Durable Consumables

Twenty-three years ago, my family bought a Land Cruiser. The leather seats have worn smooth in places now. The engine, with 190,000 miles on it, still turns over with the same solid certainty it always has. Everything about how Toyota built that truck assumed decades of use. You can feel it in the weight of the doors, in the materials they chose, in the way it's aged. The truck didn't come with a software update cycle. It came with the implicit promise that it would outlast us.

The Tesla in my driveway operates on a different premise. It has a cellular service plan. It has Full Self-Driving as a subscription. Even ownership feels less like possession and more like a temporary claim on a particular moment in the capability frontier. I expect to own it maybe four years. Not because something will break, but because by then it will feel behind. A friend once said driving a Tesla feels like driving a MacBook. Capable, frictionless, soulless.

Economists distinguish between an asset's physical life and its economic life. The physical life is how long something can function. The economic life is how long it remains useful relative to alternatives. Technology products increasingly have economic lives far shorter than their physical lives. The thing still works. It's just been obsoleted.

This is spreading. We buy homes now under the assumption of a five to seven year hold. We finance cars over 68 months while the vehicles themselves face functional obsolescence in 48. Data centers depreciate GPUs over six years, but Jensen Huang joked that when Blackwell ships, "you couldn't give Hoppers away." The hardware doesn't wear out. It gets replaced by the frontier moving forward. I've often wanted a wayback machine for LLMs just to capture the rate of depreciation. Models that felt cutting-edge six months ago now seem quaint.

Stewart Brand has this framework called pace layers. In a healthy civilization, different layers move at different speeds: fashion moves fast, then commerce, then infrastructure, then governance, then culture, then nature. Fast learns, slow remembers. Fast proposes, slow disposes. The slow layers provide stability while the fast layers innovate. Problems arise when one layer tries to move at another's pace.

What we're seeing now looks like commerce pushing into infrastructure's territory. GPUs are infrastructure-layer investments being made at commerce-layer speed. The financial structures assume one rate of change while the technology assumes another. We're buying durable goods on installment plans calibrated for permanence while the goods themselves are calibrated for replacement.

Part of this is just math. There's a concept called the discount rate: how much less we value future benefits compared to present ones. High discount rates favor short-term returns and penalize long-duration investments. When you expect rapid progress, the rational discount rate rises. Why invest in something built to last thirty years when the version ten years from now will be meaningfully better? The math says: don't.

But the math doesn't capture everything.

When Patrick Collison noticed the Woolworth Building in New York, he remarked on its gargoyles, its spandrels, its full Gothic ornamentation. It was called "the cathedral of commerce." People built it to be beautiful and to last. That assumption of permanence shaped every decision. The intention shows in the thing itself.

The Woolworth Building's gargoyles and the Land Cruiser's drivetrain are different expressions of the same impulse. One is ornamental, the other functional. But both emerge from a shared assumption: that what you're making will be around long enough to warrant the extra care.

Objects built to last carry a different intention than objects built to be replaced. You can feel that intention. The Land Cruiser's doors close with a weight that communicates something about how it was made and for whom. The materials Toyota chose, the tolerances they held, the decisions about what to over-engineer. All of it assumed decades of use. That assumption is legible in the object. And there's something else embedded in it: pride in the making. When you build for permanence, you build with the knowledge that your choices will be lived with, tested, revealed over time. That's a different kind of work than optimizing for next quarter's margin. It produces what Adam Gopnik once called "delight in the ordinary," the quiet satisfaction of things made with more care than they strictly require. Engineers and programmers feel this too. When you don't expect a feature to last, when software is ephemeral, you build differently.

When you build something expecting it to be obsolete in five years, you make different choices. You optimize for capability now, not presence across time. You don't invest in the qualities that let objects age well, gather patina, become companions. You don't build cathedrals of commerce.

Not everything suffers from this, of course. The TV in my living room is better than anything available ten years ago. Thinner, sharper, cheaper. Software improves precisely because it ships and re-ships: bugs get fixed, features evolve, it gets better in the background. Rapid iteration serves some categories well. The question isn't whether permanence is always better. It's what we lose when nothing is built to last.

Maybe the strangest thing is what's happening in software now. The tools are starting to remember. They learn what you care about, how you think, what you've said before. Use one long enough and it begins to feel less like a product and more like a place you've been building. The value isn't in the model. Rather, what accumulates between you and it: the context, the files, the scaffolding of what you represent back to it, your "world model." The underlying model keeps getting replaced. And what accumulates lives somewhere you don't control. We're building patina on borrowed surfaces. Investing in permanence we don't quite possess.

Ross Douthat, when his daughter asked why people don't build beautiful things anymore, offered two explanations: ideology changed, and labor costs rose. But there's a third: when expected lifespan shortens, investment in beauty stops penciling out. Beauty was always partly an investment in permanence. If the permanence assumption disappears, the beauty rationale weakens. And so does the rationale for craft, for over-engineering, for any care beyond the immediate.

We live differently in cities when we expect to move every few years. We relate differently to objects when they're subscriptions rather than possessions. The psychologists say our possessions become extensions of self, that we develop attachment to things simply by virtue of owning them. What happens to that attachment when ownership becomes access? When the objects in our lives are designed as temporary platforms rather than durable relationships?

I keep coming back to that Land Cruiser. The worn leather, the engine that still runs true, the way twenty-three years of use have given it character rather than just age. It carries a kind of presence that feels increasingly rare. Not nostalgia exactly. Evidence that things can be built to last, to age, to remain.

The frontier keeps moving. Progress keeps accelerating. The rational response is shorter time horizons, faster replacement cycles, less investment in permanence. I understand the logic. I just wonder what we lose when fewer objects around us are built with the assumption that they'll still be here, still working, still ours, decades from now.