Daily Briefing — Tue 13 Oct 2026

Daniel Brooks

Good morning, Walter. It's Tuesday, October thirteenth, twenty twenty-six, here in Costa Rica. Four new episodes landed in your queue since yesterday's briefing, and three of them are about money, so let's start there.

Daniel Brooks

First up, Invest Like the Best. Patrick O'Shaughnessy brings back David Vélez, the founder and CEO of Nubank, five years after their last conversation. Back then Nubank had about thirty-five million customers. Today it has around a hundred and forty million. Vélez's core argument is that fintech is still in the early innings. He puts global consumer financial services at roughly ten trillion dollars, with about ninety percent of it still running through incumbent banks and branches. He compares it to most of the world still renting movies at Blockbuster.

Daniel Brooks

The numbers he uses to back that up are striking. Nubank serves about a hundred and forty million customers with ten thousand employees. He says a comparable incumbent serves a hundred million with a hundred thousand employees and six thousand branches, so Nubank's cost to serve a customer is about three percent of a traditional bank's. In Brazil it has close to a hundred and twenty million customers, about sixty-five percent of adults, and roughly sixty percent of them use it as their main account. On the United States, he points to about fifteen trillion dollars of retail deposits earning around forty basis points. In a four to five percent rate world, he estimates that's about six hundred billion dollars a year staying with banks instead of savers.

Daniel Brooks

There's also a lot on AI inside the company. Nubank bought a Palo Alto startup called Hyperplane and built its own foundation model for credit decisions, trained on thirteen years of customer data. He also tells a candid story about missing the moment, around December twenty twenty-five, when coding models suddenly got good at Clojure, Nubank's main programming language. Once they caught up, they rebuilt about a hundred and twenty internal workflows around AI and cut management layers from ten to five. He went from one direct report to twenty. One personal note you'll appreciate: he was born in Colombia and moved to Costa Rica at eight. Nubank is now launching in the US and rolling out a stablecoin-backed account in about thirty-five countries. Worth a full listen. It's seventy-three minutes, and the economics of digital banking plus the frank AI-reorganization story make it the best episode of the day.

Daniel Brooks

Next, Slate Money. Felix Salmon interviews David Skeel, a Penn bankruptcy law professor who spent nearly eight years on Puerto Rico's oversight board, the last three and a half as chair. He has a new book on the debt crisis. The scale is the first takeaway. Skeel puts it at roughly seventy-two to seventy-four billion dollars of debt plus about fifty-five billion of unfunded pensions, which is bigger than Argentina's default. Twelve of the thirteen government entities have finished restructuring, cutting about sixty-three billion dollars of debt to twenty-seven billion. He defends the roughly two billion dollars in total fees as money well spent. The one that's still open is PREPA, the power utility, with about ten to eleven billion dollars outstanding. A deal nearly closed in twenty twenty-four, then fell apart over a court ruling on bondholder collateral. He says a big hedge fund holdout has kept the other bondholders together, and he expects a resolution late this year or early twenty twenty-seven.

Daniel Brooks

The bigger idea is that Congress should let states go bankrupt the way cities already can. Skeel argues it wouldn't push up borrowing costs across the board. Riskier states would pay more, and safe states like Utah might even pay less, because the market could tell them apart more clearly. If you follow munis or sovereign restructurings, this forty-six-minute conversation is worth your time. If not, the summary covers most of it.

Daniel Brooks

Third, Investment Management Operations, from the Capital Allocators family. Scott McDonald talks with Alberto Rizzoli, co-founder of V7 Labs, about turning investment operations into AI workflows. A common starting point is turning a deal's information memorandum into an investment committee memo, along with manager diligence and data-room reviews. A few practical points stood out. Rizzoli says about ninety-five percent of the written procedures clients bring in are only about eighty percent complete, so the real work is covering the edge cases, not the happy path. He says the step past a chat assistant is consistency, meaning the fifth run looks like the first and every claim is cited back to its source document. He reports about a twenty-times time saving on deal screening. He also argues for routing each step to whichever model is best at it and for avoiding vendor lock-in. His advice for getting started is to just jump in, try coding agents on knowledge work, and read Andrej Karpathy's short piece on Software three point oh. It's partly a product pitch, so I'd skim it. That said, given the agent workflows you're building at KSC, the parts on writing procedures and running evals are worth about twenty minutes of the forty-eight.

Daniel Brooks

Last, a quick one from The Indicator by Planet Money. This year's economics Nobel went to Robert Hall of Stanford and Michael Woodford of Columbia, for what the committee called the microeconomic foundations of macroeconomic stabilization. In plain terms, they rebuilt macro models from how people and firms actually decide to spend, save and invest, and how expectations shape those choices. That work fed the New Keynesian framework central banks still use. Most of the episode is the origin story. The prize was created by Sweden's central bank, the Riksbank, for its three hundredth anniversary, partly to burnish the bank's prestige. The Nobel family pushed back, which is why it isn't technically a Nobel Prize. It was first awarded in nineteen sixty-nine, and the central bank still pays for it. At nine minutes, it's an easy yes.

Daniel Brooks

One housekeeping note. The Spellcaster feed about Sam Bankman-Fried dropped a short promo for another Audible show. There's nothing new there, so I skipped it.

Daniel Brooks

That's your briefing for Tuesday. If you only have time for one, make it David Vélez. Have a great day, Walter.