When the Product Stops Mattering: What the Quartz Crisis Teaches AI Founders

Paul Graham's longest essay in years, published March 2026, opens with a Swiss watchmaking disaster and lands somewhere startling: we are living in the Brand Age, and AI is about to make it permanent. The implications for early-stage founders building on top of commodity intelligence are sharp enough to cut.

The argument in one paragraph

Brand is what's left when substantive product differences disappear. Technology is what makes substantive product differences disappear. Therefore, in a world of fast-improving technology, brand accumulates power at an accelerating rate. This is the central claim of Graham's "The Brand Age," and it is not a theory about luxury goods. It is a theory about what happens when your product edge gets commoditized. 1

The quartz crisis is already happening in AI

In 1968, the Swiss watchmakers thought they were in the precision-instrument business. They were wrong. Japanese manufacturers matched their accuracy, Bretton Woods collapsed and made Swiss watches three times more expensive, and then quartz movements made accuracy a commodity altogether. Between the early 1970s and early 1980s, unit sales of Swiss watches fell by almost two-thirds. Most went insolvent. 1

The ones that survived did so by becoming something different: luxury brands. And here is the uncomfortable part — the transformation was not about making better products. It was about accepting that the product race was over.

For AI founders building on top of foundation models in 2026, the parallel is uncomfortable close. Model capabilities are converging. The API costs for the same intelligence are falling roughly 10x per year (a figure Sam Altman cited in his September 2025 post "The Gentle Singularity" 2). Whatever performance advantage you have against a competitor today will likely be erased within 12-18 months — either by your own model provider's next release or by a competitor's switch to the same base model. The quartz movement is already here; you just haven't built the Nautilus yet.

The fatal path: Omega

Graham uses Omega as the cautionary tale. Omega were, in his words, "the nerds of Swiss watchmakers." When the Japanese matched their accuracy, Omega responded in the Omega way: build an even more accurate movement. The new movement ran at 45% higher frequency. In theory: more accurate. In practice: so fragile it destroyed their reputation for reliability.

"They even tried to make a better quartz movement, but there was nothing down that road but a race to the bottom. By 1981 they were insolvent and were taken over by their creditors." 1

For founders, this is the most important failure mode to recognize. The Omega trap is not stupidity — it's the logical extension of a growth strategy that worked until the day it didn't. If your moat is "we have the best model" or "we have the fastest inference" or "we have the cheapest tokens," you are Omega. Doubling down on a performance dimension the market is about to commoditize is not a moat; it is a countdown timer.

The specific mistake is not choosing performance as a dimension to compete on — it's choosing only performance, which means you have no second leg to stand on when the first one gets kicked out.

The survivable path: brand is centrifugal, design is centripetal

Graham's cleanest formulation of the core tension: Branding is centrifugal; design is centripetal.

Good design seeks the right answer, and right answers converge. If you find the optimal way to display a metric in a dashboard, every dashboard will eventually look like yours. Branding, by contrast, requires distinctiveness — which means deliberately choosing a non-optimal path to stay different. You can't have both, except in two narrow cases: when the design space is unexplored enough that the right answer is also unique (early product categories), or when it's vast enough that many right answers coexist (fine art, architecture).

Most AI applications are neither. The design space of "AI writing assistant" or "AI coding agent" or "AI analytics" is not vast, and it's not unexplored. The right answers are converging. Which means that any brand advantage you build now — before your product feature set becomes identical to everyone else's — is more durable than any performance advantage. 1

The question this raises for a founder is not "how do I maintain product differentiation?" The question is: "When the product differences disappear, what do users choose us for?"

What Patek did right (and why it took 20 years)

The surviving Swiss watchmakers did not pivot cleanly. Graham documents the awkward middle period: Patek Philippe hedging with both quartz and mechanical watches, arguing in ads that quartz watches in fancy cases were "almost as laborious to make as mechanical ones." This was them trying to have it both ways — and it didn't work.

What finally worked was committing completely to the brand story. The Calatrava "banker's watch" succeeded not because Patek found a better movement. It succeeded because a receptive audience — the ibankers of the 1980s who were being newly minted in large numbers and needed visible ways to display wealth — was ready to buy into the full narrative. 1

Two things are worth noting here for founders:

The 20-year lag is a warning, not a comfort. It took the watchmakers two decades to fully internalize the new rules. Companies that start building brand identity before commoditization hits will have a substantial head start. The ones that wait until feature parity to start thinking about brand will be building it into a headwind.
Audience readiness matters. Patek's pivot worked partly because the right audience (newly wealthy young professionals who wanted visible markers of status) was growing rapidly. Founders should ask: who is the audience that will choose brand over spec, and is that audience growing?

A skeleton pocket watch circa the golden age: the movement was the point. After the quartz crisis, buyers stopped caring about what was inside. 1

Back case of a Swiss mechanical watch, showing the intricate movement — a Speake-Marin Piccadilly with SWISS MADE engraving — The back case of a Swiss mechanical watch, showing engineering that became invisible to buyers once branding took over 1

Sam Altman on the "ring of power" dynamic

Graham's essay is about products. A post published by Sam Altman on April 10, 2026, in the wake of a Molotov cocktail thrown at his home, addresses the organizational dynamic that makes the brand-vs-commoditization problem even harder. 3

blog.samaltman.comhttps://blog.samaltman.com/2279512External link

Loading content card…

Altman's central observation about the AI industry: "Once you see AGI, you can't unsee it. It has a real 'ring of power' dynamic to it, and makes people do crazy things."

The specific crazy thing this dynamic produces, in his telling, is the desire to be the one who controls AGI. He describes this as the actual driver of conflict inside and between AI labs — not disagreements about safety approaches or product strategy, but a totalizing philosophy of control.

His proposed solution: "Orient towards sharing the technology with people broadly, and for no one to have the ring." Concrete mechanisms: individual empowerment and keeping democratic institutions more powerful than companies.

For early-stage founders, the actionable version of this is narrower and more immediate. The "ring of power" dynamic plays out at startup scale too — in the tendency to build systems where all capabilities, data, and user relationships are locked into your platform. This creates a short-term moat and a long-term brittleness. When the capabilities commoditize (they will), the locked-in integrations become friction, not value. Founders who build on top of AI while maintaining user agency and data portability are better positioned for a world where the model layer becomes infrastructure.

Altman is also unusually candid about his own version of the Omega trap: conflict-aversion. "I am not proud of being conflict-averse, which has caused great pain for me and OpenAI." The specific failure mode he describes — avoiding the hard conversation until it becomes a structural crisis — is recognizable to anyone who has run a team. 3

What "The Shape of the Essay Field" adds

Graham's shorter June 2026 essay is a meta-observation about writing, but it has a direct application to content strategy for founders. 4

His core claim: if you write for smart people about important things, you are functionally writing for young people. Not because older people aren't smart, but because the space for changing someone's thinking is larger the less they've already thought about a given topic. "There's more room to change their thinking, so there's a bigger payoff for writing about important things."

For founders building in AI, this observation maps cleanly onto the early-adopter dynamic. The most receptive audience for a new kind of AI tool is not the domain expert who has already settled into workflows and mental models. It is the younger or newer practitioner who hasn't yet locked in. This has been true in every previous wave of developer tooling. 4

Applied questions for the week

Question	Context
On what dimension does your product differ from competitors that cannot be replicated by training a better model on the same task?	The Brand Age: identifying the second leg
If a competitor had identical features at half the cost in 18 months, what would still make someone choose you?	Brand audit for commoditization pressure
Who is the growing audience for whom brand preference matters more than marginal performance differences?	Patek-style audience mapping
Are you in the middle period — hedging between performance and brand story, arguing that your AI is "almost as good"?	Omega avoidance check
In your team, which hard conversations are being avoided until they become structural problems?	Altman's conflict-aversion pattern

Sources used this week: Paul Graham, "The Brand Age" (published 2026-03); Paul Graham, "The Shape of the Essay Field" (published 2026-06); Sam Altman, untitled personal post (published 2026-04-10).