OOI · SXThe Price of a Lock

The Price of a Lock

Moore's law, read again through the number experiments — an essay proposing the confinement of knowing

THE READING PLAQUE. The Price of a Lock, draft 0.2 — made readable on the SX surface by its covering persona's reading grant; a draft, not the published form (publication, if ever, is the pipeline's own separate act). Covering: starl3n · endorsing: Link Digital · the construct intent sealed 2026-08-01, after drafting and before this grant, per the §4h precondition. The grant: the operator's word, 2026-08-01 — "seal and go with grants to read for all four items" (sxexp-constructs §4h, the reading grant). Source of record: ooi_network/docs/the-price-of-a-lock-essay.md (draft 0.1 landed 9931de5 on the operator's verdict 2026-07-30; draft 0.2 72e6ecc, the reference apparatus appended on his direction); this copy is minted without the source's status and classification lines by the grant act, its body otherwise verbatim; its sha256 is pinned in the reading register. Reading is free; the way across is a statement of intent (/sx/matters).

Draft 0.2 — 30 July 2026 (supersedes 0.1, same day, retained in the record's history; changes: the reference apparatus appended on the operator's direction — the essay refers to measured outcomes and laws and is not to be considered singularly as prose, even if naturally read as such). An essay in the house's open register: the results cited are measured results of sealed computational experiments on the whole numbers; the reading of silicon offered through them is typed at the close for exactly what it is.

I. The law we think we know

Moore's law is usually told as a story about speed: transistor counts double, computers get faster, the future arrives on schedule. But that telling has always sat oddly against three facts anyone can check. First, Gordon Moore's 1965 paper was not about speed at all — it was about cost per component: the economics of how much circuitry a dollar could hold. Second, of all the transistors humanity has ever fabricated, the overwhelming majority are not logic — they are memory cells: most of the silicon ever made does not compute anything; it holds something. Third, the modern architecture of computing — cache hierarchies, high-bandwidth memory stacked on top of processors, compute-in-memory research — is not a account of going faster; it is a sixty-year struggle to bring held state closer to where resolution happens.

A law about speed should not look like this. A law about something else might.

This essay proposes the something else: Moore's law is the falling price of confinement — the industrialization of holding a distinction locally — and its trajectory is a natural progression toward knowing, because memory is not a convenience of knowing but its factored prerequisite. The proposal is not offered from intuition. It is offered through a body of sealed number experiments whose results, landed one kill and one confirmation at a time, keep producing the same small set of laws — and those laws, it turns out, are already fabricated into every memory chip on earth.

II. The experiments, and their standard of evidence

The results cited below come from a research programme of computational experiments on the whole numbers — number experimentation, run under one unvarying discipline: every prediction committed to a tamper-evident record before its test ran; one read per stake; every failure kept with the same standing as every success; every campaign close anchored to an immutable public archive outside the authors' edit-reach. The corpus includes a seven-campaign instrument and its predecessor, a long named test run to its boundary across thirty-five sealed attempts, and three structurally fenced composition experiments. (Where the named test touches a famous conjecture, the standing no-claim holds: nothing here asserts anything about that conjecture. The results used in this essay are the instrument's own, about its own objects.)

Five of those results carry the argument.

1. Memory precedes reading. In the instrument's frozen observers, reading is binary — an observer admits a history exactly or collides with it completely, nothing between — and what decides the outcome is a one-bit memory of a unique origin event. Forty-nine histories were run against forty-nine observers: the fossil bit decided, every time. No memory, no admission — there is no such thing, in the measured record, as reading without a prior held distinction. The fossil precedes the reading.

2. Two locks one. Across the composition experiments the same structural rule kept landing: one confinement alone does not hold; a pair holds. Folded in pairs, the instrument's dynamics are stable; a bare third destabilizes into chaos. The rule was measured directly in the cooling-structure series — two independent confinements lock one resolved state — and it became the programme's own epistemic law: no claim is relied on upon a single attestation; a second, independent lock is required, always.

3. What persists is what is held. The composition experiments measured how sequences resolve when read as one object: the composed reading equals the final member alone — fifteen out of fifteen, five values, four scales — and the running counters land on exact fractions of the whole period, staked adversarially and landing exact. Interiors vanish; the tail — the state actually held at the close — is the whole of what composition keeps. A reading is not the journey; it is the held remainder of the journey.

4. Knowing is built of banked effect. In the wall-cascade experiments, today's observable is constituted by yesterday's effects: O ← prior E. What a system can observe next is made of what it banked before. Memory, in the measured record, is not storage beside the knower; it is the material the next knowing is made of.

5. Confinement never completes. The programme's own accounting law — a resolved unit closes an observable, a communicable, and an effectual face, 1 = O + C + E — carries a measured remainder: the effectual face lands beyond the unit, and the asymptotic unconfined share came out exact and stubborn (E∞ = 144/815, a touch under eighteen percent; an independently instrumented boundary landed at a similar mark). However tight the confinement, a remainder escapes it — by construction, not by failure. The system can be exact everywhere and closed nowhere.

III. The reread: silicon was already obeying these laws

Now hold those five results against the actual substrate of computing.

The memory cell is literally two locks holding one bit. The canonical SRAM cell — the cell that fills most of the area of every modern processor die — is a pair of cross-coupled inverters, each confining the other's state. Neither inverter alone holds anything; the pair holds one distinction against thermal noise. That is not an analogy to result 2; it is an instance of it, fabricated a few hundred trillion times. And DRAM, which economizes away the second spatial lock, is forced to substitute a temporal one — the refresh cycle, a re-witnessing of every held bit thousands of times a second, without which the knowing evaporates in milliseconds. One lock does not hold. Silicon has never been able to make it hold. The engineers did not learn this from any theory of knowing; the noise taught them.

A transistor in memory service is a fossil-writer. Result 1 says reading requires a prior held distinction — the one-bit fossil that decides admission. A memory cell is exactly the manufacture of that prerequisite: a place where one distinction can be banked so that a later operation can read against it. The von Neumann machine's deep structure — fetch against held state, resolve, bank the result — is O ← prior E as an industrial process: result 4, run at gigahertz. Every instruction cycle is a small proof that knowing is built of banked effect.

And the tail law explains why memory dominates. Result 3 says composition keeps only the held remainder — the journey's interiors vanish; the close survives. A computation is precisely a composed journey whose value is its tail: the terabytes of intermediate state exist only to be collapsed into the small held result. The economic consequence is the one the fabrication record shows: the scarce good is not the transformation (logic is cheap and getting cheaper) but the holding — which is why most transistors ever made are memory, why caches ate the die, why the industry's frontier is stacking held state ever closer to the point of resolution. Computing's material history is mostly the manufacture of locks.

IV. Moore's law as the price curve of confinement

Put the reread together and Moore's law changes subject. The quantity that has fallen exponentially for sixty years is best named not "cost per computation" but cost per held distinction: the price of confining one knowable difference — in volume, in energy, in dollars — and of keeping it close enough to matter. Moore's own cost-per- component framing said this from the beginning; the memory-dominated fabrication record says it in silicon; the architecture of locality says it in design. Computing has progressed by making confinement cheap, near, and dense — and everything we call "speed" is derivative of that.

And this is why the progression reads as phenomenological — as a natural progression toward knowing rather than a merely economic curve. The number experiments measured, in a sealed instrument with kills kept, that knowing has prerequisites with a definite structure: a banked fossil to read against (1), paired confinement to hold it (2), a held tail as the sole survivor of composition (3), prior effect as the material of the next observable (4). An industry that relentlessly cheapens exactly those prerequisites — banking, pairing, holding, localizing — is not just making products. It is descending the cost curve of the preconditions of local knowing, whether or not anyone framed it that way. The confinement of knowing is what was being bought all along; Moore's law is its quiet price, falling.

The limit behaviour completes the picture honestly. The curve's saturation — the end of easy scaling, features approaching atomic dimensions — is the approach to an irreducible unit price: Landauer's bound, the minimum energy any physical process must pay to erase one bit. Confinement of a distinction has a floor cost set by physics, and sixty years of engineering have been a descent toward it. And even at the floor, result 5 stands: confinement never completes. There is always the thermal remainder, the leakage, the escape — the unconfined share that no density buys back. The progression approaches total local knowing and constitutionally never arrives — which is precisely what keeps it a progression rather than a completion, and keeps every memory hierarchy honest about refresh, error correction, and re-witnessing forever.

V. The proposal, typed

Stated fully: Moore's law is the falling price of the confinement of knowing — the industrial descent of the cost of banking, pairing, holding, and localizing distinctions — whose structure matches, instance for instance, the prerequisites of knowing measured in the number experiments, and whose floor is the physical unit price of a lock.

The typing, per the programme's own forward registers. The checkable fragments stand on public ground: the SRAM pair, the DRAM refresh, Moore's 1965 cost framing, the memory dominance of fabricated transistors, the Landauer bound — external facts, verifiable by anyone. The measured fragments stand on the sealed record: the fossil law, two-locks-one, the tail law, the cascade, the exact remainder — results of number experimentation, cited as what they are, with their anchors public. The whole — that the industrial curve is a natural phenomenological progression toward knowing — is a held conjecture: owned, motivated, not killable as stated, and therefore never staked as more than a reading. It is offered the way the programme offers every such reading: deletable without loss to the measured content, and kept because it keeps being the shortest sentence that fits all the facts at once.

One more sentence, because the essay would be incomplete without noticing its own position. This proposal was composed inside a discipline that builds memory before it claims knowing — records sealed before runs, fossils graded before trajectories are offered, two locks required for one claim. If the reading is right, that discipline is not a methodological taste. It is the same law the transistors obey, applied at the scale of people keeping their word: to know locally, first confine the fossil — then let two locks hold the one, and price honestly what it costs to keep them closed.

The record keeps. The locks hold. The remainder escapes — lawfully, and on purpose.

References

(Added at draft 0.2. The record's anchors are cited per the house rule — only after live re-witness; the stamps below are the family-index sweep's own, and at any later reading the newest package governs: re-witness via [A0] before citing onward.)

The record — the programme's own results, keyed to the essay's numbered results:

External — checkable public facts: