Half of today’s dossier is the same question in different clothes: who owns the customer relationship, and what it costs to buy one that didn’t come as an inheritance. Nubank opens in the US without a single branch and states the number that replaces the physical network: $1 of expense per customer against $20 at the big American banks. Porto makes the opposite move on the same logic — it already has 44,000 brokers, the most expensive channel to build in the Brazilian market, and decided to train them as creators, media it doesn’t have to buy. One builds the channel with price because there’s no other way; the other found out the channel it already had was media. It’s the third time in a week this section lands here (Poppi betting on the channel nobody measures, Oura taking the subscription public while the rented channel gets more expensive), and the repetition is itself the data.
Cagan came at the same thread from the inside: he published today the list of what he argued in 2008 and no longer argues, and the hardest item is that the “why” he spent twenty years hammering was the wrong one — what matters isn’t why the company picked the problem, it’s why people use or drop the product, and almost nobody goes looking for whoever churned. Lookout Santa Cruz is the accounting version of that: six years, 60% ads, 40% memberships, profit. And the day’s counterpoint is the 351 ETF, the design where the gain comes from no relationship at all, it comes from reorganizing what you already own — and where sixty years of American rulemaking exist precisely to make sure that hurts.
launches
the economics of Nubank in the US — The bank announced its US operation targeting 100 to 150 million underbanked and, according to David Vélez, without niching down to Latin Americans: it will be mass market. The decision that matters is the cost one — 60 employees at the start, $1 of expense per customer against the $20 average at local banks, and revenue per customer projected at ten times that of the other countries because of exchange rate and income. Cristina Junqueira even puts the experiment’s budget in writing, a rare thing: about 1 point of the efficiency ratio, some $250 million, to find product market fit before growing. The entry product is pure price (an account paying 3.5% a year in a market where 60% to 70% of deposits yield nothing) and the global account across 35 countries is the bank saying out loud that the competitor is Revolut, not the incumbent.
product and growth
Porto is turning its brokers into influencers — The first class at Porto Academia had a thousand brokers adding up to 4 million followers; the goal is to train all 44,000 by 2030. The marketing budget split explains the thesis better than any speech: 35% on sponsorship, 17% to 20% on media buying, the rest on the broker. And the target isn’t reach, it’s cross-sell — in 2020, 900 of the 37,000 brokers sold multiple products, the goal is 10,000 this year, and the average products per customer should go from 1.1 to 2 by the end of 2026. For the broker with no camera skills there’s a generative avatar tool, and Luiz Arruda sums up the positioning in a line worth more than the deck: Porto stops being a content provider and becomes the channel’s content tool.
Strong Opinions, Loosely Held — Narrative version of a keynote where Cagan answers for what he argued in the first edition of INSPIRED (2008) and today considers a mistake. The first two sting anyone who builds: he underrated business viability to the point of burying it inside feasibility, and he chalks that up to his own career made in dev tools, one of the rare places where you can have weak business skills and survive. The third is the best — he tries products all the time, churns, and says almost no company ever comes looking for him to find out why. He also revisits the excess of problem discovery (arguing over whether it’s “the most important problem” is a question with no answer) and the appetite for predictability, with roadmap and PRD as the artifacts that make the team think it knows more than it knows.
brand and ip
Oatly Calls On Politicians To Address Climate Change On Its Packaging — Teaser: Oatly gave up a side panel of the carton, in Sweden, to The 89 Percent Project, a campaign by the journalism consortium Covering Climate Now anchored in the finding that 89% of people (80% to 89% depending on the country) want stronger government action on climate. Only the teaser is available, but the brand decision is already all there: packaging is the one media space the brand doesn’t rent from anyone, and Oatly is lending one side of it to a campaign that isn’t its own.
media and attention
“Many have said it couldn’t be done”: Lookout Santa Cruz is profitable — Six years after launch, Lookout turned a profit in 2026 on earned revenue: 60% advertising, 40% memberships, more than 40 marketing partners per market and average deals of $1,500 to $6,000. The number that teaches most isn’t the profit, it’s the second market’s — Eugene got to 60% of expenses covered by its own revenue in under two years, faster than Santa Cruz, which is the sign that there’s a playbook and not market luck. Ken Doctor, who was a media analyst and became CEO in 2020, is already announcing a third newsroom and a consulting arm for 2027. Neither leg closes the books on its own, and the bigger one is local advertising, which everybody had written off.
The Creator Economy keeps stealing away veteran journalists — The post’s open hook is Alex Heath, a veteran reporter who had already left for a Substack of their own and now becomes a partner at a VC firm while keeping the newsletter. Owens is blunt about the price: they covered the industry from an objective distance and now have skin in the game, so the reader has to recalibrate what to expect. What the press treats as a flight is, in practice, a swap of revenue model — the owned channel earns less than the deal flow it generates, and that’s why the math works. The main analysis is behind the paywall.
market and capital
351 ETFs: Tax-Free “Diversification” Is Supposed to Hurt — On July 21, at a Wall Street Tax Association meeting, the US Treasury said it may list certain Section 351 transfers into ETFs among the transactions that look “too good to be true”. The mechanism, in three pieces: 351 lets you transfer securities into a corporation and receive shares with no tax today; an ETF is a corporation for tax purposes; and 852(b)(6) lets the fund hand over appreciated securities in an in-kind redemption without recognizing gain. The authors — who disclose the conflict midway through the piece, one runs an ETF platform that does 351 seedings and the other is tax counsel on those deals — argue that sixty years of rulemaking aim at a single target: converting a concentrated position into a diversified one without paying. The part that teaches is the 1960 story, when a Denver banker pooled appreciated shares from several clients into a fund and everyone came out diversified with no tax; it raised $25.8 million, the imitators came, and Congress shut the door in 1966.
who wrote
AI Growth is Visible in GDP, If You Squint Right — Teaser: Hobart goes looking for AI growth inside the national accounts, and the title already warns how much squinting it takes. Only the summary is available (the issue also carries IP bootstrapping, provenance, state hypercapitalism and ads and routers). It gets linked because the exercise is the inverse of what circulates: capex numbers repeated with no denominator.
the rest in one line
Emily Sundberg logged the same reporter→investor pipeline on the same day, next to a NoHo bakery selling bread at $60 — only the teaser is available, and the two things are the same subject: price as a brand statement.
stalled sources
First Round Review at 319 days without publishing, Gurwinder at 257, Calculated Risk at 243, Matthew Ball at 207, Elad Gil at 144. My opinion: on fund and operator blogs, silence that long is a time allocation decision, not a pause — and it’s worth treating the source as closed until proven otherwise.