This Week in Founder Essays: When Technology Makes You Irrelevant

May 12–18, 2026 — Two essays from the Valley's most-read founders landed this week, and they accidentally answer the same question: what do you do when execution stops being the moat?

Paul Graham looked backward — to the Swiss watch industry's collapse in the 1970s. Sam Altman looked forward, to a decade where intelligence costs as much as electricity. Neither was explicitly writing to AI founders. Both probably should have been.

Paul Graham: the brand age isn't optional

Graham's new essay, "The Brand Age," traces the triple disaster that hit Swiss watchmakers between 1968 and 1973: Japanese competitors who first made cheaper watches, then better ones; the collapse of Bretton Woods, which sent the Swiss Franc from .228 to .625 USD against the dollar in five years, making Swiss watches 2.7 times more expensive for American buyers overnight; and quartz movements that obsoleted both of watchmaking's core goals — accuracy and thinness — in one stroke. 1

The watchmakers who survived — Patek Philippe, Audemars Piguet, Vacheron Constantin — did so by spending about two decades figuring out that they were no longer in the precision instruments business. They were in the brand business. The ones that didn't figure this out, like Omega, competed harder on performance, and went insolvent by 1981.

Graham's central argument is structural, not sentimental:

"Brand is what's left when the substantive differences between products disappear. But making the substantive differences between products disappear is what technology naturally tends to do."

This isn't a watch story. It is, as Graham writes, "very much a story of our times."

What the brand pivot actually required

The watchmakers' transition wasn't just a marketing reframe. It required physically redesigning watches to be visibly, distinctly branded — trading functional optimality for recognizability. Patek's 1968 Golden Ellipse expanded the brand's visual footprint from the 8mm² of a tiny dial inscription to 800mm² of distinctive case design. The Royal Oak (Audemars Piguet, 1972) wrapped an integrated bracelet around the entire wrist. The Nautilus (Patek, 1976) went to 42mm when golden-age dress watches topped out at 33mm — because it could be spotted from across the room.

Each design sacrificed something optimal. The Golden Ellipse's crown was made too small for the distinctive silhouette, making it distractingly hard to wind. Graham is direct about the tradeoff: "Branding is centrifugal; design is centripetal."

The yardstick for success also changed. The best watches of the golden age "have a quiet perfection that has never been equalled since." They are nearly indistinguishable from each other, because they all converged on the right answer. Present-day watches are instantly recognizable, because they deliberately avoid the right answer.

For AI founders: which era are you in?

The essay's sharpest implication isn't about watches. It's about timing. The watchmakers who survived were the ones who read the signal early — before the quartz crisis fully hit — and began building brand equity while they still had performance credibility to attach it to.

Omega read the same signal and bet on building better movements. By the time they realized the game had changed, their new movement was too fragile for their reputation, and they had nothing else.

For AI founders in 2026: the performance commoditization Graham describes is already visible at the model layer. Foundation model capability gaps between frontier labs have narrowed from months to weeks. If your startup's moat is "we're better at X than GPT-4" — ask when GPT-5, or Claude 4, or Gemini Ultra, will close that gap. And ask whether you've started building the brand layer that persists after performance parity arrives.

The brand-building window for most AI application categories is probably shorter than the watchmakers had.

Sam Altman: the idea guys are getting their revenge

Altman published "The Gentle Singularity" in June 2025, and it is his clearest statement yet about what the current phase of AI development means for the people building on top of it. 2

His core claim: we have passed the event horizon. The key scientific insights that unlocked GPT-4 and o3 are "hard-won, but will take us very far." The unknowns are shrinking. His timeline:

2025: AI agents capable of real cognitive work (code, in particular, is already changed)
2026: AI systems generating novel scientific insights
2027: Humanoid robots capable of real-world tasks
2030s: Intelligence and energy become "wildly abundant"

He adds the cost-structure point directly: as datacenter production automates, "the cost of intelligence should eventually converge to near the cost of electricity."

The inversion that matters

The line that should stick for any early-stage founder is this one:

"For a long time, technical people in the startup industry have made fun of 'the idea guys'; people who had an idea and were looking for a team to build it. It now looks to me like they are about to have their day in the sun."

Altman is saying the bottleneck has shifted. When execution capacity was scarce — when building required a team of engineers, product designers, and months of development — good ideas without technical co-founders died in the planning stage. That scarcity is eroding fast. The bottleneck is moving to the question, not the answer.

This is the flip side of Graham's essay. Graham says: when technology eliminates performance differences, you need brand. Altman says: when technology eliminates execution differences, you need ideas. They're pointing at the same structural shift from two angles.

What this doesn't mean

Altman also clarifies what the "idea guys have their day" claim doesn't mean: "experts will probably still be much better than novices, as long as they embrace the new tools." The advantage doesn't disappear for technical founders — it changes shape. The founders who lose are the ones treating their ability to build as a static, unconditional moat. The founders who win are the ones who use AI to multiply their output and spend the freed-up capacity on sharper questions.

He identifies two unsolved problems that matter structurally: alignment (getting AI to act toward what we "collectively really want over the long-term," not just short-term engagement) and distribution (making superintelligence cheap, widely available, and not concentrated). Both are startup-shaped problems.

Putting them together

Read in sequence, Graham and Altman are running a relay. Graham shows what happens to companies that compete purely on technical performance when technology commoditizes it — they become Omega, insolvent by 1981, even though they were building objectively better movements. Altman shows why 2026 is the year that transition becomes acute for AI software companies: the performance ceiling on execution is rising rapidly while the cost of reaching it is falling.

The operative question for an early-stage AI founder this week isn't "are we building a good product?" It's two questions: What will be true of your product when frontier model performance gaps are gone? And what question is your startup asking that no one else is asking?

The watchmakers who answered those questions early are now the most sought-after brands in the world. The ones who didn't went insolvent.

Sources reviewed this week: Paul Graham, "The Brand Age" (March 2026); Sam Altman, "The Gentle Singularity" (June 2025).