Ad Narrator | Invest like the best Summary

1. Key Themes

Banking Is Where Blockbuster-to-Netflix Is Still ~90% Ahead

Vélez frames the whole opportunity as an early-innings migration. All neobanks and fintechs combined are under 10% of a ~$10T global consumer financial services industry, while ~90% still flows through branches owned by incumbents. Nubank is the proof point that the migration can fully complete in a market.

"If you look at global consumer financial services, it's about a $10 trillion industry, and still, something like 90% of that industry goes through banking branches and is owned by incumbent banks." [00:02:27]

"We are really the only neobank in the world where we've fully surpassed the biggest incumbent bank in terms of scale... about 60% of our customers use us as their primary bank account." [00:02:27]

Cost-Structure Advantage Is the Real Moat (3% Cost to Serve, 2–3x ROE)

The compounding advantage is structural: no branches means a radically lower cost base, which funds free accounts, better yields, and still produces superior returns. Vélez uses ROE as the single metric proving the model works.

"We have 140 million customers with 10,000 employees. Our competitors have 100 million customers with 100,000 employees and 6,000 branches... our cost to serve a customer is about 3% of what a traditional incumbent bank is." [00:04:54]

"We really charge no fee, so we have less revenue, but then we end up with a return on equity that is 2 to 3x their return on equity... then you found yourself a moat." [00:38:06]

Three Problem Buckets: Access, Cost, Intelligence

Vélez's taxonomy for where financial services is broken, and where value will shift to consumers. The cost bucket is quantified with a striking US figure (~$600B/yr in unpaid yield), and intelligence is the least-understood one.

"On the access side... you still have 3 billion people in the world that don't have access to a credit product." [00:06:45]

"There's about $15 trillion of retail deposits in the US banking system, but banks pay on average about 40 basis points on yield a year... that's more or less $600 billion that are going to banks to pay for banking branches." [00:06:45]

"People don't sit down to do classes. You're not going to solve this problem of intelligence through education, you're going to solve it through AI." [00:06:45]

Own Data + Foundation Model: Scale Only Matters if Culture and Systems Exist

Nubank bought Hyperplane (Palo Alto) and built a proprietary financial-services foundation model on 20–30,000 data points per customer across 13 years. But Vélez insists scale alone is worthless without data-native systems and culture, using Blockbuster as the counterexample.

"We acquired a company called Hyperplane in Palo Alto that was already working on doing foundation models for financial services. They were having such a hard time selling to the big incumbent banks. Sales cycles were 2 years." [00:17:46]

"Blockbuster had a huge amount of data... Did that data help them build a better search algorithm? No, because they were not capturing that data, they did not have the right systems to capture data, they didn't have the right culture, data culture." [00:19:32]

The End of Cash Unlocks Credit for the Informal Economy

Digitizing transactions creates the ledger that makes underwriting possible. Vélez calls this the least-understood effect of going cashless, with trillions of dollars of latent credit for small businesses.

"That taco stand cannot get credit today... The moment that that transaction goes into a digital ledger... that taco seller can take that information to a bank and says, 'Hey, look at my cash flow. Give me access to credit.'" [00:27:43]

"Something like 70% of all the employment in the country is still in small businesses... this is trillions of dollars of credit being enabled for billions of people around the world." [00:27:43]

Open Finance, Real-Time Payments, and Stablecoins as Tailwinds

Three exogenous shifts weaken incumbent inertia: Pix/UPI-style payment rails, consumer-owned data portability, and dollar stablecoins. Vélez sees fiat currencies consolidating and stablecoins as a savior of dollar dominance.

"The moment that you can connect all your history from all your services and pass it to somebody else, you're reducing that inertia and you're enabling more competition for your data." [00:23:09]

"Today, there's about 190 currencies around the world. My view is that 10, 20 years from now, those 190 will be maybe 10." [00:23:09]

"US stablecoins, ultimately, will be the big savior, or one of the biggest saviors, of the US dollar, because it democratizes access to the US dollar." [00:26:32]

Reinventing the Whole Organization with AI: 120 Workflows, 10 Layers to 5

A detailed operating story: Nubank missed the December 2025 Opus 4.6 inflection by ~60 days, called a "code red," mapped 120 workflows, built eval infrastructure, and flattened the company dramatically.

"We mapped out 120 different workflows across the entire organization, coding is one of them, but everything from, how do you do product design, know-your-customer flows, internal audit, compliance, regulatory capital management." [00:46:06]

"I personally went from having one direct report to 20 direct reports over that period... We went from 10 layers to five." [00:46:59]

AI Will Reinvent HR: Olympic-Athlete Coaching for Every Employee

Instead of semiannual recency-biased reviews, instrument every role so AI gives daily, specific feedback, and managers see real-time performance signals rather than low-information-density 1-on-1s.

"Imagine that's the type of feedback you would be giving an Olympic athlete. That would be crazy... that is the opportunity that ultimately AI is going to create for every single person in an organization, is you have Olympic-athlete type of coaching." [00:51:21]

Customer Obsession Is the Differentiator AI Builders Neglect

Vélez argues the AI debate is dominated by efficiency, with little attention on customers. Nubank studies Amazon and Disney to engineer "magic" even in boring banking products.

"In the entire AI debate, you actually don't hear that much conversation about customers... This is where humans still are beating the AI." [00:41:47]

Founder Mode vs. CEO Mode, and Imposter Syndrome

A candid leadership thread: the founder is impatient and urgent, the CEO systematic. Vélez argues to lean toward founder mode, especially in technology shifts, and that visceral care for the company is what qualifies a founder to keep leading.

"The founder, it's impatient, huge amount of sense of urgency... The CEO will need to be more patient." [00:59:27]

"I don't think anybody in the world cares as much for this company as I do... that is what is going to make you be the right leader for the organization for a very long time." [01:08:54]

2. Contrarian Perspectives

AI Makes Incumbents-at-Scale Stronger, Not Weaker, and Startups Face a Harder Road

The popular view is that AI collapses the cost of building and lets startups topple incumbents. Vélez argues for financial services the opposite: licenses, trust, brand, and proprietary data are moats that AI cannot be prompted into existence, and a company like Nubank sits in the sweet spot because it also has the tech stack to exploit AI.

"You still cannot really get ChatGPT to get you a banking license, and you still cannot really get them to have regulators give you a license in, in 6 months instead of 2, 3 years." [00:13:52]

"Companies like us, we're sort of in that sweet spot of benefiting the most from AI, because in one way, we already have the scale and the data and the brand and the trust and the licenses." [00:13:52]

Financial Education Is the Wrong Answer; AI Intelligence Is the Right One

The consensus remedy for poor financial outcomes is literacy. Vélez says it has never worked and that intelligence-as-a-service will compete away profit pools banks currently earn from consumer inattention.

"Historically, there's been a lot of conversation about education is the answer... That never worked." [00:06:45]

The 190 Fiat Currencies Will Collapse to ~10, Many Being "Crypto Tokens" in Disguise

A bold, specific prediction with a fiscal-dysfunction rationale, paired with a belief that the dollar persists via stablecoins.

"The fiscal situation in so many countries around the world is so dire that a lot of that fiat money is not real money. It's basically crypto tokens." [00:23:09]

In an AI Era, More In-Person Work, Not Less

Against the remote-work-plus-AI narrative, Vélez argues AI replaces what an individual can do alone at home, so human value concentrates in physical collaboration and coordination.

"The individual at home will be competing with AIs and will lose that battle. But the work around collaboration and coordination, in presence physically, which is what humans need to do, is what needs to be happening inside the office." [00:44:36]

Lower the Barrier to Starting and You Get Worse Founders' Problems; the Best Go for the "Impossible"

AI makes starting easy, which Vélez says tempts founders toward easy problems. He sees the best founders now gravitating to hardware and space.

"I think AI makes it harder, not easier, just because the bar to start a company is now so much easier, you will probably just going to go and pursue problems that are easier to solve ultimately." [01:10:37]

3. Companies Identified

Nubank (Nu Holdings)

Brazil-headquartered digital bank with ~140M customers across Brazil, Mexico, and Colombia, now launching in the US. Mentioned as the subject of the episode and the only neobank to surpass its country's largest incumbent. Key evidence of scale and profitability.

"We're now the largest private financial institution in the entire continent, we have the highest Net Promoter Score. From a profitability perspective, we have the highest return of equity in a country like Brazil." [00:34:11]

Hyperplane

Palo Alto startup building foundation models for financial services, acquired by Nubank. Mentioned as the origin of Nubank's proprietary credit foundation model, and a case of great tech failing to sell to incumbent banks (2-year sales cycles).

"We acquired a company called Hyperplane in Palo Alto that was already working on doing foundation models for financial services." [00:17:46]

Amazon

Cited as the most-studied consumer-obsession model, and the source of the "your margin is my opportunity" framing for attacking high-ROE incumbents.

"We were very inspired by Amazon since the very beginning. Amazon being the more traditional consumer-obsessed company." [00:39:49]

Disney

Studied for creating "magical experiences" and communication with consumers; applied to even boring banking moments like welcome kits.

"We studied Disney very carefully because this is a company that wants to create magical experiences." [00:39:49]

Netflix

The central analogy for digitally native disruption of incumbents who held more data but lacked systems and culture.

"Somebody like Netflix was able to come in with no data and build a better algorithm and, ultimately, use that." [00:19:32]

Plaid

Open-banking data aggregator, cited as the type of infrastructure that, combined with AI, lets consumers see and optimize their entire financial lives.

"Plaid can connect 120 million accounts and you put intelligence on top of that, you're effectively allowing each customer to know how to optimize their financial life." [00:11:00]

Pix (Brazil) and UPI (India)

Real-time payment systems named as catalysts accelerating access to consumer banking rails.

"Enter a number of real-time payment systems, as you've seen in Brazil with Pix, or India with UPI." [00:06:45]

Wealthfront

Referenced by Patrick for its "self-driving money" concept as a model for AI-driven personal finance.

"Wealthfront used to have this idea of self-driving money, which comes to mind." [00:10:36]

Uber, Spotify, Barnes & Noble

Named as examples of the incumbent-to-digitally-native migration (transportation, media, retail) that Vélez says is still early in finance.

"That migration from incumbents to digitally native companies is still very, very early in financial services." [00:02:27]

Ramp, WorkOS, Rogo, Vanta, Ridgeline

Sponsor/advertised companies (finance operations, enterprise auth, finance AI, compliance automation, investment management software) read in ads; Patrick states he runs his business on Ramp.

"I run my business on Ramp and so should you." [00:00:01]

OpenAI / ChatGPT, Codex, Anthropic Opus

Named as the AI tools that crossed a capability threshold for Nubank's Clojure codebase; Opus 4.6 (December 2025) is the pivot point.

"Opus 4.6 came in in December of '25, and that was sort of the moment that a lot of people suddenly realized, 'No, this actually works.'" [00:45:22]

Clojure

Programming language heavily used at Nubank, used as the benchmark for whether AI coding tools actually work in their stack.

"We use a lot of Clojure as a programming language, and through 2025, we kept using a lot of the models for Clojure for programming specifically, and it wouldn't work." [00:45:01]

Shopify

Referenced via Tobi Lütke's use of AI to remove the gap between the CEO and ground truth.

"Tobi Lütke has talked about this, how he can now ask the company stuff because of AI can tell him stuff." [00:48:44]

4. People Identified

David Vélez

Founder and CEO of Nubank; formerly an investor. Guest. Notable for building the first neobank to surpass an incumbent and for a detailed AI-driven reorganization.

"I find that I'm more of a wartime general than a peacetime general." [00:56:41]

Micky Malka

Founder of Ribbit Capital, described by Patrick as possibly "the best fintech investor ever" and a mutual friend. Remains a Nubank investor and is doubling down on later-stage neobanks while spending more time on AI.

"We are just doubling down on the larger neobanks or neo fintechs that are now right in the sweet spot between maintaining the ability to transform themselves through AI, but now have access to these moats of scale and data and brand and licenses." [00:33:14]

Tobi Lütke

Shopify CEO, cited by Patrick as an example of a CEO using AI to eliminate the separation between leadership and ground-truth data.

"He can now ask the company stuff because of AI can tell him stuff. So, the separation between him and ground truth based on data is gone." [00:48:44]

Jeff Bezos

Referenced by Vélez through the "Amazon or Bezos view" of attacking incumbents with extraordinary margins.

"The typical Amazon view or Bezos view of that incredible ROI that even incumbent banks have... that means there is a significant opportunity to undercut." [00:38:50]

Patrick O'Shaughnessy

Host; CEO of Positive Sum. Contributes the "dictator's dilemma" framing about leaders losing contact with truth.

"I like this idea of the dictator's dilemma, that as someone gets more powerful at the top of an organization, it's harder for them to know what is real and what is true." [00:48:44]

5. Operating Insights

Make Every Senior Manager Ship Software Each Quarter

Vélez requires senior managers to build real products every quarter, to stay close to AI progress through use rather than reading. He also blocks calendar time to build himself and uses an agent that summarizes AI developments.

"We ask all of our senior managers to build products every quarter, actual software and products, so that we are integrated and very aware of what's going on." [00:53:19]

Treat AI Agents as a Formal Customer Segment

Nubank staffs a team for agents as a segment on par with under-18s or small business: building product, marketing, and SEO for agents while still marketing to humans, where trust matters.

"We have a team dedicated to understanding that segment and understanding the needs of that consumer, which is the agent." [00:43:00]

Flatten Through a Company Brain and Skip-Level Directness

Go from 10 layers to 5 and from 1 to 20 direct reports by giving the CEO a proactive, reactive information system and a norm of talking to anyone with the best information.

"I want to go talk to anyone that I have a question with, and there is no respecting of a chain of command." [00:50:10]

Build Eval Infrastructure Before Reinventing Workflows

When reinventing 120 workflows with AI, Nubank created eval infrastructure to measure progress at each step, rather than adopting tools on faith.

"We had to create a lot of eval infrastructure to be able to measure at every single step how these workflows get reinvented." [00:46:34]

Start Credit Products Tiny ($20 Lines) and Coach Customers Up

A product tactic for building creditworthiness: ultra-small first lines, then an AI copilot coaching behavior, creating the data and the customer simultaneously.

"The first line was $20, a $20 credit line. When banks saw that product, they thought it was absurd... those $20 became a 100, and the 100 became a 1,000 over a period of time." [00:29:27]

6. Overlooked Insights

A 60-Day Blind Spot Was the Gap Between Nubank and Best-in-Class AI Adopters

Said almost in passing: even a technology-native, data-driven company missed the December 2025 model inflection for two months, and that short lag mattered materially. It implies adoption velocity is now measured in weeks, and that only a "code red" reaction closed the gap.

"We didn't catch that moment, the switch moment in December. We only catched it 60 days after... And it was 60 days where it made a huge amount of difference between us and best-in-class companies that were using AI." [00:45:22]

Nubank's Product Is Already a Competitor to Incumbents' Fees: Daily Interventions on 15 Million Customers

Buried in the yield discussion is a live example of AI-enabled profit-pool destruction: proactively telling ~15M customers about due dates at rival banks charging 15% daily interest and refinancing them. This is the mechanism by which intelligence strips fee revenue from competitors and deepens Nubank's primary-account status.

"About 15 million of our customers, every day, we're telling them, 'Hey, your bank account, your credit card in that other bank is due tomorrow. You don't have any money there, and they're going to charge you a 15% interest rate per day. Let us give you some money so you can pay that credit card.'" [00:11:00]