The brand age is your enemy — and how to opt out of it

Paul Graham published two essays in early 2026 that, read together, form an unusually sharp briefing for anyone building a startup right now. The longer one, "The Brand Age," is about the structural conflict between brand and product. The shorter one, "How to Convert Between Wealth and Income Tax," is a quick lesson in a math error that even sophisticated people routinely make. Both repay close reading.

Published March 2026 1

The essay starts with Swiss watches, but it's really about you.

Between the early 1970s and the early 1980s, unit sales of Swiss watches fell by almost two-thirds. The industry had been hit simultaneously by three compounding crises: Japanese manufacturers had caught up on quality and undercut them on price; the collapse of the Bretton Woods agreement sent the Swiss franc soaring (from 0.228 USD to 0.625 USD — Swiss watches became 2.7× more expensive for American buyers almost overnight); and quartz movements made accurate timekeeping a commodity. The two things watchmakers had competed on for a century — accuracy and thinness — were suddenly irrelevant.

The firms that survived did something drastic. They stopped selling engineering and started selling brand.

Graham traces exactly how this transformation happened — the Patek Golden Ellipse in 1968, the Audemars Piguet Royal Oak in 1972, the Patek Nautilus in 1976 — and what each move represented. The shift wasn't gradual; it was the deliberate, strategic decision to make watches large and visually distinctive instead of small and technically excellent.

The essay's analytical core is a single sentence:

Good product design converges. Watchmakers spent 500 years making watches smaller and thinner because there was a functional right answer and eventually most competitors reached it. The best mechanical watches of the golden age (roughly 1945–1970) were nearly indistinguishable from the outside — 32–33mm, minimal, quiet. Names were printed in letters less than a millimeter tall.

Branding by definition must diverge. To signal brand, you have to look different from competitors. Since competitors have already found the functional optimum, looking different means moving away from it. The Nautilus was 42mm and had gratuitous knobs on either side of the dial. It looked distinctive. It was not better engineered.

Graham is careful about the exceptions. Brand and good design can coexist in two narrow situations: (1) when the design space is enormous, as in fine art — there's enough room for many distinct "right answers"; (2) when the territory is largely unexplored and a first mover can claim the functional optimum as their distinctive identity, at least temporarily.

Outside those two situations, brand is only achievable by sacrificing design quality.

Graham doesn't apply the argument to startups in the essay, but the implication is close to the surface, and it's worth making explicit.

Early-stage startups almost universally benefit from being in a golden age — a period when smart people are working on interesting unsolved functional problems. The AI/LLM era has the hallmarks of one. The problems are hard and not yet solved. Competitive advantage still comes from doing the thing better, not from spending on advertising.

The risk the essay implicitly flags: brand pressure starts surprisingly early. The moment a founder starts asking "how do we make this feel premium?" or "what does our visual identity say about us?" before the product has a hard technical moat, they are drifting toward the Swiss watch path. Brand investment in a product-competition phase is a tax on the engineering work that actually compounds.

The harder corollary: you can't always tell which phase you're in. The Swiss watch industry didn't know quartz would arrive when it did. Founders who built competitive advantages in, say, "fine-tuning custom models" in late 2022 found the floor move under them when base models improved. Graham's prescription isn't to ignore brand entirely — it's the heuristic: follow interesting problems. If you're working on a problem that genuinely attracts talented, honest people, you're probably in a product-competition phase. If you notice most of your competitive differentiation conversation is about perception rather than capability, check whether you've drifted into the brand age prematurely.

If you're early-stage and feel pressure from investors or team members to invest in brand positioning: run the quick test Graham's essay implies. Ask whether your product has reached the functional optimum in its category or whether there is still meaningful differentiation available through capability improvement. If the answer is capability, defer brand investment. Patek Philippe didn't thrive by spending on advertising in 1970 — they spent on case design, which was a proxy for brand only in retrospect, after the performance argument became moot.

The advice is especially pointed for AI founders: the current frontier is not exhausted. The functional optimum for most AI applications is not remotely approached. Brand competition at this stage is premature enclosure.

Published May 22, 2026 2

This one is shorter — about 700 words — but makes a point that's relevant to any founder thinking about long-term capital formation, fundraising structure, or jurisdictional tax planning.

The core math: 1% wealth tax ≈ 20% income tax.

The conversion formula is straightforward. A wealth tax is assessed annually on the total stock of capital. An income tax is assessed on the flow — on returns. If the risk-free rate of return is 5%, then a 1% annual tax on wealth is equivalent, mathematically, to a 20% tax on the returns that wealth generates. (1% / 5% = 20× multiplier.)

Graham illustrates this with a worked example: $100 principal, 5% return, 20% income tax, and $100 principal, 5% return, 1% wealth tax. Both leave you with $104 after one year.

The policy implication is pointed. Denmark currently has the world's highest marginal income tax rate at 60.5%. The US federal top rate is 37%; the median US state rate is Oklahoma's at 4.75%. A state adding 20% in income tax would produce a combined marginal rate of ~61.75% — highest in the world. Politicians who describe a "mere 1% wealth tax" as a modest proposal don't know they're proposing the equivalent.

For early-stage founders, the immediate relevance isn't the political argument — it's the mathematical framework. Wealth taxes and income taxes are often presented as alternatives on a spectrum from "fair to capital" to "fair to labor." Graham's essay shows they're the same instrument, just expressed at different rates and measured against different bases.

The practical consequence: if you're making decisions about where to incorporate, where to base yourself, or how to structure carried interest or equity vesting in a jurisdiction that has or is considering a wealth tax, apply the 20× conversion before comparing rates. A jurisdiction offering a "0.5% annual net worth tax" with lower income tax isn't obviously better than one with a flat income tax — the equivalence matters.

Graham notes that 5% is an optimistic assumption for the risk-free rate; 4% may be more realistic, which would make the conversion factor 25× (1% wealth tax = 25% income tax at a 4% risk-free return), a stronger result than the headline number.

The two essays don't cite each other, but they share a method: find the conversion that most people don't know exists, state it plainly, and trace the implications.

In "The Brand Age," the hidden conversion is between "brand investment" and "product capability foregone." They look like different currencies — one is about identity and perception, the other about engineering — but in a field that hasn't yet reached its functional ceiling, they trade at a real exchange rate, and founders should know what it is.

In "How to Convert Between Wealth and Income Tax," the hidden conversion is literal: 1% wealth tax = 20% income tax at a 5% risk-free return. Politicians ignore it because the rate numbers look small. Founders who understand the math will make cleaner decisions about jurisdiction, capital structure, and exit planning.

Graham's output rate has slowed in recent years — these two essays represent the bulk of what he published between January and June 2026. Both are worth reading in full.