This week’s reading list is about the distance between what is good enough to act on and what is good enough to claim you know. The Dark Energy Survey watched the sky for 758 nights and came back with a three sigma preference for dark energy weakening over time, which is interesting, and is not the five sigma physicists require before saying so out loud. Markets are less fastidious. Jensen Huang resigned, Nvidia lost 14 per cent in a week, and Morningstar’s Colin Laidley notes that this one company had contributed more than nine percentage points of the American market’s 69 per cent return since May 2023. Howard Marks is candid that the honest answer on AI returns is largely a matter of faith. Ernie Tedeschi finds the thing everybody blames for the graduate job market is probably less responsible than working from home, and calls his own model suggestive rather than causal. Bethany McLean adds Nvidia’s disclosures up to US$530.5 billion, a total the company does not publish. The Reserve Bank’s November minutes record a six to three split and one line worth the whole document: data centre developers had raised debt equivalent to at least a fifth of the increase in business debt this year. The Bank for International Settlements has assembled 568 currencies since 1850; the median lasts 33 years. And an AI settled a famous conjecture its own author cannot follow. The useful lesson for investors is that the bar for acting is almost always lower than the bar for knowing. The best investors are not pretending otherwise. They are sizing positions accordingly.
📚 Minutes of the Monetary Policy Board Meeting, 2 and 3 November 2026: Six members voted to lift the cash rate 25 basis points to 4.85 per cent and three preferred to hold at 4.60, with inflation now expected to remain above 3 per cent until late 2027 and to return to the midpoint of the target range in mid 2028. Crude oil and refined petroleum prices had risen 20 to 25 per cent since August. The line to sit with is this: data centre developers had raised debt equivalent to at least one fifth of the increase in business debt this year, which makes the global AI build out a visible feature of our own credit aggregates. (Reserve Bank of Australia)
📚 Gaining (and losing) currency: a global database of currencies, 1850 to 2026: Iñaki Aldasoro, Jon Frost, Rudraksh Kansal and Catherine R Schenk have built the first systematic record of currency births and deaths, documenting 568 currencies of which 163 are still in use. The median lifespan is 33 years against a mean of approximately 52 in the paper’s own summary statistics, and the gap tells you how skewed the distribution is: 23 per cent of currencies lasted less than a decade while 12 per cent survived more than a century. The most common reason for creating a currency is the independence of a jurisdiction, at 28 per cent of cases, while the most common reason for ending one is a name or unit change, at 35 per cent. (Bank for International Settlements)
📚 The Comprehensive Guide to the Bank of Nvidia: Bethany McLean, who has form on circular financing, assembles what Nvidia has promised to whom. Her own tally of the disclosures in Nvidia’s latest 10-Q comes to US$530.5 billion of future commitments and guarantees, a total the company does not itself report, made up of US$366 billion of future commitments, US$56 billion of additional commitments and US$108.5 billion of guarantees. Hunterbrook Capital advises one or more investment funds and states that it held no position related to this article at the time of publication. (Hunterbrook Media)
📚 What Makes Up the Cosmos? A 20-Year Survey Hints That Something Is Missing.: Charlie Wood on the Dark Energy Survey, which ran for 758 nights between 2013 and 2019, extended past its planned finish to make up for observations lost to cloud during a strong El Niño season. It reports a three sigma preference for a theory in which dark energy weakens over time, against the five sigma standard the field holds itself to before announcing a discovery. Twenty years of work, and the discipline is still declining to call it. (Quanta Magazine)
📚 AI proved a major quantum complexity conjecture. What does it mean?: Ethan Siegel on the Aaronson-Ambainis conjecture, longstanding and widely thought to be true, and now settled by a proof an AI found. Scott Aaronson, who noticed in 2008 that a solution would follow from it, is reported as quite impressed while not himself understanding the proof as it was presented. Siegel’s verdict on the practical consequence is admirably flat: the impact on practical quantum computing is zero, because what has been ruled out is quantum advantage for yes or no problems relative to random oracles, while the speedup still expected in quantum chemistry and materials science is untouched. (Big Think)
📚 Soldiers of Misfortune: The Drawings of Urs Graf the Elder: Miya Tokumitsu on a Swiss Renaissance artist, trained as a goldsmith and working in woodcut, engraving and etching, who fought in at least four military campaigns as a mercenary and fled the city in 1518 after an assault so violent that his victim was permanently disabled. Set his soldiers beside Dürer’s stoic, steadfast Christian knight and the contrast is the point: one is an ideal almost nobody could meet, the other is what the work actually looked like. A reminder that the people inside a boom rarely resemble the picture painted of it afterwards. (The Public Domain Review)
📚 Models of Football Capitalism: Alexandre Afonso applies a trade balance framework to the football transfer market and finds national specialisations hiding in plain sight. English clubs spent 11.3 billion euros on foreign players and sold 4.7 billion, a deficit of 6.6 billion, while France sold 5.0 billion and bought 3.3 billion and Germany sold 4.0 billion against 3.2 billion. Portugal and the Netherlands sit among the feeders, his own category for those who sell at least 25 per cent more than they buy and balance the books by developing talent and selling it on. (Alexandre Afonso’s Political Economy Newsletter)
