A register is what separates a monitoring instrument from commentary. Anyone can say a number moved. The register states, in advance, what movement would count, which direction the evidence can actually run, and the date FP1 reports what happened. Nothing here is graded after the fact, and nothing is re-thresholded mid-window.

How to read this

Each entry names four things and nothing more: what is tracked, the threshold that counts as movement, the direction in which the dimension can be proven, and the grade date. Readings are stated as of the pipeline's most recent weekly close, which lags the publication date.

The direction line matters more than it looks. Most of what FP1 tracks can resolve in only one direction: a capability can be shown to exist but its absence can never be shown; a constraint can be shown to hold, while its past failure proves nothing about the future. Where a dimension is one-way, the absence of evidence is logged as unresolved, never as a weak negative reading.

If a threshold turns out to be wrong, it is not quietly improved. The signal is voided in public, in the Reading, and re-registered with a new date.

Tracked dimensions

1. The FNC-1 / SOX spread

FieldValue
What it tracksWhether capital moving toward the companies actually building the AI buildout is diverging from the chip trade.
Reading−80.9 points, as of the weekly close of 20 July 2026.
ThresholdCloses to −40 points, or widens beyond −120 points.
Provable directionBoth. The spread is a measured quantity and can move either way.
Grade date31 August 2026.
Stated limitationThe FNC-1 proxy basket rebases on a rolling twelve-month window, so the spread is comparable across issues only approximately. The Tier 2 expansion replaces this with a fixed base date; until it ships, this limitation is carried with every reading of this dimension.

2. The Belief Index

FieldValue
What it tracksHow strongly prediction markets price the continuation of the AI boom, on a 0–100 acceleration scale.
Reading49.7, as of the weekly close of 20 July 2026. Below 50 since June.
ThresholdTwo consecutive weekly readings above 55, or any single reading below 40.
Provable directionBoth. The index is a measured composite of live market prices.
Grade date31 August 2026.
Stated limitationThe panel is four Polymarket markets, three of which resolve on 31 December 2026. As those approach resolution their prices compress toward certainty and the index carries less information. The panel is scheduled for expansion before that compression binds.

Not registered, and why

The Biological substrate. It currently rests on a single listed name carrying a fifteen percent weight. That is too thin to carry a threshold: one company's idiosyncratic news would trip a dimension presented as a read on a whole substrate. It is reported in the Reading, with the concentration named, and it is not registered until the Tier 2 expansion gives the substrate enough constituents to be read rather than merely sampled.

The FNC-1 level. Under the current rolling rebase, the level is not comparable week to week, so a threshold on it could not be honestly crossed. The spread is registered instead, because it survives the rebase.

Both exclusions are the register working as intended. An instrument that registers everything it can compute, rather than everything it can defend, fails on its first grade date.

Grading

On each grade date, FP1 publishes what the dimension read, whether the threshold was crossed, and what that implies, in the Reading and on this page. A dimension that did not move is reported as not having moved. Most weeks, that is the honest result, and a register that produces movement on a publication schedule is not measuring anything.

Method: NCB-004 documents how an FP1 signal is constructed, scored, and graded, including the conditions under which the method itself is revised or retired.