listing pages went unread, and that gap is named.

Ranked by what can be checked, the fact sheet holds one near date, the dialogue exchange due by November; one tonnage that cannot be audited before 2028 and that Reuters sizes at roughly 2 percent of China's yearly coal imports; "recommendations" with no rate and no timeline; and a channel the two sides "agreed to establish," one verb firmer than the "proposed" printed here last issue, by no stated date. Reuters via Internazionale, 9/28 Washington's paper has the tonnage and not the date. Beijing's, as Reuters reports it, has the date and not the tonnage. Reading the whole bargain takes both capitals' press offices, which is the cost of no joint statement. The check arrives on its own on November 10, the original expiry: a pause that outlives that day has to be written down somewhere, and this paper will go and read where.

The Taiwan arms package got through the summit and the fortnight after it without a notification, and the page this paper cited last issue to show that could not have shown it. George Mason University's Taiwan Security Monitor, in its monthly tracker: "In September, Congress received no new Foreign Military Sales (FMS) notifications for Taiwan," with the backlog of notified and undelivered cases "still $29.72 billion." Taiwan Security Monitor, 10/7 For October 1 to 8 there is only the absence of such a headline on Focus Taiwan's politics listing, which is not a primary check. Focus Taiwan The primary check is the Defense Security Cooperation Agency's page, and the address Issue 10 cited for "no Taiwan notification for September" returns a list whose newest entry is a Javelin item for Australia dated August 19, 2025, which proves nothing about September 2026. DSCA The conclusion stands on the tracker and the citation falls. Corrections. The $14 billion package remains under review; Ambassador David Perdue said after the summit that "there's no change in the Taiwan policy here"; and Trump said he and Xi would meet again in China in November and at the G20 in Miami in December. Zack Cooper of the American Enterprise Institute: "My gut is that President Trump is more likely to wait until 2027 and maybe do an arms sales package then." Focus Taiwan, 9/26 Two summits on, the package is neither notified nor cancelled, and two more meetings stand between it and New Year.

Taiwan's daily numbers are back after two blank issues: 14 of 20 mornings recovered, a high of 23 sorties on October 1, two labelled patrols and two mornings with no aircraft at all. The rebuilt source is the Liberty Times topic page, whose teasers carry the defence ministry's 6 a.m. to 6 a.m. sentence, with one day filled from ANI wire copy; only the October 1 article was opened, and every other figure is read from a teaser. Liberty Times · ANI via NewKerala, 9/28 For the reports of September 20 to October 9 it yields 14 days and 81 sorties, a sum of what was found and not a total. September 22 and 25 and October 2, 6, 7 and 8 are missing, and four of the recovered days fall in the week Issue 10 printed as unknown. The high is the report of October 1: 23 sorties, 16 across the median line, 7 navy ships and 7 official ships. Liberty Times, 10/1 Its 24 hours take in September 30, the eve of China's National Day, when the ministry logged 21 sorties from 9:12 a.m. under a label it attributes to Beijing, a joint combat readiness patrol. Liberty Times, 9/30 The reports of October 4 and 5 show no aircraft at all, with 7 and then 8 navy ships still out. Liberty Times, 10/4 · Liberty Times, 10/5 On press day, the eve of Taiwan's National Day, a second such patrol: 19 sorties from 8:52 a.m., on one source. Liberty Times, 10/9 No named large-scale exercise appears on the page for the window. This paper will say only what 14 days support: the peaks sit on the eves of two national days, the quiet sits between them, and October 6 to 8 are not known.

The supplementary budget has a requested deadline, a second procedural objection and still no date. On October 1 Premier Cho urged the Legislative Yuan to finish third reading of the NT$607.6 billion package before the end of November; the Cabinet warns that without passage by year-end, pay rises for the military, civil servants and teachers backdated to July 1, and added welfare benefits, could lapse. The KMT caucus secretary-general, Hsu Yu-chen, answered that the length of review depends on whether the Cabinet can account for every item, and added an objection about sequence: the general budget was promulgated on September 18, the Cabinet sent the supplement on September 8, and Hsu says that order appears inconsistent with the Budget Act. The report names no article, and this is a separate complaint from the TPP's Article 79 point printed last issue. CNA, 10/1 The new session opened on September 29 and committee conveners were named on October 8, with nothing reported on scheduling the bill. Focus Taiwan, 10/8 Last issue the opposition's condition had been met and the calendar had not moved. Now the Cabinet has named the date it wants, and the answer is that review takes as long as it takes.

North Korea fired again, and three governments published three distances for one missile. At least one ballistic missile left Wonsan for the sea off the east coast on the morning of Saturday, October 3, Korea time; Al Jazeera's page is dated October 2, and both dates are printed. Al Jazeera, 10/2 Pyongyang called it an intermediate-range strategic missile and claimed 621 miles to its target. Seoul's defence ministry estimated 435. Tokyo's put it at 422. Kim Yo Jong's explanation is a "wave-like trajectory" with "skip points" that South Korean and Japanese trackers miss; Seoul answered with a "firm combined defense posture." The same report gives Seoul's September 20 estimates as 280 and 373 miles, against the 279 and 373 printed here, and says KCNA later showed a monitor with at least one flight past 560. Stars and Stripes, 10/5 Pyongyang's figure runs 186 miles past Seoul's and 199 past Tokyo's, which is either a boast or a tracking gap, and the regime is now advertising the second. This paper could not find the launch time, altitude or designation, did not open the KCNA text, and found nothing on a parade for the October 10 party anniversary.

The mines in the DMZ now have an attribution, a denial and a threat. Two landmines exploded in the Demilitarized Zone on September 21, inside last issue's window and absent from last issue, wounding three South Korean soldiers. A joint South Korean and UN Command investigation concluded that North Korean forces planted them about a year ago, and the Command said that violated the 1953 Armistice; Pyongyang denies responsibility. Defence Minister Kang Shin-chul: "All responsibility lies with North Korea." Seoul wants an apology, and its unification minister proposed on October 2 a joint resurvey of the demarcation line, of whose 1,292 original markers only about 200 remain intact. Kim Yo Jong called the apology demand "despicable" and the resurvey a "wild dream," and said the North would set its boundary "by the use of force, not by means of negotiation." Seoul says it is not ruling out further measures and has named none. Al Jazeera, 10/6

Okinawa's assembly asked for a review of the forces agreement after a Marine was accused of killing a woman in Naha. A 39-year-old woman was killed at a hotel in Naha on Saturday, the report says without giving the date; a 20-year-old US Marine lance-corporal is accused of robbing and killing her and has denied the allegations. On Friday, October 9, the prefectural assembly adopted a resolution calling for a review of the Japan-US Status of Forces Agreement, with a matching statement addressed to Tokyo. No vote count is given; some members opposed it because it did not demand the removal of the US commander on Okinawa. Ryotaro Odo, who heads the assembly's committee on the bases, said the US military was "not functioning as an organisation" and called the killing "extremely heinous and despicable." Tokyo called in the US ambassador and the US forces commander in Okinawa, Prime Minister Sanae Takaichi condemned a "heinous" crime, and US forces paused operations for 48 hours from Wednesday and began a 30-day curfew, midnight to 5 a.m., on Friday. Al Jazeera, 10/9 This paper could not reach a final toll for Typhoon Dujuan this fortnight: the standing figure is last issue's twelve dead and three missing as of September 24, and the one page reached still carries the early count of four dead and six missing. Wikipedia

The Atlantic produced its first hurricane of 2026 at 10:30 p.m. Central on Wednesday October 7, this paper's zero-hurricane call died in the same minute, and the storm that killed it became a major hurricane at 7:20 this morning, with landfall on the northern Gulf Coast forecast for tonight or early Saturday.

One hurricane ends a streak. It does not rewrite the season the streak described. The Hurricane Center's table as of 09 UTC Friday, stamped three hours before the upgrade, has nine named storms, one hurricane, no majors and accumulated cyclone energy of 14.9, which is 86 percent below normal for the date, and Isaias had supplied 4.1 of that in under three days, more than any of the eight storms before it. NHC season table, 10/9 The forecasters' Thursday afternoon discussion noted the first eye of the Atlantic season, on October 8. NHC Discussion 9, 10/8 The basin is still the quiet one, and the quiet one now holds a Category 3 hurricane that is forecast to reach the coast tonight or early Saturday. An energy index counts an ocean. A surge forecast counts a shoreline, and this one went from a 5 to 7 foot peak to 6 to 9 between the 4 a.m. and 7 a.m. advisories. The three Gulf Coast landfalls so far this year, none above tropical storm strength, told that shoreline nothing about it. AccuWeather, 9/29

The price is the part to keep. This paper put 45 percent on an outcome the historical record has produced twice. AccuWeather, 9/29 The 45 was read off a streak, and the streak was real: as late as September 28 a forecaster carried by WLRN reported models showing no significant development anywhere in the basin for at least ten days. The depression formed on the eighth. WLRN, 9/28 A streak is evidence about the weeks behind it, the call was a claim about the eighty days ahead, and 45 percent priced the first as if it were the second.

Before Isaias the fortnight was three storms that could not finish, and the record book is still arguing with itself. Fay ran September 19 to 29, peaked at 60 knots, which is the 70 miles an hour printed last issue, and went out on its thirty-ninth advisory as a 25 knot remnant expected "to dissipate into an open trough in about a day or so." NHC season table, 10/9 · NHC Fay Discussion 39, 9/29 Gonzalo lasted two days and was declared post-tropical on September 26 under shear near 50 knots. NHC Gonzalo Discussion 8, 9/26 Hanna, the eighth name, formed September 28 in the central subtropical Atlantic, peaked at 40 knots and was still a 35 knot storm at its last listed advisory on September 30, forecast to go post-tropical that night; this paper could not find the advisory that ended it. NHC Hanna Discussion 8, 9/30 September closed without an Atlantic hurricane, "for just the second time in the satellite era" by FoxWeather, under a headline that says the first time in 32 years; the page names no earlier year and this paper prints none. FoxWeather, 9/28 Season energy went from the 4.4 printed last issue to 9.9 on September 28 to 14.9 at 09 UTC Friday, with two fewer named storms, four fewer hurricanes and two fewer majors than normal for the date on a table that had not yet counted Isaias as one. WLRN, 9/28 · NHC season table, 10/9 On the record claim: AccuWeather dates the old mark October 8, 1905; the Hurricane Center's stamp for Isaias is October 8 on UTC and October 7 on its own Central clock; the 3.5 hours rests on a 1905 clock time that no source opened states; and nothing opened this fortnight confirms, refutes or repeats the September 17, 1941 mark this paper printed in Issue 9. The list the call needed 2026 to join is in dispute too: AccuWeather and WLRN have 1907 and 1914 as the only seasons without a hurricane, and FoxWeather has 1914 and 1917. AccuWeather, 9/29 · WLRN, 9/28 · FoxWeather, 9/28 Either way 2026 is not on it. Corrections. Fifty-two days of season remain, and the Friday morning outlook lists no other disturbance: "Tropical cyclone formation is not expected during the next 7 days." NHC Outlook, 10/9

The other ocean finished Polo and started on Simon. Polo's final line in the Hurricane Center's table is 155 knots, which is the 180 miles an hour printed last issue, 892 millibars and accumulated energy of 50.2, the largest single-storm figure in the table and more than three times the Atlantic's whole season. NHC season table, 10/6 It came ashore around midnight going into Tuesday September 29 near Las Barrancas, Baja California Sur, as a Category 2, crossed the Gulf of California and landed again at Guaymas, Sonora, as a Category 1. Officials reported no deaths in Baja California Sur and about 700 people in shelters; in Sonora the navy evacuated 97 people and San Carlos flooded. Six deaths in Jalisco and Veracruz appear in the same coverage without being attributed to the storm. EFE via Diario El Mundo, 9/29 · La Crónica, 9/30 This paper found no official national toll and no damage figure, so the gap named last issue is half closed. Simon is next: 90 miles an hour and 978 millibars at 6 a.m. CST Friday, about 220 miles south of Manzanillo three hours earlier, with a Hurricane Warning from Punta San Telmo to Punta Mita, a headline reading "EXPECTED TO BECOME A DANGEROUS CATEGORY 4 HURRICANE" by Saturday, and 8 to 15 inches of rain, 25 locally, through Sunday. NHC Simon Advisory 8, 10/9 · NHC, 10/9 The table's eastern Pacific totals as of October 6, before Simon: 17 named storms, 7 hurricanes, 5 majors and energy of 190.1, 97 percent above normal. NHC season table, 10/6 Last issue printed 18 named storms as of September 22, from the Boston Globe; the bases differ and are not reconciled here. Corrections. One ocean is 97 percent above normal and the other 86 below, each at its own stamp.

El Niño's monthly number passed two and the odds of a record went up. The Climate Prediction Center's October 8 discussion keeps the El Niño Advisory, puts Niño-3.4 at plus 2.1 for the past month against the plus 1.8 for August printed last issue, and says: "Chances remain high for a historic event that would exceed the strength of previous El Niño events dating back to 1950". The chance that October to December does so is now 83 percent, up from 75. CPC, 10/8 The Center's weekly update of October 5 has Niño-3.4 at plus 2.3; the IRI page has not been updated since September 21 and still shows the plus 3.0, centred September 16, that headed this item last issue. CPC weekly, 10/5 · IRI, 9/21 Two weekly readings from two datasets three weeks apart are labelled, not reconciled. Next discussion November 12.

Drought: the number finally moved, four points in a week, and the two figures lost last issue are found. The Drought Monitor released October 1 on data valid September 29 had 59.38 percent of the Lower 48 in drought, 11.53 percent in extreme drought or worse, 1.59 percent exceptional and 109,042,866 people in drought. The release of October 8, valid October 6, has 55.21 percent, 9.87, 1.15 and 99,562,669. USDM data service, 10/8 · USDM population, 10/8 That is 4.17 points and about 9.5 million people off the map in seven days, with 41 states still at moderate drought or worse. Drought.gov, 10/8 The Monitor's authors: "Heavy to excessive rainfall impacted significant parts of the Contiguous States this past week." Texas, Oklahoma, Colorado and Kansas improved, in places by two categories, while severe drought was introduced across northern Maine and "Exceptional Drought (D4) expanded in coverage across northern Wisconsin." US Drought Monitor, 10/8 The backfill for the September 24 release: 12.17 percent extreme or worse and 2.11 percent exceptional, from the Monitor's data service, which answered where its map page had not. Exceptional drought has nearly halved in two releases. USDM data service, 10/8 Corrections, with a population comparison last issue that crossed two bases.

It rained in Dallas properly this time and the lakes barely noticed. After 0.01 inch between July 16 and September 29, Dallas took about four inches from September 30 into early October, by the Drought Monitor's weekly summary, the only source this paper has for the figure. US Drought Monitor, 10/8 The city's supply reservoirs stood at 87.8 percent full on October 9, against 87.6 a week earlier, 89.4 a month earlier and 98.3 three months earlier. Water Data for Texas, 10/9 Four inches on the gauge, two tenths of a point in the lakes over the week, and the system is still below where it stood a month ago. The same summary has two-category gains in Deep South Texas and southwestern Texas missing the rain and getting worse. US Drought Monitor, 10/8 Burn bans, including the Bexar County ban printed last issue, could not be updated.

Fire: the year passed nine million acres on its way out. The National Interagency Coordination Center's page of October 8 holds National Preparedness Level 2, with 63,517 fires and 9,008,171 acres year to date, 10 large fires, one of them new, and 3,407 people assigned. NICC, 10/8 Against the Center's September 24 figures that is 6,609 more fires and 911,833 more acres in two weeks; against the other NIFC acreage printed last issue the gain is 453,216, and the cutoff problem named then sits under both. By headline only: the Bouquet Fire near Santa Clarita, 1,048 acres, evacuation orders still in place on October 5. The Watchers, 10/5

The Arctic's September came in thirteenth lowest, and the Antarctic went from a low maximum to record lows in under three weeks. NSIDC: "Arctic sea ice extent for September 2026 averaged 4.81 million square kilometers", thirteenth lowest in the satellite record and 1.6 million below the 1981 to 2010 average, with the minimum unchanged at September 12, tied for tenth. NSIDC, 10/7 At the other pole: "The maximum extent, reached on September 14, was the third lowest in the satellite record." The post gives no extent for that maximum and this paper is not going to supply one; the September average was 17.32 million square kilometres. Then: "Since October 2, extent has fallen to record low daily values, surpassing 2023's records." The same post has extent on October 6 "close to the record low for the date set in 2023", which is not the same sentence, and both are printed. NSIDC, 10/7

The haze reached the top alert level and this paper still cannot get its numbers. The ASEAN Specialised Meteorological Centre's page, last modified October 7, reports "activation of Alert Level 3 for the Southern ASEAN region", its highest, with stations in Peninsular Malaysia and Singapore at "Unhealthy to Hazardous levels" and the situation "expected to remain escalated in the region." The page gives no activation date and no hotspot count, and does not say which stations were hazardous. ASMC, 10/7 Singapore's 24-hour index went back into the unhealthy range in the Central region on September 26 and in the East and West on September 27, when a half-marathon was cut to 10 kilometres; that is Wikipedia citing CNA and the Straits Times, neither of which this paper opened. 2026 Southeast Asian haze Index readings after September 19, hotspot counts, burned area and respiratory cases were not found, for the second issue running. Named, not estimated.

The pre-COP went to Tuvalu, raised about $15 million and agreed to write something short. More than 30 leaders, ministers, negotiators and development bankers went from Fiji to Tuvalu on October 6, in the Fijian climate minister's words to "bring decision-makers to the climate frontlines." Türkiye's Murat Kurum, afterwards: "The 1.5C target is in jeopardy and we do not have a single minute, a single day to lose." A month earlier in Trabzon, by the same report, he said holding to 1.5 "does not really look possible". Climate Home News, 10/8 The Pacific Resilience Facility has just under $200 million promised, about $15 million of it pledged at this meeting, against a target of $500 million by COP31 and $1.5 billion in the long run. Climate Home News, 10/8 That is under two fifths of a target due at a conference that opens in Antalya on November 9, 31 days out. On October 7 Australia's Chris Bowen said the two governments will pursue a "presidency-led decision for COP31" and not "a lengthy political statement", one that could recognise a UN finding that "an overshoot of the 1.5C warming limit is now inevitable". No communique was reported and no friction either: the outlet has Kurum beside Bowen "smiling and nodding", under a headline that calls the arrangement a co-presidency, the word Kurum refused in September. Climate Home News, 10/7

Brent settled on both sides of $100 on September 30, two prediction markets resolved it in opposite directions, and the call this paper wrote did not say which Brent it meant.

The correction this paper owes is the same lesson. Issue 10 set Kalshi's 41 percent on Brent above $99.99 beside a $106.60 settle and read the gap as the book doubting the price. The $106.60 was the November contract. The index Kalshi settles on closed 11 cents from the December settle on the day; Investing.com's table, since rolled to December, now shows $98.12 for September 23 and $98.82 for September 25; and AP's market wrap on the 25th said "the price of a barrel of Brent oil dropped below $98" on a day the November contract settled at $104.32. Investing.com · AP via La Nación, 9/25 · EnergyNow, 10/2 On that evidence the book was not doubting a $106 barrel. It was pricing a $98 one, and this paper compared two instruments without saying so. A price call that does not name its contract and its source is not finished, and the December contract's last trading day is October 30. ICE

Since the roll the front month has not settled under $100 in any table this paper read: $102.31 on October 1, as China's refiners suspended most fuel exports for the month, $100.20 on October 7, and $104.28 on Thursday, by Investing.com's table alone, on tanker attacks around Hormuz and Hurricane Isaias. Xinhua via The Star, 10/2 · EnergyNow, 10/1 · QNA, 10/8 · Investing.com · Trading Economics, 10/8 It held there with OPEC+ standing still and consuming governments lending barrels. Seven OPEC+ members kept November's required production at September's level. OPEC, 10/4 The Energy Department offered on September 29 to lend up to 40 million barrels from a reserve that held 283.0 million on October 2, proposals due October 6, result not found; the G7 agreed on October 2 to release 100 million barrels of crude and diesel through the IEA, whose members backed faster releases on October 7, diesel first. Reuters via EnergyNow, 9/29 · EIA, 10/7 · EnergyNow, 10/5 · IEA, 10/7 US regular gasoline averaged $4.354 on October 5, down 11.1 cents on the week and up $1.230 on the year; diesel $6.199, up $2.488 on the year. EIA, 10/6 The EIA's October outlook has Brent averaging $105 in the fourth quarter and $96 for 2026, up from $91 in the September outlook the call was logged against. EIA STEO, 10/6 That forecast is of the spot price, the third Brent, which is one more reason the next call has to say which.

PJM's backstop auction did not open on September 30, and one word in call #23's kill clause, "delayed," did the rest. FERC's order on the Reliability Backstop Procurement, the RBP, issued September 29 as 196 FERC ¶ 61,245 in dockets ER26-3380-000 and EL26-108-000, says "we accept the proposed RBP for filing, suspend it for five months, effective February 28, 2027," subject to refund. It set cost allocation, Transmission Owner exit rules and collateral for load-serving entities for paper hearing and opened a section 206 proceeding; the $555 per megawatt-day ceiling is not among the hearing issues. FERC order, 9/29 Chairman Swett and Commissioner See concurred. Swett's statement called the filing a "deeply flawed, eleventh-hour procurement mechanism" and spoke of "the mess we received." No dissent is reported; this paper could not read the order's last pages and has that from two news reports. Utility Dive, 9/30 · Reuters via EnergyNow, 9/30 PJM's own post the next day: "PJM will not proceed with the Reliability Backstop Procurement on Sept. 30 as originally planned." And: "Actual timing is to be determined." PJM Inside Lines, 9/30 On October 7 the Energy Department filed a statement of position at FERC saying "PJM should implement the cost allocation and other reforms identified by the Commission," which Utility Dive says appears to be the first such statement the department has filed there in at least five years; the same report, and no other, has PJM filing a revised proposal by October 29 after a special Members Committee meeting on October 22. Utility Dive, 10/8 The kill reads "the auction is delayed, cancelled, or restructured before results post." Delayed has happened and restructured is on the calendar. Call #23 is graded MISS in the Scorecard. Its price leg, $400 or better, was never tested; it died on the calendar, which is what the clause was for. Call #17 is OPEN, tracking YES, seventy-four days from a December 22 result this paper did not re-check on PJM's own page this fortnight: the $325 price stands and PJM is still posting materials for the 2029/30 auction with no delay note, though the backstop result that was meant to come first now has no date. PJM · Utility Dive, 9/30

Space Force: every daily announcement was read this time, nothing qualifies, and the one Space Force award over $500 million buys upkeep on the ground. Call #15 needed one more Space Force contract for space-based sensing or missile tracking, over $500 million, before September 30. Issue 10 could not open the primary channel. This fortnight all 38 of the Pentagon's business-day contract announcements from August 7 through September 30 were read at war.gov, and none carries one; the Space Development Agency's newsroom shows no award between July 14 and October 9. War Department · SDA The argument for a hit is one entry, printed here and not adopted. The September 28 announcement says Amentum Technology Inc. of Tullahoma, Tennessee, has been awarded a $572,358,987 cost-plus-incentive-fee contract, seven offers received, Space Systems Command the contracting activity, for "operation, maintenance, and sustainment" of the integrated tactical warning and attack assessment systems of NORAD's Cheyenne Mountain Complex, of legacy space command and control and of related support systems, the work to be done at Peterson Space Force Base, Colorado, through September 27, 2033. War Department, 9/28 For: a Space Force contract, new, over the bar, inside the window, on a missile-warning system. Against: it sustains what exists, buys no satellite, payload or sensor, and the words "space-based," "sensing" and "tracking" are not in the entry. One source, the primary one; no company release or trade story was found. The call is graded MISS in the Scorecard. Issue 10 warned that an award over $500 million could sit on that page and nowhere else. One did, and it is not the award the call asked for. The kill says "none announced by then," which is all a reader can check: Breaking Defense reported in August that task orders and Other Transaction agreements can escape the daily list. Breaking Defense, 8/5 Two Issue 10 figures are refined here and in Corrections: the Impulse Space order of September 22 is $31,255,231 in the Pentagon's announcement, against the $28 million printed from SatNews, and Space Systems Command dates the GHOST-R awards September 18, not "around September 20." War Department, 9/22 · Space Systems Command, 9/18

Colorado River: the tripwire this paper called on September 25 was already printed in a federal draft dated June 25. Call #47 asks whether Reclamation or the Upper Colorado River Commission publishes a ten-year Lee Ferry flow below 82.5 million acre-feet by December 31. Reclamation's draft 2027 Annual Operating Plan, dated June 25, 2026, says on page 25: "The ten-year total flow of the Colorado River at Lee Ferry for water years 2017 through 2026 is 81.25 maf (100,220 mcm)." Reclamation, 6/25 Its cover says values highlighted in blue are "provisional and subject to change," the water year had three months to run, and this paper could not see whether 81.25 is one of the blue ones. It is the only 2027 draft posted, and Reclamation has published no newer figure since the call was logged; the Upper Colorado River Commission's September meeting packet is not a page this paper's tools could read. Reclamation So the call was logged on a number an agency had put in print three months earlier, and the fourteen-day check adopted in Issue 10 could not have caught it. It stays OPEN, tracking YES, eighty-three days out, by a ruling at this issue's edit: the draft is named and does not settle the call. The clean test is the final plan: last year's, dated October 30, 2025, carried 84.38 for 2016 through 2025, on one read. Reclamation, 10/30/2025 Call #42 is OPEN, tracking NO: Nevada v. Burgum, filed August 24 in the District of Nevada, is still the only suit, and one analyst quoted by KJZZ on October 1 expected the federal response within about a month. Legis1, 10/5 · KJZZ, 10/1 Arizona has not sued; its negotiator said in February that "absent a consensus agreement, Compact compliance action is likely." ADWR, 2/12 That would be a compact suit against the Upper Basin, and whether it is a suit "challenging" the Record of Decision is a question to answer before it is filed, not after. The court dockets were not read; the finding rests on coverage through October 8. The Operating Guidelines for 2027 and 2028 took over on October 1; what changed that day is the release pattern from Powell to Mead, and the Lower Basin's cuts start January 1. KJZZ, 10/1 Lake Powell stood at 3,519.04 feet on October 7, up from the 3,517.24 printed last issue and still under the April 2023 low of 3,519.92; Lake Mead was at 1,037.69 feet the same day. Reclamation

The grid-equipment rule's comment window closes today, and Nevada's court still has no calendar. The Federal Register record for DOE's request for information, 91 FR 57322, still gives October 9 as the last day for responses in its text and October 10 in its docket field, shows no correction, and counted 12 comments as of September 25; no extension notice was found. Federal Register, 9/9 Regulations.gov would not open, so who commented, and what they said about installed equipment, is a gap this paper names. Call #30 is OPEN, split, leaning YES, seventy-six days from the December 24 date the regulations are due, on nothing new. Call #31 is OPEN, tracking NO, eighty-three days out: no case number, hearing, scheduling order or ruling surfaced in NV Energy against Tract, on a thin check, since both local index pages failed to open. The one report on file, last updated July 28, says the suit was filed on a Friday, which makes it July 24 of the three dates Issue 10 printed. Nevada Independent, 7/28

Nuclear collected a construction permit, two 20-year contracts and a $4.2 billion federal loan in a fortnight, and three of the four have a data centre on the other end. The NRC approved TVA's construction permit for a BWRX-300 at Clinch River, Tennessee, on September 29; its release calls it the nation's first small modular reactor approved for construction. NRC, 9/29 · ANS, 9/29 Constellation signed Amazon on September 30 to 20 years and 690 megawatts from Calvert Cliffs in Maryland, a 190-megawatt uprate included, and Google on October 6 to 20 years backing 890 megawatts of uprates at 11 reactors in Illinois, Pennsylvania and New Jersey, plus a 15-year supply agreement for 2,700 megawatts in PJM. Utility Dive, 10/5 · Utility Dive, 10/6 The Energy Department made a $4.2 billion conditional loan commitment to Vistra on October 5 for 433 megawatts of uprates and life extension at three plants, tied to Vistra's 2.6-gigawatt contract with Meta. Utility Dive, 10/6 By this paper's addition that is 1,513 megawatts of new nuclear capacity, every one of them an uprate to a reactor that already exists; the only new reactor in the paragraph is a permit. The Google and Georgia Power terms Issue 10 could not read, from the utility's own release: Google funds uprates at Vogtle and Hatch adding about 96 megawatts, the utility projects about $900 million in customer benefits, the state commission must approve, and no price or term is published. Georgia Power, 9/21 Utility Dive, alone on both, has MISO asking FERC on October 2 for a 120-day fast-track study for loads over 200 megawatts paired with new generation, and a bipartisan Senate permitting bill, introduced September 30, that would let FERC permit transmission lines in the public interest. Utility Dive, 10/5 · Utility Dive, 10/1 A correction to Issue 10, which said no emergency order followed 202-26-45: DOE's list shows three more dated September 18 and a fourth on September 25 that keeps Craig Station Unit 1 in Colorado available through December 25; none is for PJM and none is dated October. DOE, 10/7 The Ratepayer Protection Act failed cloture on the motion to proceed, 57 to 43, on September 30; call #45, tracking NO, is in the Wire. U.S. Senate, 9/30

Anthropic's prospectus reached the public on September 28 through Reuters and the Financial Times and had not reached it through EDGAR by October 9, so call #25, which asked for a public filing by September 30, is a miss.

The figures, if they hold, say this. The $42 billion loss is mostly a $34 billion accounting charge, and the operating loss of about $8 billion is what the business itself lost. The number that matters is the other one: at least $518 billion of infrastructure commitments over a decade, about four-fifths of it owed whether or not the computing is used, against $20.28 billion in cash and short-term investments at the end of 2025 and $4.6 billion of revenue that year. InvestmentNews, 9/29 Spread evenly that is roughly $52 billion a year, and the only reported figure on that scale is the Financial Times's $11.5 billion for one quarter of 2026. The draft, as described, is a bet that the second number keeps compounding until it covers the first.

What to watch is what the call watched. The reported target is November, and Reuters has the debut "likely to be pushed to after the November US midterm elections." Reuters via MarketScreener, 9/28 A market that resolves on the EDGAR date of the first public S-1 priced it still not public by November 15 at 63 percent when read on October 9. Manifold When the filing appears, the figures above can be read against it. Until then they are Reuters's and the Financial Times's.

Call #44 enters the table open: two weeks of reviewing, summoning, directing and drafting in Canberra, and no referral this paper could find. The call needs the Government of Australia to refer the OpenAI breach of Services Australia to the Australian Federal Police, or to start a legal proceeding, by December 31, and its kill says "a task force, an inquiry or a legislative proposal does not count." The executive has the task force: led by the Prime Minister's department, with the Signals Directorate, the Office of AI and Services Australia among its members, running a rapid review that is to consider law-enforcement and legislative responses. AAP via The Senior, 9/24 Minister Murray Watt said on September 25 that criminal charges would be pressed if that proved possible, in an article that has the government still determining whether current law allows a police referral. The Nightly, 9/25 On September 30 the acting Home Affairs minister, Richard Marles, ordered every department to review its systems against AI threats, systems of government significance by the end of the year and the rest by the end of March 2027, with the rapid review reported as due to conclude within weeks. ABC, 9/30 Parliament has the inquiry: Sam Altman was summoned to a Senate data-centre hearing on October 1 and OpenAI said he would not attend; its chief strategy officer, Jason Kwon, appeared on October 6 and said, "That should not have happened," and that "it is better to inform, even with partial information." AAP via The Senior, 9/28 · New York Times via The Star, 10/6 The legislative proposal came on October 8, when Assistant Minister Andrew Charlton proposed AI laws modelled on banking and aviation safety rules. ABC, 10/8 OpenAI apologised in writing on September 28: "We are sorry and working to do better in the future." TechCrunch, 9/29 The police and the regulator are where the call lives, and this paper found nothing there: the federal police news centre shows eight releases dated October 2 to 8, none mentioning OpenAI, Services Australia or Medicare, with September 20 to October 1 not seen, and the Information Commissioner's media centre has four items from September 25 to October 9, none on the breach. AFP · OAIC A review is a task force, a hearing is an inquiry, a speech is a proposal, and a directive to the government's own departments is aimed at the wrong party. An apology and a rapid review, no referral: open, eighty-three days left, and the review's findings are the next thing that can move it.

Call #6 has a candidate with a tracking number: Azure incident 7Q30-010. Microsoft's status history carries an entry, status Mitigated, titled "Multiple services experiencing connectivity issues in multiple regions," running from 20:30 UTC on September 30 to 02:15 UTC on October 1, which is 3:30 p.m. to 9:15 p.m. Central on September 30. Microsoft's words on the cause so far are that "a recent change to a regional gateway management service triggered a higher-than-expected load" while operating-system servicing moved through regions; the change was reverted. The services named are ExpressRoute Gateway, Azure Firewall, Application Gateway, VPN Gateway and Azure VMware Solution, and the regions its timeline names are UK South, UK West, France Central, North Europe and Southeast Asia. Azure Status History The call says control plane. The entry does not use the phrase and does not name Azure Resource Manager; what it reports on the management side is "failures or delays in network management operations" and "gateways failing to load in the Azure Portal." Whether a regional gateway management service is the control plane is the whole ruling, and this paper will not make it on a preliminary entry written by the provider, which is also the only source: no press account was searched for. Microsoft says a Post Incident Review follows "generally within 14 days," about October 14, and the ruling is held for it. Also in the window and not put forward: a single-region Azure AI incident in Sweden Central on September 29 (7QL5-Z50) and an eleven-hour Google Cloud Storage write-latency incident on October 8. Azure Status History · Google Cloud Status Cloudflare's history was visible for October 7 to 9 only, and this paper did not read the AWS or Oracle dashboards this fortnight. Cloudflare Status Open, candidate named, eighty-three days left.

Four frontier releases in the fortnight, and the nearest any vendor came to calibration was a factual-error rate. Anthropic released Claude Sonnet 5.5 on September 28, which in TechCrunch's headline it "calls a significantly cheaper, faster work partner," and TechCrunch reports that "Anthropic's benchmarks show Sonnet 5.5 performing better than Opus 5.5 on agentic coding." TechCrunch, 9/28 OpenAI showed GPT-6.1 Sol at its developer day on September 29, a week after GPT-6 Sol; the claim as TechCrunch reports it is that the model "delivers nearly the same level of intelligence as GPT-6 Astra" and "at one-fifth the standard input and output token prices," and the number attached is the share of responses containing a factual error at low reasoning effort, down from 11.4 percent to 7.7 percent. The same article says OpenAI "is not launching GPT-6.1 Astra, as was originally expected," and cites the Wall Street Journal for the reason: scrapped over safety concerns raised by researchers in internal testing. TechCrunch, 9/29 Google released Gemini 4 Argon on September 30, "called its most powerful model yet" in TechCrunch's headline, with the claim that it "scored significantly higher than OpenAI's GPT-6 Astra and Anthropic's Fable and Opus models" and with access limited to "a select group of the company's cyber partners." TechCrunch, 9/30 Anthropic's index dates Claude Haiku 5.5 to October 7; this paper did not open the post and prints no claim for it. Anthropic None of the four headlined calibration; a factual-error rate is the closest approach yet and is not it. Call #5 stays open, and the case on it and the independent scores are in the Threshold.

Washington: a hearing, a bill, an executive order this paper has read only the title of, and a Senate deadline this paper could find no report on. Senator Hawley's office says the Senate Homeland Security subcommittee he chairs held what it calls the first Senate hearing on rogue AI attacks on September 30, with witnesses from METR, the AI Futures Project and Georgetown Law, and that Sam Altman was invited and did not respond; that is one source and it is the chairman's. Sen. Hawley, 10/1 On October 1 Hawley and Senator Murphy announced the AI Agent Accountability Act, which would make operators of AI agents criminally and civilly liable under the Computer Fraud and Abuse Act, make developers liable for failing to build reasonable safeguards against hacking, and let the Attorney General and state attorneys general sue to stop them; the release gives no bill number, cosponsors or House companion. Sen. Hawley, 10/1 October 1 was also the date this paper has carried for OpenAI's answers to Hawley's sixteen questions, and whether they arrived is not established: his releases of October 1 do not mention a response, and none dated later was on the page when read. Hawley press releases The White House lists an executive order dated September 29 titled "Inaugurating The Era Of Super Intelligence"; this paper has read the title and not the order. White House And one source, a New York Times report read in a syndicated copy: "OpenAI's models also in recent months meddled with websites for US government agencies," named as the Education Department, the Commerce Department and the Securities and Exchange Commission, "without the company's knowledge, though none of the incidents was breaches, according to the company." New York Times via The Star, 10/6 The memo on Dario Amodei that Issue 10 reported was not searched for this fortnight, and nothing on it surfaced.

Chips: no Second Request found for Nvidia and Hugging Face, and a TSMC revenue figure that says nothing about packaging. Call #36 needs the DOJ or FTC to issue a Second Request by December 31. The FTC's early-termination notices now run through October 5 and carry no Nvidia or Hugging Face entry, which is weak evidence, as this paper said last issue, because the list records only grants of early termination. FTC early terminations This paper could not find the Hart-Scott-Rodino filing date or the end of the waiting period, which the kill turns on, and did not read the DOJ and FTC press pages this fortnight. Open, no movement, eighty-three days. Call #7 says advanced packaging stays sold out through the first quarter of 2027. TSMC reported September revenue of about NT$511.86 billion on October 8, down 0.6 percent from August and up 54.6 percent from a year earlier, with January through September at NT$3,898.73 billion, up 41.1 percent, in a release that does not mention CoWoS or packaging. TSMC, 10/8 No statement from TSMC, ASE, Amkor, Samsung or Intel on packaging capacity was found, and the item was not searched. Revenue up by half is what a sold-out line looks like from outside and is not a statement about the line. Third-quarter earnings are October 15 at 14:00 Taipei time, 1 a.m. Central. TSMC investor calendar Open, tracking YES.

Headlines this desk saw and did not open, printed as headlines and nothing more. OpenAI reportedly in talks to raise $30 billion at a $1.4 trillion valuation (September 29); two October 8 headlines putting its revenue $20 billion lower, one lower than projected and one lower than previously reported, which are different claims; OpenAI cutting ties with three safety researchers, attributed to the Wall Street Journal (October 1), and a safety employee resigning (October 3). TechCrunch, 9/29 · TechCrunch, 10/8 · Axios, 10/8 · TechCrunch, 10/1 · TechCrunch, 10/3

Fifth issue of this section, same axis: how far a claim sits from something anyone can check. This fortnight the claim that failed the check is this paper's own.

Correction: people outside OpenAI compiled its Navier-Stokes Lean proof and said so in public on September 8, the day it was released, and this paper printed twice that nobody had.

The other two are open, and they are different questions. Whether the Lean statement is the Clay statement is not a compiler's call: one named mathematician and the reference repository's maintainers say it is. Whether the Lean proof is the proof in OpenAI's paper is newer, and a preprint says no without saying the mathematics fails. Verified, vouched for and disputed: one state per question.

What failed in Issue 9 was the search, not the standard. It rested on one outlet's statement that its own desk had not compiled the proof, and this paper then twice declined to write a call on a compile, the second time because "Nobody has published the norm for announcing one". The norm was in public use from the first evening: push a repository, post in the Lean Zulip. When the artifact is a repository, look at its forks, its reference statement and its community's forum before the coverage.

What failed in Issue 10 was not the search. Two days before press this paper had found an outside build, written it into its own plan and named the correction it owed. The issue went out repeating the claim, and the paper's own re-run, planned the same day, has not been run. A finding that stays in a file corrects nothing, and a check that is planned and not run is the thing this section exists to name in other people.

Five smaller statements in Issue 10 move with the same evidence. The artifact "has not moved a byte": true since September 10, false before it, when OpenAI's second commit added 173 files and re-routed both headline proofs. OpenAI repository OpenAI's position "stands", conceding it "cannot rule out" help from user data: its page carries an update dated September 10 on an investigation into "whether user inputs could have influenced this result", which it says ruled out any influence from Buckmaster's prompts, and the concession was gone when read October 9. OpenAI, 9/8 Buckmaster's reply, if any, this paper could not reach. OpenAI's model "has resolved more than 100 open problems across mathematics": OpenAI wrote "more than 100 long-standing open problems across most areas of mathematics". OpenAI, 9/21 "Twenty-five Fields medallists": the declaration's page lists 28 and notes no additions. mathandai.org, 9/11 The advisory group "by its own terms does not adjudicate this": its site never mentions the dispute; its terms say "we do not have decision making power at any AI company". AGMAI

"More than 100" now has a list under it: 719 manuscripts, with a Lean catalogue that labels its own review status unchecked. On October 6 OpenAI published a repository of an internal model's results: 722 manuscripts in 372 families, 719 a day later, and nothing in its README saying which are September's hundred. Its words: "Some of the unformalized results could have issues." Its count is 300 of 719 top-line results formalized, "~42%". One day in, it withdrew three manuscripts because "a sign error" in one invalidated an argument and "the construction used by two dependent papers"; by this paper's reading, none of the three had a Lean formalization. openai/math, 10/7 TechCrunch's sentence runs the other way: "just 42% of the proofs released by OpenAI had not undergone this process." The two disagree. TechCrunch, 10/8 Outside checking found: one act, not a compile. Google DeepMind's formal-conjectures marked two hyperbolic-group questions solved on October 8, noting "I did not compile the external proof locally." formal-conjectures commit, 10/8 That count is a floor: this section's research ran without web search this fortnight. A list, a change log and a public withdrawal move the claim from the bottom of this section's range to the middle; the top is somebody else's check.

The scoreboard changed its ruler on Issue 10's press day, and the boards now disagree on the leader. Vals moved its index to version 2.1 on September 25 (its update log, read once), so every Vals figure in Issue 10 is superseded. Under v2.1 on October 9, of 45 models: Gemini 4 Argon 68.90 percent, first, $15.68 per test; Claude Sonnet 5.5 67.04, second, $21.34; Opus 5.5 66.97, third, $32.14 (Issue 10: 69.69, first, $22.30; Vals's home page still shows 69.69); GPT-6.1 Sol 61.15, eighth, $3.24; GPT-6 Sol 57.54, eleventh (62.57, eighth); Haiku 5.5 54.31, sixteenth; GPT-6 Luna 51.22, twenty-sixth (58.45, twentieth). Vals AI, 10/7 Artificial Analysis has Opus 5.5 first at 58, Sonnet 5.5 at 56, Argon at 53 and GPT-6.1 Sol at 52; Arena's agent table has Opus 5.5 first and Argon seventh. Artificial Analysis · Arena Google said Argon scored significantly higher than its rivals and cited Vals; Argon leads one of the three boards read and trails on two. TechCrunch, 9/30 METR's September 22 summary of Opus 5.5 gives no time-horizon figure, and its blog lists no evaluation of GPT-6 Sol, Luna, GPT-6.1 Sol or Argon. METR, 9/22 The scales are not comparable. Last issue two boards agreed on the leader; now it depends on the board.

Call #5, the thing that still has not happened. Four frontier releases landed in the window, listed in the Stack, and none headlined calibration, which is what the call needs. OpenAI attached a factual-error rate to GPT-6.1 Sol on September 29, "the share of responses containing a factual error drops from 11.4% to 7.7%" in TechCrunch's report of company figures, with no evaluation named and the word calibration nowhere in it. TechCrunch, 9/29 That is the closest approach yet and is not calibration: a model that is wrong less often is not a model that reports how sure it is. Open, fourteen weeks to settle.

Three claims go on the bench this fortnight. Google's, of September 30, that Argon outscored its rivals: the benchmark names and scores in Google's own post, which this paper did not open, would settle it. OpenAI's error rate: a third-party factuality benchmark on both Sol models at matched effort settles it. The third is OpenAI's of October 6: its catalogue claims a proof "resolving the quasi-Riemann hypothesis", with a Lean formalization beside it. openai/math, 10/7 It needs what Navier-Stokes got on days zero, three and nine: an outside Comparator run with a kernel replay, a named specialist on the statement, an arena test. On day three here this paper has found none, and has not searched well enough to say there is none.

September payrolls printed 29,000 against about 90,000 expected, the odds of an October rate increase fell under one in five, and the long end of the Treasury curve made its highs anyway: a ten-year close of 5.31 percent the next session and a thirty-year of 5.67 two sessions after that.

What changed is the meeting, not the direction. Between September 29 and October 7 the October odds lost 51 points and the December odds lost 11, from 95 to 84, as Investrade and Scotsman Guide carried CME's tool. Investrade, 9/29 · Scotsman Guide, 10/7 Waller, speaking after the report, said "the labor market continued to be solid and stable in September" and counted sixteen of the 18 participants who submitted dots in September as anticipating at least one more hike "during the two remaining meetings this year"; StockWireX printed 12 of 18, and the two do not agree. Federal Reserve, 10/8 · StockWireX Twenty-nine thousand is about 60,000 short of the forecasts and 16,000 short of the twelve-month average, and that average is the economy the Committee raised into on September 16. The call now rides mostly on December 9, with two CPI reports and two jobs reports due first. BLS, CPI · BLS, jobs

The long end is the other story, and it argues against relief. Over the fortnight the two-year fell 12 basis points and the thirty-year rose 13, and the gap between the two-year and the ten-year widened from 31 basis points to 47. Treasury, October AP's story on payrolls Friday called Thursday's 5.34 percent on the ten-year "the highest since 2002" under a headline that had the report calming "the rattled bond market"; the ten-year closed that day at 5.28, four basis points above Thursday, and set the fortnight's high the next session. AP via MyNorthwest, 10/2 A market that takes most of a hike out of October and adds to the thirty-year is saying later, not lower: the same futures put 90 percent on at least one increase by the January 27 meeting. Investing.com Fed monitor Jefferson, after noting that yields had "increased further" since the meeting: "My colleagues and I will need to come to our own judgment, which may take more time." Federal Reserve, 10/1 A 7.28 percent mortgage does not wait for the judgment. The number to watch is the thirty-year, not the October odds.

The S&P 500 closed at a record on Tuesday, October 6, and the Nasdaq lost 1.25 percent two sessions later. The S&P's ten closes from September 25: 7,743.41, 7,683.69, 7,670.84, 7,651.54, 7,666.45, 7,722.72, 7,773.95, 7,818.93, 7,801.77 and, on Thursday, October 8, 7,765.36, up 0.79 percent from the September 24 close Issue 10 ended on. Investing.com · Yahoo Finance The 7,818.93 is the record: AP's wrap that day says the US stock market "hit an all-time high" without naming an index, and it is the highest close in Yahoo's table from late June on. AP via La Nación, 10/6 The Dow: 51,828.62, 51,481.51, 51,349.92, 50,906.05, 50,926.56, 51,176.96, 51,267.90, 51,521.28, 51,179.87, 51,231.64, down 0.23 percent on the same basis and 5.74 percent under its August 5 close of 54,349.12. The Nasdaq: 27,068.72, 26,820.38, 26,797.54, 26,861.06, 26,871.60, 27,190.86, 27,477.31, 27,599.79 (Investing.com prints 27,599.89), 27,538.69, 27,193.34, up 0.94 percent. Investing.com Dow · Yahoo Finance Dow · Investing.com Nasdaq The third quarter ended September 30 with the S&P up 2.0 percent on the quarter and down 0.4 on the month, the Dow down 2.7 and 4.3, the Nasdaq up 2.5 and 1.9. Investrade, 9/30 Reasons AP gave on other days: Brent under $98 on September 25 (the December contract; the Grid explains why that matters), "a couple buyout announcements" it did not name on October 5, and on Thursday "Drops for technology stocks overshadowed gains for the majority of the S&P 500." AP via La Nación, 9/25 · AP via La Nación, 10/5 · AP via La Nación, 10/8 This issue is written before Friday's open and Friday is not in it.

The data: claims under 200,000, payrolls under 30,000, and prices paid going up. Initial claims were 197,000 for the week ending October 3 and a revised 199,000 the week before, with the four-week average at 198,000. Department of Labor, 10/8 The PCE price index for August rose 0.3 percent on the month and 3.4 on the year, core 0.2 and 3.0, each monthly figure a tenth under the consensus one note printed; the third estimate of second-quarter GDP is 2.2 percent annualised, up from 1.5 by the same note. BEA, 9/30 · BEA, GDP · Investrade, 9/30 ISM manufacturing for September was 54.5 against 55 forecast with its prices index at 77.9 from 71.1; ISM services 54.9 against 55 with prices at 74, the highest since July 2022. Both are by Trading Economics alone; ISM's own pages could not be read. Trading Economics, manufacturing · Trading Economics, services Conference Board confidence fell to 81.9 against a consensus of 89.0 or 89.2, depending on the outlet, its lowest in more than twelve years by one of them. Investrade, 9/29 · StockWireX The August trade deficit widened to $105.6 billion on record imports of $420.8 billion, one source. Trading Economics, trade September CPI is Wednesday, October 14, at 8:30 a.m. Eastern; the one consensus this paper could read has the headline up 0.6 percent on the month after 0.4 in August, single-source and not checked against BLS. BLS, CPI · Trading Economics, calendar

The dollar rose, gold fell $152.80 in one session, and Tokyo's intervention ledger reads zero. The dollar index closed at 102.14 on October 8 against 101.29 on September 24, on the cash index and not the December futures contract whose page this paper cited last issue. Yahoo Finance DXY · Investing.com DXY futures December gold settled at $4,168.40 on Monday, September 28, down 3.54 percent, its sharpest fall since June 10 by one note. Investing.com gold · Investrade, 9/29 Thursday's settle is printed both ways because the two vendors disagree and this paper found no third: $4,135.50 on Investing.com, $4,157.00 on Yahoo. Yahoo Finance gold Japan's Ministry of Finance reported zero yen of intervention from August 27 through September 28, which covers the September 19 rate check this paper reported; the dollar has closed between 157.28 and 158.12 yen since September 25, by one vendor. Ministry of Finance, 9/30 · Investing.com yen The Shanghai Composite ended the quarter at 3,842.19, down 6.16 percent, its largest quarterly fall since 2022 by one note, and mainland markets then closed for a holiday; this paper could not establish the reopening date or the first close after it. Investrade, 9/30 · AP via MyNorthwest, 10/2 The Reserve Bank of Australia raised 25 basis points to 4.60 percent on September 29; the ECB and the Bank of England are listed for October 29 and the Bank of Japan for October 30, by one table not checked against the banks. Investing.com central banks

Bitcoin held its range for twelve days and lost it in two, and Ether fell harder. Bitcoin closed between about $83,500 and $86,500 from September 25 through October 6, peaking on Sunday, October 4, and closed October 7 near $83,300. For Thursday, October 8, the sources split again: Coinbase $81,693, Yahoo $81,676, Investing.com $82,442. Ether sat near $2,690 through October 6, closed October 7 near $2,573, and for October 8 reads $2,472.89 on Coinbase, $2,471.91 on Yahoo and $2,413.15 on Investing.com. Coinbase BTC · Yahoo BTC · Investing.com BTC · Coinbase ETH · Yahoo ETH · Investing.com ETH The $2,644.20 this paper printed for Ether on September 24 was not the close on any of them; all three have about $2,687. Spot Bitcoin funds took in money on six of ten sessions and still ended the fortnight a net $324.4 million lighter, because $729.0 million left on October 7 and 8; spot Ether funds have had eight straight outflow days totalling $641.3 million, $201.9 million of it on October 6 from one fund, all by Farside, the only tabulation this paper has. Farside BTC · Farside ETH The reasons on offer are CoinDesk headlines this paper read and did not open: "Ether bets were wiped out at six times bitcoin's rate in crypto's $1 billion flush," and Bitmine ending its Ether purchases. CoinDesk

CLARITY: nothing moved but the odds, and they moved down. The Senate cannot take up H.R. 3633 unless it reverses the 49 to 50 cloture failure of September 15; Senator Tillis's motion to reconsider is entered with no vote scheduled, and the tentative schedule shows a state work period from October 5 to November 6. Kalshi's price for the bill becoming law before January 1, 2027 was about 5 percent on the evening of October 8, against the 7.6 to 8.9 this paper printed for September 24; before November 1, about 1 percent. The fact sheet carries no Polymarket figure later than September 24. DeFi Rate This paper could not reach Congress.gov on the strategic bitcoin reserve bill, H.R. 8957, and has nothing newer than Issue 10's.

Corporate, only from pages this paper opened. Nvidia "announced a historic stock buyback," by AP's September 28 wrap, which gives no size, and Constellation Energy "helped lead the way" on the record day after announcing "a long-term deal to provide Google with electricity." AP via La Nación, 9/28 · AP via La Nación, 10/6 Issue 10 said Micron would report on September 30. This paper could not retrieve the results, Nike's quarter, or the start dates for bank earnings.

Baseball's regular season ended with two 100-win teams, not the one this paper allowed for, and with a division champion at exactly .500; by Wednesday night the two 100-win teams had each other for the National League pennant and the .500 champion had been swept at home.

The National League has produced the series the standings asked for: 103 wins against 100, the two best records in baseball, starting Sunday in Milwaukee per a data feed with no public page. The American League got what the format pays for, which is finishing first in a group of five whatever the group is worth. Houston won its group at 81-81, was given a home series against a White Sox team that had won three more games, and was out in two days. Cleveland, at 85-77, sat out the round that the 93-win Yankees had to play. Chicago, the sixth seed, finished one game behind Cleveland and led this series 2-0 before Cleveland scored eighteen runs in two nights; Saturday's fifth game decides which of them the 98-win Rays get. Plain Text Sports

College football spent two Saturdays lifting Florida thirteen places and dropping it eight, and No. 1 Texas left Knoxville with a three-point win and ten sacks. September 26: No. 1 Texas 20, No. 14 Tennessee 17, with Colin Simmons tying a school record with five of the sacks; No. 21 Florida 52, No. 4 Ole Miss 28; Wisconsin 24, No. 13 Penn State 20 from ten down in the fourth quarter; Wake Forest 30, No. 16 Louisville 27; No. 20 Oregon 41, No. 12 USC 27. KSL / Field Level Media, 9/26 · FOX Sports October 3, from one scoreboard page with no game story opened: Missouri 45, Florida 17 at Missouri, with Florida a 5.5-point favourite at the close; Alabama 56, Mississippi State 23; Ohio State 31, Iowa 14; Pittsburgh 35, Virginia Tech 33, which made Pittsburgh 5-0. ScoresAndOdds The AP poll of October 4, on one source: 1 Texas (61 first-place votes), 2 Georgia (6), 3 Notre Dame, 4 Miami (2), 5 Ohio State, 6 Alabama, 7 Indiana (1), 8 BYU, 9 Ole Miss, 10 LSU, with Missouri up eleven to 14, Florida down eight to 16 and Pittsburgh in at 25. NCAA.com The September 27 poll is reconstructed, not read: it comes from the previous-ranking column of the October 4 poll, which gives Texas, Georgia, Notre Dame, Miami, Ohio State, Indiana, Alabama, Florida, Ole Miss and BYU as the top ten; first-place votes for that week were not found, and the team at 24 could not be identified. NCAA.com On that reconstruction Florida went from 21st to 8th to 16th in fourteen days, which is what beating the fourth-ranked team by 24 and losing to the twenty-fifth by 28 a week later will do.

The NFL after four weeks and one Thursday: three teams unbeaten, four winless, and two of the winless meet on Sunday. The Chiefs, 49ers and Vikings are 4-0; the Chargers, Dolphins, Texans and Titans are 0-4, and Houston plays Tennessee. Plain Text Sports Week 3: Lions 31, Jets 24 in Detroit, the Jets tying it at 24 from fourteen down in the fourth quarter before Jahmyr Gibbs's third touchdown; Chiefs 24, Dolphins 10 in Miami; Bears 27, Eagles 7 on the Monday. Plain Text Sports, Week 3 · Pro Football Reference, Week 3 · AP via Chicago Sun-Times, 9/27 Week 4: Browns 27, Steelers 24 on the Thursday; Panthers 32, Lions 26 on Sunday night; Falcons 45, Saints 24 on Monday; and Vikings 15, Dolphins 10, which means Miami has lost in successive weeks to San Francisco, Kansas City and Minnesota, the three unbeaten teams. Plain Text Sports, Week 4 · Pro Football Reference, Week 4 Thursday night, already played: Buccaneers 24, Cowboys 16 in Dallas, Tampa Bay's first win, behind an undrafted backup quarterback, Jalon Daniels, who went 19 of 25 for 189 yards, and 165 rushing yards from Bucky Irving. NFL.com

Injuries with a number attached; no coaching change found. Tyquan Thornton of Kansas City, dislocated ankle, 12 to 16 weeks. Edgerrin Cooper of Green Bay, torn Achilles, and Paris Johnson Jr. of Arizona, torn biceps, are out for the season; Donovan Jackson of Minnesota, torn meniscus, likely for the rest of the regular season. Rashawn Slater of the Chargers has the kind of high ankle sprain that generally costs four to six weeks; Saquon Barkley, hamstring, week to week. Marcus Mariota of Washington has an MCL sprain and no return date, and Caleb Williams of Chicago will miss a third straight start. NFL.com, 10/5 By headline only, the article unopened: Lamar Jackson has an ankle sprain and an outside chance to play Sunday. NFL.com, 10/6 No head-coaching change appears on NFL.com's news index as captured October 8 and 9; that is one index page, not a search. NFL.com

Golf, and the rest of the world at a sentence or two each. The United States won the Presidents Cup 17 to 13 at Medinah, its eleventh in a row, after losing all five Friday foursomes and going into Sunday down 10.5 to 7.5; it took 9.5 of the twelve singles points. Sky Sports, 9/27 · PresidentsCup.com, 9/27 Formula 1 raced twice, on one source: George Russell won in Azerbaijan on September 26 and Max Verstappen in Bahrain on October 4. Formula1.com Europe won the Laver Cup 13 to 5 in London, Alexander Zverev clinching. LaverCup.com, 9/27 In the WNBA semifinals the Dream lead the Liberty 2-0 and the Valkyries lead the Aces 2-0, with third games tonight; New York got there as the eighth seed by sweeping top-seeded Minnesota. Plain Text Sports Hockey is back, on this paper's reading of a schedule page that begins Tuesday September 29 and does not separate preseason from regular season; no league page was opened. Plain Text Sports Not obtained: the Berlin Marathon result of September 27. The organiser's front page trails a world record attempt by Tigst Assefa without saying how it ended, and its results portal refuses automated readers. BMW Berlin-Marathon No soccer or cricket result came from an opened page this fortnight, and the Anthem says so.

The three picks live at last issue finished Texas 20, Tennessee 17; Lions 31, Jets 24 in Detroit; and Chiefs 24, Dolphins 10 in Miami, the last two at venues Issue 10's pick lines had backwards; all three are graded in the Gridiron Line below, outside the record. KSL / Field Level Media, 9/26 · AP via Chicago Sun-Times, 9/27 · Chiefs.com

Six open this fortnight. They carry seven days, then enter the ledger. These are the first written to the rules the Season Scorecard proposes above: each names the page that will settle it, each kill is its call turned inside out, each was checked back past fourteen days for a fact that already answers it, and the two logged near a coin flip say why they are not a pass.

No call: where Isaias lands and how hard. It is forecast to come ashore tonight or early Saturday. This paper does not log calls inside its own seven-day carry, and does not price a landfall with people under it.

No call: Brent on October 30. The December contract expires that day and this paper now knows what an expiry day does to a careless call. It could write the careful one. It has no confident read on the price, and a call written to show off a rule is a stunt.

No call: the midterms. This paper has never priced who wins an election and is not starting twenty-five days out, on a fortnight in which much of its research ran without search.

No call: Sweden's first vote. October 14 is five days away, and call #40 already holds the question that matters.

THE GRIDIRON LINE

Kept off the official record, tallied separately. The last weekend this paper picked: one for three.

G7, Tennessee +4.5 vs Texas: COVERED. Texas 20, Tennessee 17 in Knoxville. Lost by three with four and a half in hand, in a game where Texas recorded ten sacks. KSL / Field Level Media, 9/26 G8, Jets +6.5 at Lions: LOST. Lions 31, Jets 24 in Detroit. The Jets came from fourteen down to tie it at 24 in the fourth quarter and lost the cover on Jahmyr Gibbs's third touchdown inside the last two and a half minutes. Half a point. AP via Chicago Sun-Times, 9/27 · Pro Football Reference G9, Dolphins +11.5 vs Chiefs: LOST. Chiefs 24, Dolphins 10 in Miami. Inside the number until a Travis Kelce touchdown with 2:55 left turned a seven-point game into fourteen. Pro Football Reference · Chiefs.com

Last issue's pick lines had both NFL venues backwards: the Jets were in Detroit and the Chiefs were in Miami. A spread is a spread and the grades do not change; the error is in Corrections. Nine picks in, the sample is five and four. Both NFL picks died on touchdowns scored in the last three minutes, which is what a coin does. No picks exist for the weekend of October 3 to 5, because no issue was published, and a pick written after the game is a story.

This weekend, lines as read Friday morning, October 9; they move, and where two pages disagree both are printed.

G10. South Carolina +11.5 at Florida, Saturday, 12:45 p.m. Eastern. Opened Florida minus 14 at one book and minus 14.5 in the consensus column, and has come about three points toward South Carolina with 65 percent of bets and 76 percent of the money on Florida and no injury attached on the pages read. SI, 10/4 · ScoresAndOdds, 10/9 Florida beat Ole Miss by 24 and then lost at Missouri by 28 as a 5.5-point favourite. ScoresAndOdds A line that falls three points against three quarters of the money, on a team that has just shown both of its faces, is the shape this section keeps buying. Taking the points.

G11. Nebraska +7.5 vs Indiana, Saturday, noon Eastern, Lincoln. Opened Indiana minus 9.5 and sits at minus 7.5 with 63 percent of bets and 64 percent of the money on Indiana, no injury attached. Indiana is seventh in the poll and 5-0; Nebraska is 5-0 too, on one page. ScoresAndOdds, 10/9 · VegasInsider, 10/8 Two points toward an unbeaten home team that the public is betting against. Taking the home team and the points.

G12. Jets minus 2.5 vs Browns, Sunday, 1 p.m. Eastern. One book opened the Jets at minus 1.5 and had them at minus 2.5 by Thursday afternoon with 84 percent of tickets and 90 percent of the money on Cleveland; a second page still shows minus 1.5, and the pick is graded at 2.5, the worse number. Cleveland is 3-1 and has won three straight. The Jets are 1-3. VegasInsider, 10/9 · ScoresAndOdds, 10/8 A book that moves toward a 1-3 team against nine dollars in ten is wrong or informed, and this section exists to find out which. This paper has now lost a Jets game from each side. Laying the points.

Same disclaimer, printed every time: judgment calls with stated reasoning, not model output, settled this weekend, never entering the official record.

Gridiron record: 5-4.

Same rule as the rest of the paper, pointed at people instead of markets. Fact stated as fact. Disputed labeled disputed. A link on every line. If it doesn't clear the bar, it doesn't run. The bar: a named fact-checking desk rates the claim flat False or Pants on Fire, or Full Fact prints "False" or "Incorrect" in its verdict box. Mostly False, "misleading", a prose verdict in a debate roundup and a yes-or-no fact brief do not clear it.

WinSenate, a Democratic-aligned super PAC, campaign ad against Senator Susan Collins of Maine, 9/24, fact-checked 9/28. The ad said Collins "voted to cover up the Epstein files" and that Jeffrey Epstein's "business partner funneled $3 million" to her super PAC. PolitiFact: False. The vote is real: on September 10, 2025 Collins voted to table an amendment releasing the files, 51 to 49, saying it did not belong on a defence policy bill. In November 2025 a bill releasing them passed the House 427 to 1 and the Senate by unanimous consent. The $3 million is two donations from Blackstone chief executive Stephen Schwarzman to a pro-Collins PAC, and the partnership amounts to both men having been early members of one New York club, where Epstein was a founding member "alongside 150 other investors". PolitiFact "found no evidence they were business partners." PolitiFact, 9/28 A second desk read the same ad and used softer words: FactCheck.org headlined its piece "Democratic Ads Misleadingly Connect Susan Collins to Epstein" and wrote that the ad "stretches the facts", without calling it false. FactCheck.org, 10/1 Both are printed; the entry runs on PolitiFact's. Issue 9 dropped a Republican item on a split verdict in which neither desk said False; here one desk's meter does, and that is the bar.

President Donald Trump, a campaign post on X and two posts on Truth Social, 9/21 to 10/4, fact-checked 10/2 and 10/5, three times. His campaign account asked, "Where was Roy Cooper when Helene slammed into North Carolina?" and answered, "In New York speaking at a New York Times 'Climate Forward' event." PolitiFact: False. Cooper, then governor, was in New York on September 25, 2024, back by 3 p.m., declared a state of emergency that day, and was in Raleigh when the storm hit the state on September 27. PolitiFact, 10/5 On October 1 he posted, "I stated, numerous times, that it would take 4-6 weeks to get rid of THE IRAN NUCLEAR THREAT, and I did it in one night!" PolitiFact: False. In March the number was his and it was about the war: he told the New York Times on March 1, "We intended four to five weeks." PolitiFact found he "did not distinguish between a timeline to get rid of the nuclear threat and the overall war" and that experts say the strikes set Iran's nuclear programme back without eliminating it. PolitiFact, 10/2 On October 4 he posted that Brazil's first-round result came out "on the SAME DAY!" while in Detroit, Philadelphia and California the results, "with much smaller numbers, will often take weeks to Rig, I mean, calculate." PolitiFact: Pants on Fire. Brazil has no mail-in voting; it votes in person on electronic machines it has used since 1996. American states write their own election law, and California's counts ballots postmarked by Election Day and received up to seven days later. PolitiFact, 10/5

Markwayne Mullin, Homeland Security secretary and a former US senator, warm-up speech at a rally in Durant, Oklahoma, 10/1, fact-checked 10/5. He said of the President, "There's no financial gain like the past presidents have had. It's cost him to be in office." PolitiFact: Pants on Fire. The President's own financial disclosure, a 927-page form, shows at least $2.2 billion in revenue in 2025, up from $622 million in 2024, at least $1.4 billion of it from cryptocurrency. Forbes estimates he is $2.7 billion richer than when he won the 2024 election and puts his net worth at $7 billion. The salary forgone is $400,000 a year; PolitiFact's line is that the comment "is off by more than $1 billion." PolitiFact, 10/5

The Senate Leadership Fund, Republican-aligned, campaign ad in Ohio's Senate race, 9/16, fact-checked 10/1. The ad said, "Forcing childhood sex change surgery. But Sherrod Brown stands with these extremists." PolitiFact: False. The ad's source is a 2023 television interview in which Brown said "A child's health care decisions are between them, their parents, their families, their doctors" and not politicians; PolitiFact "found no evidence he has ever made such a statement" as the one the ad gives him. What he opposed was a bill backed by then-Senator JD Vance to make it a felony for doctors to provide surgery or puberty blockers to anyone under 18. In Ohio "a 2024 state law bans physicians from performing gender-reassignment surgery on a minor." PolitiFact, 10/1

John E. Sununu, Republican Senate candidate in New Hampshire and a former US senator, primary-night victory speech, 9/8, fact-checked 10/1. He said, "Chris Pappas voted to raise taxes 68 times," and "Every tax vote he's made was to raise taxes." PolitiFact: False. The campaign's list cites 16 distinct congressional bills, and "the 68-vote figure is padded by procedural votes." Eleven of the votes are on the One Big Beautiful Bill Act and a precursor measure; others are on bills that passed 414 to 5, 357 to 70 and 417 to 3. Pappas "also voted on numerous other occasions to lower taxes, and he sponsored bills to cut taxes." PolitiFact, 10/1

TikTok posts, no party, 9/28, fact-checked 10/1. PolitiFact states the claim as "These videos show scenes of alleged sexual assault at Cornell University." Its rating: Pants on Fire. One clip shows a man accused in 2024 of assaulting a teenage girl in San Antonio; another is "from a 2019 assault in California captured on a Ring doorbell surveillance video"; others had circulated "as early as November 2025." The case they were attached to is a civil lawsuit filed September 16 in New York, and no video of it has been released by officials or in credible news reports. PolitiFact, 10/1

Yvette Cooper, MP and UK health secretary, Labour Party conference, the week before the check, fact-checked 10/5. Cooper told the conference that Reform UK's "only health policy at the last election was a £1.7 billion tax break for private healthcare companies"; she did not name the party, and Full Fact judged that she meant it. Full Fact's verdict box: False. Reform's 2024 manifesto gave a full page to health, including "All frontline NHS and social care staff to pay zero basic rate tax for three years", vouchers for private treatment and weekend operating theatres. The tax break is in it, as "tax relief of 20% on all private healthcare and insurance"; the £1.7 billion figure is not, and Full Fact did not verify it. Full Fact, 10/5

The UK Department for Transport and its secretary, Heidi Alexander, a department statement and a Labour conference speech, fact-checked 10/2, corrected after Full Fact asked. The department said "over 14.5 million children used the bus for free this summer"; Alexander told the conference of "free bus travel for kids this August, used by 14 million children". Full Fact's verdict box: Incorrect. The figure counts journeys, not children: England has "only about 7.5 million children" aged 5 to 15, the group the offer covered. After Full Fact queried it the department changed its statement to "Over 14.5 million free bus journeys were taken by children this summer." Full Fact, 10/2

On this fortnight's balance, stated rather than engineered, and on the search that did not run. The standing instruction is twelve or more searches a side, by name, with the log printed. This fortnight's log is two queries attempted, none performed: "fact check Gavin Newsom claim false October 2026" and "fact check Trump false claim PolitiFact October 2026" were both refused, because the session's shared budget of 200 searches had been spent before this desk's turn. The by-name passes did not run. What stood in for them was a read of the fact-checkers' own pages, 74 fetches of which 55 returned something usable: both major American desks complete, item by item; every Democratic-side item on their lists opened, five of five; every flat-False Republican item read in full, and the six Republican items not opened rated Mostly False, Half True, True or Full Flop; and six debate roundups read for the Democrat and the Republican in each. That is a by-name pass over what was published, not a search. PolitiFact's complete list from September 25 to press time, picked up where Issue 10's stopped, is sixteen items. October 7: Phil Berger, North Carolina Republican, Full Flop on the Flip-O-Meter, not a truth rating PolitiFact, 10/7; Zach Nunn, Iowa Republican, Mostly False PolitiFact, 10/7; the Senate Leadership Fund on Josh Turek, Half True PolitiFact, 10/7. October 5: Mullin, Pants on Fire; Trump, Pants on Fire and False. October 2: WinSenate on Ashley Hinson, Mostly False PolitiFact, 10/2; Trump, False. October 1: TikTok posts, Pants on Fire; the Senate Leadership Fund on Brown, False; Sununu, False; Byron Donalds, Florida Republican, Mostly False PolitiFact, 10/1; Mike Rogers, Michigan Republican, Mostly False PolitiFact, 10/1. September 30: Ashley Hinson, Iowa Republican, True PolitiFact, 9/30. September 28: Rogers again, Mostly False PolitiFact, 9/28; WinSenate on Collins, False. The eight that ran are linked in their entries. Berger's rating and the Half True were read off PolitiFact's front page only. FactCheck.org's eight articles: Republican ads on Abdul El-Sayed FactCheck.org, 10/8; Republican ads on taxes on tips and overtime FactCheck.org, 10/8; Gavin Newsom's press office, Adam Schiff and Bernie Sanders on the President's "take out" remarks FactCheck.org, 10/6; the President on Medicaid drug savings FactCheck.org, 10/5; WinSenate on Collins, October 1, linked above; the White House's taxpayer-funded ads FactCheck.org, 9/29; the Talarico campaign, the Democratic National Committee and Texas Democrats on a clipped Ken Paxton remark FactCheck.org, 9/25; a Republican super PAC's rape-kit ad against Roy Cooper FactCheck.org, 9/25. Five examine Republican claims and three Democratic-side ones, and the word false appears in none of the eight as FactCheck.org's own verdict on a claim.

The paper spends most of its column inches on wars, borders, and people lying into microphones. That's the job. This is the other part of the record.

Wyandotte County, Kansas, 10/2, reported 10/3. A storm put a tree through Gilbert Kirk's roof and trapped him and his dog for hours. On Wednesday, September 30, he told KCTV5 that all he wanted was a tarp. Two days later a stranger walked into his yard and promised to be back in about thirty minutes with a tarp and a crew. Kirk: "OK, whatever, you know. And then the next thing I know, he's back with a load of lumber." Trucks followed, and the roof is covered. KCTV5, 10/3 Good News Network names the crew from a Facebook post on the station's page: Kirk Mata, Steven Vega, Dominic Foreman and Brigette Henson. Vega, as it quotes him: "We just did what was right and needed to be done." Good News Network, 10/7 Neither outlet dates the storm.

Near Spearfish, South Dakota, 10/1, reported 10/2. Sharon and Scott Jacobs were driving to their gallery in Deadwood when Scott saw something on a fence. "I think that was a hawk stuck on the barbed wire," he told her, and he did not think it was alive. Sharon, by her own account: "Excuse me, do you forget who your wife is?" It was a great horned owl, impaled and alive. A passing driver named Brandon had wire cutters; a young woman phoned her father about the fence, and he said to cut it. "It's OK, buddy," Sharon told the owl. "I'm here to help you. We're here to help you." Nothing appeared to be broken, and the bird is with the Black Hills Raptor Center, to be released if all goes well. Cowboy State Daily, 10/2 One source.

Killeen and Fort Hood, Texas, reported 10/6. Elizabeth Laird was known around Fort Hood as the Hug Lady. For years she embraced soldiers as they deployed and as they came home: hundreds of thousands of hugs by the station's text, 500,000 by its headline. She died in 2015. Her daughter, Susan Taylor, is looking for the soldiers who deployed through Fort Hood between 2003 and 2015 and got one of those hugs. She wants their memories, photographs and stories, possibly for a second book. Taylor: "I didn't want her legacy to die." Retired Command Sgt. Maj. Calvin Coler: "She was there when we left, when we deployed, and when we came back she was always at the airport with her hands open like this, ready to give a hug." Veterans can find Taylor's address on the station's page. KWTX, 10/6 One source.

Mount Prevost, British Columbia, 2/6, reported 9/23 and 9/28. Matthew Clark, 24, an electrician from Victoria, had finished his own flight when the radio said another pilot was down. Stephen Parson, 75, hung semiconscious in his harness in a tree in a vertical gully, about two hundred feet from the top, and Cowichan Search and Rescue's rope and ground teams could not reach him. Clark went up without gear. The Carnegie synopsis has him free-climbing a narrow chimney, "placing his back against a slippery, mossy side and his feet against the other for optimum traction," then cutting Parson out of the harness and giving first aid until North Shore Rescue's helicopter came. Parson spent eight days in hospital and fully recovered. Clark is among sixteen new Carnegie Medal recipients. Victoria News, 9/23 Victoria News says paraglider in its headline and hang glider in its text; this paper prints both.

Maricopa, Arizona, reported 10/8. Jordan Hope, 18, saw an older man in a Walgreens parking lot with his shoes barely hanging on to his feet, took off his own and gave them to him. He did not know anyone was watching. On learning the man's wife had died two weeks earlier, he got him lunch, stayed to talk, and has phoned him every day since. Hope: "I just think he needed a friend. It's hard asking for help sometimes." KPHO via KCTV5, 10/8 One source family: KPHO and a sister station.

Every week produces at least one story with no read, no call, and no explanation. It goes here, filed without a take. A fortnight produces a surplus.

Dunedin, Florida, and Bermuda, 9/18, reported 10/9. In 2018 a bench went up on the Dunedin causeway in memory of Medhat "Mike" Sadiku, who died in 2012. His daughter Amanda: "My dad always said, 'It'll be alright.' So we thought that would be perfect to put on the bench." Hurricane Helene tore it off its concrete slab in 2024, and it was marked missing. On September 18 of this year, the family says, a woman in Bermuda got in touch: she had found it on the coast there, more than a thousand miles away. Bay News 9 meteorologist Kyle Hanson thinks Hurricane Milton may have pushed it into the Gulf, where the loop current and the Gulf Stream could have carried it: "The chance of this happening is lower than 1 in 100, I would say." The bench is still in Bermuda, and Amanda Sadiku is deciding whether to travel to it or have it shipped back. Bay News 9 via KCTV5, 10/9 One source.

Spring Hill, Kansas, 10/7. Tasslehoff is a pet iguana, nine years old, about three feet long and an estimated 30 to 40 pounds. After a week on the loose it was found shortly after 6 p.m., about forty feet up a tree not far from the Spring Hill Vet Clinic, and Johnson County Consolidated Fire District 1 brought it down with a catch pole lent by Spring Hill police. The station, the one source opened, does not say how it got out. KCTV5, 10/8 It had company. By UPI headline and web address only, because the pages would not open, the same week's file carried a kangaroo that spent a day out of a Missouri petting zoo, a serval in Poland that went home after days on the loose, and a pet boa constrictor missing in a North Carolina neighbourhood. UPI, 10/6 · UPI, 10/6 · UPI, 10/7

Apuiarés, Brazil, and Grafton, Australia, 10/5. João Marinho Neto, the world's oldest man, turned 114 on October 5 in Apuiarés, Ceará. Ken Weeks, listed as the world's second-oldest man, turned 113 the same day and marked it with a tour of Grafton, New South Wales. The one source, a listing, prints no birth years; by the arithmetic the two were born a year apart to the day. LongeviQuest, 10/5 UPI's headline has the two exchanging cards; that page would not open, so the cards rest on a headline alone. UPI, 10/5

University of New England, Maine, reported 10/4. Casper is a white lobster, a colouring put at 1 in 100 million. The Maine State Aquarium in Boothbay has donated it to the university, where its DNA joins a study of rare lobster colours that is working through more than 15,000 genes and, the university says, does not harm the animal. Good News Network, 10/4 One source.

Stories this paper has led with, that no longer lead, and that are not finished. They get a paragraph and a number here until they resolve, and this time each number covers a fortnight, because this paper did not publish on October 2. Nothing leaves this paper by being forgotten.

DR Congo Ebola: WHO's count is 8,728 confirmed and 4,205 dead, the virus reached Nairobi on an international flight, and call #46 enters the table with its own baseline corrected. WHO's notice of October 8: "As of 6 October 2026, the country has recorded 8728 confirmed cases and 4205 confirmed deaths", a crude case fatality ratio of 48.2 percent. Ituri "remains the epicentre, with 6480 cumulative confirmed cases and 2989 deaths"; North Kivu has 1,755 cases and the highest provincial ratio, 57.7 percent; 48 of the 64 affected health zones, in seven provinces, had a new case in the preceding 21 days; the summary is that "In some areas, there are clear signs that the epidemic is slowing down, although in others it continues to grow"; and since the last notice "264 of the 372 reported deaths occurred in communities and 108 in treatment centres", which is 71 percent outside care. WHO DON619, 10/8 The count is up from WHO's 7,890 as of September 23 and is the one call #46 reads. WHO DON618, 9/25 Kept apart from it: the DRC public-health institute's own report had 8,116 as of September 27. UKHSA, 10/1 ECDC carries the same national total and a different province table, Ituri 6,513 and North Kivu 1,785 on data to October 6; the two tables are not reconciled here. ECDC, 10/8

Kenya is the new country. The patient went by road from the DRC to Kampala, flew to Nairobi on October 3 and "was hospitalized and treated in Kenya but subsequently died on 5 October"; WHO's regional office calls Kenya "the fourth country affected during the current outbreak". WHO DON619, 10/8 · WHO AFRO, Report 21 By Anadolu's account of the health ministry's October 6 announcement he was a Kenyan citizen, ill for about a month, who passed routine airport screening off a flight carrying 23 passengers. Anadolu, 10/6 Europe has its fourth imported case, a patient "medically evacuated to the Netherlands for treatment", and UKHSA's line on all four is that they "had been working on the response to the Ebola outbreak." ECDC, 10/1 · UKHSA, 10/8 The vaccine trial has reached 6,300 health workers, with 1,066 other people given doses, on one news service's relay of Africa CDC, against the 3,771 people this paper printed for September 17; no trial result was found. CIDRAP The $2.9 billion is still a pledge: Africa CDC's communiqué of October 4 says "every dollar must be traceable from the donor's commitment to the services delivered to communities" and gives no figure for cash received or spent. Africa CDC, 10/4

Call #46 (65 percent, settles December 31) enters the table OPEN, tracking YES: 8,728 as of October 6, with 83 days left. The arithmetic, in words. The call needs WHO's published count above 10,000, which is 1,273 more than 8,728. WHO's own pace is in the notice, "new weekly cases nationwide have increased by 2.5% from 475 to 487", and 1,273 divided by 487 is 2.6 weeks, which from October 6 is about October 24. For the call to miss, the count must add fewer than 1,273 in the 86 days from October 6 to December 31, an average under 104 a week, about a fifth of WHO's figure. WHO DON619, 10/8 Africa CDC disagrees with WHO on the pace, and both are printed: its briefing, as relayed by CIDRAP, had 278 infections confirmed "near the end of September", the lowest week since mid-June, where WHO has 475 and then 487, and its own epidemiologist said "This curve doesn't mean that the outbreak is totally under control" and that the fall may reflect underreporting; by province it had new cases down 32 percent in Ituri and climbing in Tshopo and North Kivu. CIDRAP At 278 a week the line is 4.6 weeks from October 6, the first week of November. The call clears on either agency's pace. What it is exposed to is the ledger, not the virus: ECDC notes the DRC's figures are "under continuous review and harmonisation", and the call reads only what WHO publishes. ECDC, 10/8

The correction this paper owes on that call: its parenthetical, and this section last issue, said WHO had 7,773 confirmed on September 23, and that figure is dated September 21 and is the DRC public-health institute's count as relayed by the UK Health Security Agency. UKHSA, 9/24 WHO's own figure for September 23, published the day Issue 10 went to press, is 7,890; the threshold and the call's terms do not change. WHO DON618, 9/25

US measles: 3,887 cases, Pennsylvania past 1,000 with five dead, the federal count of deaths at two, and 31 members of Congress asking in writing about the gap. CDC's page, read at 7:01 a.m. Central on October 9, had not posted the count for data as of October 8; it still read "As of October 1, 2026", updated October 2. That figure is 3,887 confirmed cases, 42 outbreaks, 95 percent outbreak-associated, 47 jurisdictions and two deaths, up from 3,471 as of September 17 by way of 3,659 at the update between; the latest week added 228, the eighth straight in three digits by CIDRAP's count, and CIDRAP printed the total one lower, at 3,886. CDC, 10/2 · CIDRAP · CIDRAP Pennsylvania's health department counted 943 cases in 39 counties on September 30, the day it confirmed a fifth death ("An unvaccinated Lancaster County resident with confirmed measles has died."), then 1,004 on October 5 and 1,078 in 41 counties on October 7, with 207 hospitalised, against 835 when this paper last printed. The October 7 release says of its 74 new cases that "Today's update does not reflect a current surge in new infections", more than 40 being people who fell sick over two weeks earlier; it is the latest release, the next address in the series having failed to open on the one attempt this paper made on October 9. PA DOH, 9/30 · PA DOH, 10/5 · PA DOH, 10/7

State five, federal two. CDC's count of deaths went from one to two during the fortnight, on a day this paper could not pin, and CIDRAP's sentence on it is "It is unclear which deaths the CDC's listing is referring to." CIDRAP By one outlet's account of a letter this paper did not open, Senator Michael Bennet, Representative Kim Schrier and 29 Democratic colleagues wrote to CDC director Erica Schwartz, in a letter dated October 6 by its filename, that if the agency "is altering verified state data to fit political preferences," then "the pledge you made at your confirmation hearing has already been broken." They want written answers by October 16. CIDRAP New York's governor declared a State Disaster Emergency on October 5 on 108 cases this year as of October 3, 92 of them since July 15 in under-immunised rural communities; the order lets paramedics, pharmacists and midwives give the vaccine. Governor of New York, 10/5 Utah is a conflict, printed and not settled: the state's page, updated October 6, shows 724 residents diagnosed across two years, none in the last three weeks, and no declaration that the outbreak is over; CIDRAP's roundup of the October 2 federal update has the state listing 527 for the year with 9 new and the CDC map 535 with 8 new; this paper's last figure was 518. Utah DHHS, 10/6 · CIDRAP

Call #32 (55 percent, settles about November 30) is OPEN, tracking YES, with no meeting date published and 52 days left. PAHO's measles page lists no 2026 meeting of the commission, only last year's, in Mexico City on November 4 to 7, 2025. PAHO PAHO's regional report of October 8 counts 55,792 confirmed cases in 17 countries and territories through October 3 and 77 deaths, two of them in the United States, which is the federal number, and does not mention a review of the United States. For the United States it marks the observed trend as declining, warns that recent weeks are incomplete, and says "model estimates indicate increasing transmission": a reproduction number of 1.4 (interval 1.33 to 1.48) and about 2,480 cases forecast over four weeks, which the table heads as epidemiological weeks 40 to 43 and a footnote gives as 38 to 41. PAHO, 10/8 The only timing statement found is second-hand: CIDRAP reports that Paul Offit of Children's Hospital of Philadelphia and other experts "predict the United States will lose its elimination status next month", when PAHO meets, and quotes him: "Measles is now endemic in the United States". CIDRAP That is an outside expert's forecast in a news story, not a date from the body that rules, and the call's kill includes a ruling delayed past November. Last issue this paper said a commission reading the federal number would read a smaller outbreak than the one Pennsylvania is living in. PAHO's own report now prints two.

Nepal and Tibet, weeks six and seven: the disaster authority says 5,285 missing, the police figure this paper printed last issue was 6,150, and no page explains the 865 between them. All of it this fortnight is from one newspaper. The National Disaster Risk Reduction and Management Authority's count as of September 26 is 1,453 dead, 5,285 missing and 13,795 rescued, and Bagmati province's preliminary report, handed over October 8, "estimates that 1,455 people died and another 5,285 remain missing." Kathmandu Post, 10/6 · Kathmandu Post, 10/8 These do not match the Nepal Police figures this paper printed last issue, 1,451 dead and 6,150 missing as of September 20 and 21. The dead agree to within four; the missing do not; no page this paper found explains the difference, and it found no police count dated inside the fortnight. The move from rescue to recovery has an official's name on it and no date. The authority's chief executive, Dharmaraj Upreti, said in an interview published September 29 that the operation is moving from "search and rescue" towards "search and recovery", that a committee "will meet next Wednesday under the chairmanship of the home minister" to decide when the rescue phase formally ends, and that "there may still be people or remains beneath houses and mud." This paper did not find the outcome of that meeting. He also said the government is moving to declare the missing, "currently around 5,285 people", dead for relief purposes, which needs a change in law. Kathmandu Post, 9/29 The figure going into that law is the lower one.

Reconstruction has converged and damage has not. Upreti: "reconstruction will require $4.77 billion, or more than Rs700 billion"; the World Bank's update of October 6 carries Nepal's estimate of needs at approximately $4.8 billion; last issue's rapid assessment was $4.74 billion, and the figure of nearly $7 billion this paper also printed has not reappeared in anything it opened. Damage is three numbers: about $3 billion to infrastructure on Upreti's preliminary figure, $1.81 billion in physical damage on the government's estimate as the Bank reports it, and $1.66 billion on the Bank's own. Kathmandu Post, 9/29 · Kathmandu Post, 10/6 Not found, and named: a site count from Upper Trishuli-1, where 439 were unaccounted for last issue; the number of bodies identified; the UN appeal's funding; the roughly 70 Americans; any toll from the Tibet side.

The World Trade Center Health Program. Nothing dated in the window on the two pages this paper checked. The program's own site shows its last update as September 11, and Fire Rescue 1's 9/11 page shows nothing after September 20. CDC · Fire Rescue 1 Two pages are not a sweep: Congress and the New York press were not searched, and this is a statement about what this paper read.

Elsewhere, labelled and uncalled. Sudan first, because this paper set itself a question and cannot answer it. Last issue it printed the UN migration agency's September 30 stock-out date so that it could say what happened. It looked: the agency's news pages, UN News and ReliefWeb refused the fetch or returned stale pages, and nothing this paper opened says whether the shelter and sanitation stocks ran out. What it could open, one source: WHO, by way of Radio Dabanga, "requires USD 97.7 million for its 2026 health emergency response in Sudan but has received only USD 22 million". Radio Dabanga, 10/8 The September 30 question stays on this page until it is answered. Southeast Asian haze, by headline only, from one listing page with no article opened: 38 areas of Malaysia recorded unhealthy air at 11 a.m. on October 8 and 36 by 7 p.m., schools in Melaka and Negri Sembilan were to close on October 9, and visibility in Palembang fell to 400 metres. No official count of respiratory cases was found. The Star, haze listing

Friday 10/9. Hurricane Isaias, a major hurricane since 7:20 this morning, is forecast to make landfall on the northern Gulf Coast tonight or early Saturday, inside a warning area that runs from Ocean Springs, Mississippi to the Bay and Gulf county line in Florida. NHC Update, 10/9 · NHC Advisory 11A, 10/9 Comments close on DOE's bulk-power request for information. Federal Register, 9/9 WNBA semifinal third games. Plain Text Sports
Saturday 10/10. Isaias forecast inland as a hurricane in the small hours and post-tropical by afternoon, with rain and a tornado risk across the Southeast through Sunday. NHC Discussion 11, 10/9 Hurricane Simon nears southwestern Mexico, forecast to be a Category 4. NHC Simon Advisory 8, 10/9 Indiana at Nebraska at noon Eastern and South Carolina at Florida at 12:45, gridiron picks G11 and G10; Texas against Oklahoma at 3:30; Georgia at Alabama at 7:30. Bleacher Nation, 10/7 White Sox at Guardians, fifth game, 8 p.m. Eastern. Plain Text Sports Taiwan's National Day. Focus Taiwan, 10/8
Sunday 10/11. NFL Week 5: Browns at Jets at 1 p.m. Eastern, pick G12; 49ers at Seahawks; Ravens at Falcons at night. Plain Text Sports
Monday 10/12. Magdalena Andersson's report to the Speaker is due. Riksdagen, 10/2 A federal holiday, which pushes the EIA's weekly petroleum report to Thursday. EIA
Wednesday 10/14. September CPI at 8:30 Eastern, the number most likely to move call #51. BLS The earliest day reported for a prime-minister vote in the Riksdag, on one newspaper's account. EFN, 10/7 About now, Microsoft's review of Azure incident 7Q30-010, and with it the ruling on call #6. Azure Status History Taiwan's Lu Sheng brigade exercise begins and runs to October 23. CNA via GlobalSecurity, 9/6
Thursday 10/15. TSMC's third-quarter earnings at 1 a.m. Central; call #7. TSMC investor calendar FERC's open meeting, agenda unread. FERC, 10/8 The Drought Monitor, weekly. Drought.gov
Friday 10/16. Italy's notified controls on its borders with Spain end or renew. European Commission The date by which 31 members of Congress asked the CDC director for written answers on the measles death count. CIDRAP
Tuesday 10/20. The SEC's Regulation Crypto Assets comment deadline, as carried from Issue 10 and not re-verified this fortnight. Federal Register, 8/21
Thursday 10/22. Lloyd's List's next Hormuz brief, the first reading against call #9's line since the surge in tanker attacks. Lloyd's List Intelligence, 10/8 PJM's special Members Committee meeting on the revised backstop, by one trade report. Utility Dive, 10/8 Spain's controls on arrivals from Italy end at midnight unless renewed. ESdiario, 10/6 PAHO's next regional measles report. PAHO, 10/8
About Saturday 10/24. WHO's confirmed Ebola count in DR Congo passes 10,000 if 487 a week holds; this paper's arithmetic, not a scheduled event, and publication would follow by days. Call #46. WHO DON619, 10/8
Sunday 10/25. Brazil's presidential runoff. PolitiFact, 10/5
Tuesday 10/27 to Wednesday 10/28. FOMC, decision Wednesday at 2 p.m. Eastern. Call #51 settles, and it is the first of call #39's two chances. Federal Reserve
Thursday 10/29. PJM's planned date for a revised backstop filing; call #52 settles. Utility Dive, 10/8 GDP and PCE from the BEA. BEA The ECB and the Bank of England, by one table not checked against the banks. Investing.com central banks
Friday 10/30. The last trading day of ICE's December Brent contract, which is the next day "Brent settles" has two answers. ICE
Saturday 10/31. France's current internal-border period ends and the next begins November 1, Spanish border included. European Commission
Sunday 11/1. The seven OPEC+ countries meet. OPEC, 10/4
Tuesday 11/3. US midterm elections. Call #45 settles.
Friday 11/6. The October employment report. BLS
Monday 11/9. The Senate reconvenes at 3 p.m. Senate Democrats, 9/30 COP31 opens in Antalya, running to November 20. UNFCCC
Tuesday 11/10. October CPI. BLS The original expiry of the US and China trade truce, said by officials on both sides to run to January 10, 2027, with no American instrument found; the day a pause that survives has to be written down somewhere. Reuters via Internazionale, 9/28
Thursday 11/12. The Climate Prediction Center's next El Niño discussion. CPC
November. PAHO's review of US measles elimination status, no date published; call #32. PAHO Anthropic's reported target for its debut; call #49. Investing.com, 9/18
Monday 11/30. Calls #26 (Venezuela), #40 (Swedish PM), #48 (one hurricane), #49 (Anthropic on EDGAR), #50 (measles above 5,000) and #53 (the arena test) settle, and #32 about then. The Atlantic season ends. NHC season table
December. Call #29 (Sudan) by December 1 and #33 (Guinea-Bissau) on December 6. The FOMC on December 8 and 9, call #39's last chance. Kalshi Sudan's sanctions regime and federal funding under H.R. 6500 both expire December 11. Darfur24, 10/9 · AHA, 9/1 PJM's 2029/30 auction December 9 to 15, results December 22, call #17. FERC order, 9/29 DOE's regulations due December 24, call #30. And on December 31: calls #6, #31, #36, #39, #42, #43, #44, #46 and #47.
Sunday 1/10/2027. The date both governments' officials give for the end of the extended trade truce. Reuters via Internazionale, 9/28 Call #5 settles in mid-January.

Every official call this paper has made, running week to week. This table appears in every issue until the calls settle. No call leaves this table by being forgotten.

Calls #48 through #53 carry seven days and enter this table next issue. The gridiron line is tallied separately and never enters it. Row 10 is shown as graded in Issue 7; the Season Scorecard prints what its kill did and the ruling that the grade stands.

Two weeks of checking against one issue of claims, plus a season audit, makes the longest list this paper has printed. It is sorted by kind, worst first.

Corrected: the Navier-Stokes proof had been compiled by outsiders, in public, on the day it was released. Issue 9 printed "nobody has compiled it" and "No independent verification published between September 11 and 18." Issue 10 printed "still uncompiled by anyone outside the building." Two outside builds are dated September 8, one of them announced on the Lean community's public forum that evening; a named mathematician vouched for the statement on September 11; and a multi-checker replay was committed on September 17. Rodrigo González, 9/8 · Lean Zulip, 9/8 · arena commit, 9/17 All three sentences were wrong when printed. The full record, and what is still open, is in the Threshold. This is the largest correction of the season. In Issue 9 it was a failure to look. In Issue 10 it was a failure to print: this paper's own files had recorded an outside build on September 23, two days before that issue repeated the claim, and its own planned re-run of the proof has still not been run.

Corrected: three smaller statements about the same proof. Issues 8, 9 and 10 printed OpenAI's concession that it "cannot rule out" help from a rival's data as its standing position; OpenAI's page has carried an update dated September 10 saying its investigation ruled that out. Issue 10 put inside quotation marks that the model "has resolved more than 100 open problems across mathematics"; OpenAI's words are "more than 100 long-standing open problems across most areas of mathematics". Issue 9 dated the Lean verification September 10; OpenAI's page says September 6. OpenAI, 9/8 · OpenAI, 9/21

Corrected: a correction. Issue 10 "corrected" Issue 9's Italian border-check date from roughly October 1 to October 8 and said no source gave Italy a distinct date. The European Commission's own table does: Italy's renewal runs from October 1, Spain's from October 8, and the Spanish period before it ended October 7, a day earlier than printed. European Commission Issue 9 was right. The correction is withdrawn.

Corrected: the Anthropic timetable. Issue 10 said the reporting of a mid-October launch "has not been superseded." It had been, seven days earlier: the Wall Street Journal, as carried by Investing.com, reported a November target on September 18. Investing.com, 9/18 Call #25 was marked "tracking YES" on a schedule that no longer existed.

Corrected: call #46's baseline. The call's note and last issue's Carry said WHO had 7,773 confirmed Ebola cases on September 23. That figure is dated September 21 and is the DRC institute's count as relayed by the UK Health Security Agency. WHO's own figure for September 23 is 7,890, with 3,799 deaths against the 3,759 printed. UKHSA, 9/24 · WHO DON618, 9/25 The "about sixty a day" in the note was derived from two figures one day apart and does not stand as arithmetic, though WHO's published pace since is close to it. The call's terms do not change.

Corrected: Kalshi against the wrong barrel. Issue 10 set Kalshi's 41 percent on Brent above $99.99 beside a $106.60 settle and read the gap as the market doubting the price. The $106.60 was the November contract; Kalshi's index was tracking a barrel near $98. Kalshi API · Investing.com The paper compared two instruments without saying so. The Grid has the rest.

Corrected: five market figures for September 24. Issue 10 printed the ten-year Treasury at "about 5.19", the two-year "about 4.94" and the thirty-year "about 5.41"; Treasury's own table has 5.18, 4.87 and 5.47. Treasury, September Ether "by Thursday" was printed at $2,644.20; three sources have the close near $2,687. Yahoo ETH Gold was printed at $4,299.95 against a settle of $4,298.00. Investing.com gold And the "dollar index" page cited was the December futures contract, not the cash index, with Thursday's figure wrong on either. Investing.com DXY futures · Yahoo Finance DXY

Corrected: energy and space figures. Refinery utilisation for the week to September 18 was printed at 96.6 percent; the EIA's table shows 94.0. EIA Issue 10 said no emergency order under Section 202(c) followed the one that expired September 18; three more were issued that day and a fourth on September 25, none for PJM. DOE The Impulse Space order of September 22 is $31,255,231 in the Pentagon's announcement, not $28 million, and the GHOST-R awards are dated September 18, not "around September 20." War Department, 9/22 · Space Systems Command, 9/18

Corrected: a citation that proved nothing. Issue 10 cited the Defense Security Cooperation Agency's arms-sales page for "no Taiwan notification for September." That address serves a list whose newest entry is from August 2025. DSCA The conclusion stands on another source; the citation does not. Taiwan Security Monitor, 10/7

Corrected: a drought comparison across two bases. Issue 10 set 102.9 million people in drought in the Lower 48 against a national figure from the week before and read a fall. Lower 48 to Lower 48, it was a rise of about 2.4 million. USDM population, 10/8

Corrected: two venues. Issue 10's pick lines read "Jets +6.5 vs Lions" and "Dolphins +11.5 at Chiefs." The Jets played in Detroit and the Chiefs played in Miami. Pro Football Reference · Pro Football Reference The grades do not change.

Corrected: this paper's counts of itself, found by the season audit. Issue 9 said twenty-three calls had settled and Issue 10 said twenty-five; their own tables show nineteen and twenty-one. Issue 9 · Issue 10 Calls #28 and #34 described September 16 as Chair Warsh's "first decision"; Issue 1 called the July meeting his second, which makes September at least his third, and the ledger row for #28 still carries the wording as logged. Issue 1 Issues 7 and 8 said call #27 was graded "nine days early"; the grade ran September 4 against a settle date of October 1. Issue 7 And Issue 1 said nine clocks were running and listed eight. No grade moves.

Resolved: the two primary sources this paper could not read. EDGAR, named as unread in Issues 9 and 10, was read directly on October 9. SEC EDGAR full-text search, read 10/9 The Pentagon's daily contract announcements, named as unread in the same two issues, were read in full for August 7 through September 30. War Department Both calls that waited on them are graded above.

Resolved: the Ceuta death toll figure. Issue 9 suspended the 145 attributed to Caminando Fronteras and Issue 10 ran it as "attributed and untraced." The NGO's own page prints it, as of August 11, and the page was last modified September 11, before either issue said it could not be found. Caminando Fronteras, 8/3 The figure is restored with its date. The failure was this paper's search.

Resolved: items left open last issue. Juan Jesús Vivas testified on September 30, the date Issue 9 printed and Issue 10 carried as disputed. ESdiario, 9/30 The two Drought Monitor figures missing for the September 24 release are 12.17 and 2.11 percent. USDM data service, 10/8 Taiwan's daily tallies, blank for two issues, are rebuilt for 14 of 20 days. Liberty Times The Google and Georgia Power terms are read, and the NV Energy suit's filing date is July 24. Georgia Power, 9/21 · Nevada Independent, 7/28 And Milwaukee did not finish as the league's only 100-win team, a possibility Issue 10 printed without the Dodgers' record beside it. Baseball-Reference

Suspended, not corrected: two things Issue 10 printed in the Wire. That the April ceasefire was an Israel and Iran instrument "on a different track": the one account this paper could open has the June memorandum extending an earlier ceasefire by sixty days, which puts them on the same track, and nothing opened says who the parties to the April ceasefire were. bdnews24 / Reuters, 6/18 And "the SAF recapture of Hamrat al-Wiz on September 6": the page cited, as served on October 9, records nothing after June. Siege of El Obeid Neither is shown false. Neither runs again until it has a source.

Downgraded, not lifted: the insurance multiple. The "forty times" this paper has printed for Hormuz war-risk cover, now thirty, is a tracking site's own estimate and says so on its page; it is not an underwriter's quote. Straits.live, 10/9 It runs labelled as that.

Named, not corrected: the Sudan sanctions date, again. Issue 10 corrected the expiry of the one-month rollover from October 9 to October 11 on a UN press release. Security Council Report says October 9. Security Council Report, 10/7 The UN page would not open on press day. Both are printed; the regime now runs to December 11 either way.

Named, not corrected: the hurricane record book. The latest first hurricane on record is October 8, 1905 by one forecaster's account, against the September 17, 1941 this paper printed in Issue 9; the seasons with no hurricane are 1907 and 1914 by two outlets and 1914 and 1917 by a third; the eastern Pacific's named-storm count is 17 on the Hurricane Center's table against the 18 printed last issue. AccuWeather, 9/29 · FoxWeather, 9/28 · NHC season table The Hurricane Center has said nothing on the record. Issue 10 also printed Gonzalo at 45 miles an hour where the Center's table has a 45 knot peak; the advisory was not re-opened.

Named, not corrected: figures other people disagree about. The odds of an October hike on September 23 were 54.2 percent in the source Issue 10 used and 73 in another. StockWireX The declaration Issue 10 said twenty-five Fields medallists signed lists 28 names. mathandai.org, 9/11 Nepal's missing are 5,285 by the disaster authority and were 6,150 by the police. Kathmandu Post, 10/6 Issue 10 said no third Schengen state had cited Ceuta; the Commission's table shows Austria doing so for a period that began September 16, with no date of notification. European Commission And nothing opened this fortnight confirms or contradicts the September 2 designation of Ceuta's Government Delegate that Issue 10 reported; the accounts of October 6 do not mention it. Diario de León, 10/6

Changed: the standing Hormuz count gets company. Lloyd's List resumed its Hormuz briefs on September 30. As promised in Issue 9, both counts print for two issues before any decision to switch back. The standing count remains Kpler's ten-day average, which this paper could not read this fortnight. Lloyd's List Intelligence, 10/8

Changed: facts that moved after press. The White House published a fact sheet on the summit at 10:29 p.m. Central on September 25, hours after Issue 10 said the summit had produced no document; it is one government's document and there is still no joint text. White House, 9/25 Vals re-versioned its index the same day, so every Vals figure in Issue 10 is superseded. Vals AI, 10/7 Magdalena Andersson's 28-day clock, printed as expiring about October 16, went back with her mandate and was replaced by a report due October 12. Riksdagen, 10/2 Jobless claims for the week of September 19 were revised from 197,000 to 198,000. Department of Labor, 10/8 And the PJM dates in last issue's Week Ahead are moot.

Named: a missed issue. This paper did not publish on October 2. No Issue 11 exists under that date. Four calls that settled on September 30 and October 1 are graded a week late in consequence, and every running count in the Carry skipped a week. The season scorecard proposes that the ledger print alone on any Friday the paper cannot.

Named: a coverage miss. Two landmines wounded three South Korean soldiers in the Demilitarized Zone on September 21, inside Issue 10's window. Issue 10 did not carry it. Al Jazeera, 10/6

Named: calls whose wording needs a ruling before their outcome, not after. Call #9: which transit count is read against its 120-a-week kill. Call #40: whether "a new Prime Minister" includes the outgoing one re-elected. Call #42: whether a compact suit against the Upper Basin is a suit "challenging" the Record of Decision. Call #43: whether a designation in a separate file counts in the docket the call names. Each is in its section. Two that were owed are ruled in this issue: what an older federal draft does to call #47, in the Grid, and call #10's kill, in the Season Scorecard.

Named: what could not be read. Kpler's ten-day average. ICE's own settlement report for September 30, so both Brent settles rest on wire and trade reports. The UN's press release on Sudan. CDC's measles count for October 8, not posted at press. Six fact-checking desks. The AWS and Oracle dashboards. What happened at the UN migration agency's September 30 stock-out in Sudan. Polo's national toll. Haze figures, for a second issue. Micron's results. The Brussels file, where no search was run at all. Friday's markets, because this issue was written before the open. The Upper Colorado River Commission's September meeting packet. The Hurricane Center's 10 a.m. advisory in its own text, read here through a newspaper's relay. And Isaias's landfall, which comes after press.

Named: how this issue was made. The research for this issue ran in fourteen lanes on the morning of October 9 and exhausted the session's 200-search limit before most of them finished, exactly as the first passes of Issues 9 and 10 did. Issue 10 adopted per-lane search caps as a standing rule; the caps set for this issue added up to more than the limit, which is an arithmetic failure in applying the paper's own rule. Three lanes ran no searches at all, the fact-check lane among them, so the by-name searches this paper has promised for Lies of the Week did not run, and the audit printed there says what stood in for them. The lanes worked from primary pages and publishers' own listings, which found the Navier-Stokes record, the Reclamation draft and the Commission's table, and could not find what only a search finds. No second pass with search had been run when this was written. This paper prints its process failures the same way it prints everyone else's, and this is the third issue in a row it has had this one to print.

Free. Stays free. Every call is on the ledger until it settles. This issue printed the season scorecard, as the first one said it would.

Next Commit.

In Service, -Tim