One new story. One old one. Data first, leap second.

NEW. On 7 October 2026 the Federal Reserve released the minutes of the 15-16 September FOMC meeting. Buried in the market operations section is the record of what the United States did in the currency market in late July: it bought yen. The minutes state that the New York Fed's Desk acted "purely as fiscal agent for the U.S. Treasury," using U.S. Treasury funds, and that "the System Open Market Account portfolio was not involved." Two paragraphs later the Committee records that "there were no intervention operations in foreign currencies for the System's account during the intermeeting period." The statutory machinery behind that is 31 U.S.C. 5302, whose stabilization fund traces to the Gold Reserve Act of 30 January 1934, and under which "decisions of the Secretary are final and may not be reviewed by another officer or employee of the Government."

Japan's Ministry of Finance put its own intervention total for 30 July to 26 August 2026 at 15,399.3 billion yen. The American figure is not public: the Treasury and Federal Reserve quarterly foreign exchange report for the third quarter has not been released, and the most recent one, covering April to June, says "the Federal Reserve and the U.S. Treasury did not intervene in foreign exchange markets during the quarter." The exact trading day is reported but not established, and this episode says so on air.

The leap, refused: none of this establishes a deliberate devaluation. The minutes attribute the dollar's decline to "narrowing interest rate differentials and improving foreign growth," the same meeting raised the federal funds target range to 3-3/4 to 4 percent on a 12-0 vote, and the IMF's 30 September data brief, which reported the dollar's reserve share falling to 56.70 percent, also reports that "claims in U.S. dollars were largely unchanged in 2026Q2."

OLD. On 29 September 1995 the American Institutes for Research delivered "An Evaluation of Remote Viewing: Research and Applications" to the CIA, the review that ended the Star Gate programme. The panel was built to disagree: Jessica Utts, a statistician sympathetic to the paranormal, and Ray Hyman, a skeptical psychologist. Both versions of this story are wrong. Utts concluded that "psychic functioning has been well established." Hyman conceded that "something other than chance departures from the null hypothesis has occurred in these experiments," then explained why that is not the same as a demonstration. What closed the programme was the operational record: when judging asked for specific information, "weak or insignificant effects were obtained," and across more than two hundred taskings from 1986 to fiscal year 1995, "in no case had the information provided ever been used to guide intelligence operations." The viewers also told interviewers that they and their managers edited written reports to match known facts about the target.

Tiers are labelled out loud throughout: what is documented, what is reported but not established, and the leap. The pairing of the two stories is editorial and the episode says so.

Primary sources

• Minutes of the Federal Open Market Committee, 15-16 September 2026, released 7 October 2026: https://www.federalreserve.gov/monetarypolicy/fomcminutes20260916.htm
• Treasury and Federal Reserve Foreign Exchange Operations, April-June 2026, Federal Reserve Bank of New York: https://www.newyorkfed.org/medialibrary/media/newsevents/news/markets/2026/q2-2026-fx-quarterly-report.pdf
• Ministry of Finance, Japan, Foreign Exchange Intervention Operations, 30 July to 26 August 2026, released 28 August 2026: https://www.mof.go.jp/english/policy/international_policy/reference/feio/monthly/20260828e.html
• 31 U.S.C. 5302, Stabilizing exchange rates and arrangements: https://www.law.cornell.edu/uscode/text/31/5302
• IMF Data Brief, Currency Composition of Official Foreign Exchange Reserves, 2026Q2, released 30 September 2026: https://data.imf.org/en/news/imf%20data%20brief%20september%2030
• Mumford MD, Rose AM, Goslin DA, "An Evaluation of Remote Viewing: Research and Applications", American Institutes for Research, 29 September 1995, read from the National Security Archive copy (NSAEBB 438): https://nsarchive2.gwu.edu/NSAEBB/NSAEBB438/docs/doc_57.pdf

No guests. No ads. Every claim has a retrievable document behind it, and where one does not, the episode says which tier it is standing on.

One new story. One old one. Data first. Leap second.

NEW. A randomized simulation study published 22 September 2026 put a real-time, voice-based AI diagnostic assistant into 260 simulated primary care consultations with 13 board-certified physicians. Top-3 diagnostic accuracy rose from 62.3% to 74.6%. The same paper reports that in the 51 consultations where the assistant's operative suggestion was wrong and visible, the physician's first diagnosis matched the wrong suggestion in 29 of them. The control arm was a "resource-restricted" doctor with no references, guidelines, search or colleagues. On the easiest quarter of cases the unassisted arm came out ahead. The study was funded by the company that built the system, and one author is a shareholder. All of that is printed in the paper, along with a public dataset.

OLD. Britain's Post Office prosecuted its own subpostmasters on the strength of figures produced by the Horizon accounting system. In Bates v Post Office (No. 6: Horizon Issues) [2019] EWHC 3408 (QB), Mr Justice Fraser found that "Legacy Horizon was not remotely robust", that Horizon did not alert subpostmasters to bugs, and that Fujitsu could insert, alter or delete branch transaction data remotely, without the subpostmaster's knowledge and without their consent, in a way that looked as though the subpostmaster had done it. He described the Post Office's position as "the 21st century equivalent of maintaining that the earth is flat." Parliament then quashed the convictions by statute. As of 30 June 2026 the Ministry of Justice counts 619 people and 2,249 convictions quashed by the Act in England and Wales.

Sources

• Cusacovich I, Pinilla M, Garcia Castro R. "Real-Time Artificial Intelligence Diagnostic Copilot in Simulated Primary Care Consultations: Randomized Simulation Study." JMIR Formative Research 2026;10:e104579. DOI 10.2196/104579. PMID 42771885. Full text read via NCBI.
• Bates and Others v Post Office Limited (No. 6: "Horizon Issues") [2019] EWHC 3408 (QB), 16 December 2019. Full judgment: https://www.judiciary.uk/wp-content/uploads/2019/12/bates-v-post-office-judgment.pdf
• Post Office (Horizon System) Offences Act 2024 (c. 14): https://www.legislation.gov.uk/ukpga/2024/14/contents
• Explanatory Notes to the Act: https://www.legislation.gov.uk/ukpga/2024/14/pdfs/ukpgaen_20240014_en.pdf
• Ministry of Justice, "Quashed convictions management information: 2 July 2026", updated 24 September 2026.

Tier note: everything above was read from the primary document. The Horizon Inquiry's own findings on who knew what were not read for this episode and are not relied on.

One new story. One old one. Data first. Leap second.

NEW. On September 22, 2026, Cell published "A geothermal amoeba sets a new upper temperature limit for eukaryotes." A single-celled organism pulled from a hot spring in Lassen Volcanic National Park divides at 63 degrees Celsius and crawls at 64, three degrees above the ceiling long cited for complex life. The paper is careful: it prints the replicate variability at the record temperature in the very next sentence, and says the true limit is still unknown. The NASA release built on top of it says the discovery "proves that assumption wrong." Reading both side by side is the episode.

OLD. December 2, 2010: NASA announced a bacterium from Mono Lake that "substitutes arsenic for phosphorus in its cell components," and said the finding "will alter biology textbooks." The paper's own methods disclosed the background phosphate contamination that would eventually sink it. Eight technical comments landed within six months. Two refutations landed within twenty months. On July 24, 2025, Science retracted the paper. No fraud, no misconduct, and the authors still disagree.

Primary sources: Rappaport et al., Cell, DOI 10.1016/j.cell.2026.08.043 (PMID 42772279); NASA release, science.nasa.gov, September 22, 2026; Wolfe-Simon et al., Science 332(6034):1163-1166, DOI 10.1126/science.1197258 (PMID 21127214, now flagged RETRACTED); NASA Release 10-320 and Media Advisory M10-167; eight Technical Comments and authors' Response, Science 332(6034):1149; Erb et al., Science 337(6093):467-470; Reaves et al., Science 337(6093):470-473; Retraction, Science 389(6758):357, DOI 10.1126/science.adu5488 (PMID 40705905); Thorp and Vinson, editors' post, July 24, 2025.

The retraction notice itself sits behind a paywall and was not read for this episode; that limitation is stated on air, and any wording attributed to the notice comes through secondary reporting.