Long-form reporting and commentary on banking, finance, and technology
A cohort of machine-learning engineers, hired in the aftermath of the 2023 banking crisis, now occupies positions of extraordinary influence within the Fed. Their models do not merely advise — they increasingly constrain the range of choices available to the governors themselves.
Continue reading ▸The ninth floor of the Eccles Building in Washington, D.C., has never been known for its warmth. The corridors are linoleum, the lighting fluorescent, the décor an exercise in federal austerity. Yet it was here, in a windowless conference room designated simply as 9-C, that the Federal Reserve made what may prove to be the most consequential institutional decision since Paul Volcker raised interest rates to twenty per cent in 1981.
The decision was not announced. There was no press conference, no Congressional testimony, no carefully worded statement from the Chair. What happened was at once more mundane and more radical: the Fed hired forty-three machine-learning engineers in a single quarter, embedded them within the Division of Monetary Affairs, and gave them access to the same real-time data feeds that inform the governors' rate decisions. Within eighteen months, the models these engineers built were producing forecasts that outperformed the staff's consensus projections on seven of nine key indicators.
With Brent crude above $110 a barrel and tariff-driven core inflation at three per cent, the Federal Reserve confronts a supply-side dilemma that its instruments were never designed to resolve, and history offers cold comfort to those who seek precedent for the way out.
Continue reading ▸The final implementation of the Basel IV framework has produced precisely the distortion that its critics predicted: a migration of credit risk from regulated banks to an unregulated shadow sector that dwarfs the one which preceded the 2008 crisis.
Continue reading ▸For three decades, corporate treasurers operated on the assumption that rates would return to their historical mean. That assumption is dead — and the repricing of risk that follows may prove the defining financial event of the decade.
Continue reading ▸Washington's ambitions for a CBDC have advanced far beyond what the public discourse suggests. Internal documents reveal a system designed not merely to digitise money, but to embed fiscal policy directly into the medium of exchange.
Continue reading ▸There is a phrase that has become the all-purpose reassurance in every board presentation about AI deployment. It deserves a great deal more scrutiny than it receives.
Continue reading ▸When eurozone inflation ranged from 7.1 per cent in France to 22.5 per cent in Estonia, the ECB set one rate for all. Robert Mundell anticipated this problem in 1961. The institution has still not solved it.
Continue reading ▸While the world’s attention fixed on Japan’s historic departure from negative interest rates, the central bank was quietly building an AI research programme that may prove equally consequential.
Continue reading ▸JPMorgan operates 450 AI use cases across 200,000 employees. BNY Mellon's staff are building their own agents at scale. Goldman Sachs has moved from assistant to autonomous agent. This is not the chatbot story. It is an infrastructure story.
Continue reading ▸The Federal Reserve’s characterisation of post-pandemic inflation as transitory was not merely a communications failure. It was a diagnostic error of the first order, one whose historical precedents are more instructive than the institution’s defenders have been willing to acknowledge.
Continue reading ▸European lenders are pushing regulators to recognise machine-learning credit models as grounds for lower capital requirements. The argument has merit. The risks of conceding it are larger still.
Continue reading ▸When Goldman Sachs deploys autonomous AI agents for trade accounting and KYC, it is not merely automating tasks. It is automating the application of judgment to structured problems, which is a rather different thing.
Continue reading ▸On the first of January 2026, China reclassified its e-CNY wallet balances as commercial bank liabilities. The digital yuan survived its first decade. Whether what survived can still be called a central bank digital currency is a more interesting question.
Continue reading ▸Vice Chair Bowman's March 2026 re-proposal eliminates the nineteen per cent capital hike that defined the Basel III Endgame since 2023. The industry won. The harder question is what the supervisors are conceding.
Continue reading ▸The failure of Overend, Gurney & Company on the 10th of May last has furnished a lesson of which the gravity cannot be overstated. That a house so long established, so universally trusted, and so intimately woven into the fabric of British commerce should have fallen — and fallen not from external calamity but from the recklessness of its own partners — ought to give pause to every man who presumes to direct the employment of other people's money.
We republish the original 1866 analysis in full, together with a modern commentary examining what Overend, Gurney reveals about the failure of SVB and the fragility of confidence in the age of digital bank runs.
Collateralised loan obligations have become the dominant buyer of leveraged loans in the United States, yet the mechanics of these structures remain opaque to most officials charged with monitoring systemic risk. This investigation examines the hidden leverage, the concentration risk, and the assumption that has held the entire edifice together.
Continue reading ▸The secondary market for PE stakes has tripled in volume since 2022, driven by limited partners desperate to rebalance portfolios overweight in illiquid assets. But discounts exceeding 30 per cent of NAV raise an uncomfortable question: if PE is worth what the managers claim, why are sophisticated investors willing to sell at a loss?
Continue reading ▸While the West debated the merits of CBDCs, India built the world's most successful real-time payments system. UPI now processes more than fourteen billion transactions per month — more than Visa and Mastercard combined within the country — and its architecture is being adopted from Singapore to Brazil.
Continue reading ▸SWIFT's dominance as the backbone of cross-border payments is under pressure from three competing blockchain-based settlement networks. Each promises lower costs, faster settlement, and reduced counterparty risk. This analysis examines whether any of them can deliver on those promises without sacrificing the regulatory compliance that underpins trust.
Continue reading ▸