First Principles and Structural Memory.

On September 28, 2008, the fourth Falcon 1 reached orbit. The three before it had failed, and SpaceX had cash for little beyond that single attempt. Three months later, on December 23, NASA awarded the company a $1.6 billion contract to resupply the International Space Station. The next day—Christmas Eve—Tesla closed a $40 million convertible financing hours before it would have missed payroll.

Nothing about the physics of rockets or lithium-ion cars changed that winter. The equations of September were the equations of December. What changed was the environment around the engineering: a customer arrived, capital arrived, and two companies crossed from one side of a closing funding window to the other. Two kinds of truth were operating at once—one about what is possible, one about what is survivable—and neither could substitute for the other.

This essay is about the relationship between those two truths, and the two thinking instruments they require.

The power of deleting history

First-principles thinking, properly defined, means reasoning upward from established facts rather than sideways from precedent. Its most visible practitioner, Elon Musk, describes it as boiling a problem down to fundamental truths and rebuilding from there, instead of reasoning by analogy. Applied to cost, the method is ruthless: a product is a bill of materials, plus energy, plus labor. If the raw materials of a rocket are a small fraction of its market price, the remainder is organization, habit, and margin—which is to say, an invitation.

The method deserves its reputation. Launch costs and battery costs did not fall because someone extrapolated a trend line; they fell because someone refused to accept that the prevailing price reflected a physical constraint. “That is how it has always been done” carries no information about what is actually possible, and first principles is the discipline of ignoring it.

Notice what gives the method its power: it deletes history deliberately. Precedent is treated as contamination—a record of other people’s constraints, not yours. On engineering questions, that stance is correct. The tensile strength of an alloy does not care what the industry believes about it.

What the method cannot see

But first principles interrogates the object. It has nothing to say about the environment the object must survive in: the funding cycle, the competing narratives, the depth and mood of capital, the regulatory temperature, the market’s current standard of belief. These are not engineering variables, and no amount of decomposition into materials and energy will reveal them.

Markets are not physics. They have no conservation laws. They are coordination systems built from expectation, imitation, and memory—reflexive systems in which the observers are also the participants, and in which believing something can temporarily make it true. In physics, history is noise around the signal. In markets, history is the mechanism: the 1930 bank run and the 2023 bank run are the same machine wearing different clothes, and the machine only exists because participants remember, anticipate, and imitate one another.

In physics, history is noise. In markets, history is the mechanism.

This blindness produces a specific failure mode: technically correct, financially dead. The graveyard of startups is full of working products. A first-principles case can be entirely right about the decade and still be wrong about the two years a company can actually finance. Being right about the destination has never guaranteed that a particular vehicle arrives—a lesson the railway and fiber investors of earlier cycles paid to learn on everyone else’s behalf.

Structural memory

The counterpart instrument needs a name, so call it structural memory: the discipline of remembering relationships rather than events. Not chart patterns, not “what happened to a company like ours”—relationships. Investment cycles create capacity and then correct. Narratives organize attention and then exhaust themselves. Liquidity expands until it becomes the only variable that matters. I have written about why these relationships persist beneath every technology wave and how to read their alignment; the point here is what kind of tool that reading is.

Structural memory consults history for exactly what first principles discards: what each phase of a cycle is willing to fund, how belief spreads through a market, how long capital waits, and what happens when it stops waiting. It does not predict events. It recognizes configurations.

The symmetry between the two instruments is easy to miss because they point in opposite directions. Both are anti-consensus tools. First principles refuses the consensus of practice: “everyone builds it this way.” Structural memory refuses the consensus of novelty: “this time is different.” They are different heresies, but they demand the same independence of mind. One deletes precedent to find what is buildable. The other consults precedent to find what is fundable, adoptable, and timeable.

The founder needs both

Sequence matters. First principles sets the destination: what should exist, and what it must eventually cost. Structural memory sets the route and the departure time: which phase of the cycle will reward the product, which narrative can carry it to the people who need to believe in it, and whether available liquidity can bridge the distance between prototype and adoption.

Return to the winter of 2008. The engineering that reached orbit was first-principles work—a launch vehicle designed from physics up, at a fraction of incumbent cost. But what saved the company was structural: an anchor customer and a financing that arrived before the window shut. Tesla’s powertrain did not rescue Tesla that December; a convertible note did. In both cases the product was necessary and insufficient. The physics had been true all year. Survival was decided by the funding environment, in the same week credit markets were deciding the fate of far larger institutions.

Each instrument, alone, has a characteristic failure. All structure and no physics produces derivative companies chasing whatever narrative happens to be liquid—an archetype every crypto cycle manufactures in bulk. All physics and no structure produces brilliant prototypes that arrive too early and die in the gap between installation and deployment, when the correction empties the funding market precisely as the technology begins to work. The first failure wastes a cycle. The second wastes a breakthrough.

For a founder, the division of labor is practical. Use first principles on the product: cost floors, performance ceilings, what the incumbent price structure is actually made of. Use structural memory on everything that surrounds the product: when to raise and how much, which story to tell to which capital, what evidence this phase of the market demands, and what would signal that the window is closing. The two questions belong in every serious diligence process, and they are not the same question: What do the physics permit? And: What does this moment reward?

Delete history to design the product. Consult history to survive the market.

Neither instrument is a temperament, though each attracts one. The engineer’s contempt for precedent and the historian’s respect for it are usually found in different people, which is one reason founding teams need more than one kind of mind. The rarest founders hold both at once: enough disregard for how things have been done to build something genuinely new, and enough regard for how markets have always behaved to keep that new thing alive until the world is ready to pay for it.

Sources & further reading

The 2008 events are documented in Tesla’s announcement of its December 2008 financing and the GAO’s 2009 review of NASA’s commercial resupply contracts. The installation-and-deployment framing follows Carlota Perez. This essay bridges two companion pieces on this site: one on why market structure persists, one on how to read it.