Trace Density
ρα
How tightly a recursion has packed its history
Trace density is a single reading taken over the α-trace: not how much history an attractor has accumulated, but how densely that history is encoded — how packed, interlocked, and feature-rich the record is. Where the α-trace is the structured object, trace density is a scalar measure over it. It rises with each completed compression event, it tracks closely with how hard the structure is to reconfigure, and — like all Tier-3 descriptors — it describes; it does not enforce.
The plain idea
An attractor's α-trace is the record of what its recursion has selected and kept. But two attractors with archives of the same size can be very different: one may be loose and redundant, the other tightly interlocked, every part of it load-bearing. Trace density is the number that captures that difference. It asks not how large is the record but how much history is packed into it.
The right image is compression in the ordinary sense. A file can be large and mostly air, or small and dense with meaning. Trace density reads the density, not the size. A vast genome with long stretches of nothing carries a low trace density; a compact one where nearly every base is conserved and doing work carries a high one. The measure is about the packing, never the volume of the medium that holds it.
“Trace” here is narrower than the everyday word
In ordinary use, a trace is almost anything left behind — a deer track in mud, an ant trail, droppings in the yard, a worn footpath, a scar, a fingerprint. That is a vast domain: nearly anything that was leaves some mark that it was. The framework does not mean this. An α-trace is not “any mark left behind”; it is specifically the residue of a completed compression event inside a sovereign recursion — the record of what a running, self-sustaining loop selected and kept. That qualifier does the work, and it rules most ordinary “traces” out.
A quick test settles any candidate. It is an α-trace only if all three hold:
1. there is a sovereign recursion — a loop that holds itself together;
2. that recursion compressed its own configuration — kept what worked, shed what did not;
3. the thing left behind is the residue of that compression.
All three → α-trace, and it can carry a density. Any one fails → it is a mark, a byproduct, or an imprint — real, sometimes enormous, but not the kernel’s object, and it has no trace density.
Is an α-trace
A genome — the cell is a sovereign attractor, and its genome is the packed record of what a long lineage’s recursion selected and kept. The trained weights of a network, the skill in a trained hand, the conventions a living institution runs on: each is a record a live loop laid down about its own selection history.
Is not
A deer track in mud — the mud runs no recursion and selected nothing; it just deformed. Droppings are an output, not a loop’s compressed history of itself. A fingerprint, a worn path, an eroded scar: marks the world happens to hold, not compression residue. Real evidence of a past — but not α-traces, and nothing to take a density over.
So the everyday sense is the intuition pump — a trace really is “a record of what was” — and the framework borrows that intuition, then clamps it to one kind of record: the kind a self-sustaining loop leaves about what it selected. The gargantuan domain gets you in the door; the narrow definition is what makes trace density a usable measure instead of a synonym for “the past.”
Where it comes from: repeated successful selection
Trace density is built the same way the α-trace is built — one completed compression event at a time. Each time the recursion runs, keeps what worked, and sheds what did not, the residue of that selection settles into the archive, and the archive gets a little denser.
It increases with repeated successful selection. Every completed compression event contributes to ρα. A structure that has been through many rounds of selection, keeping the survivors each time, ends up with a densely encoded record precisely because so much has been tested and retained.
It starts at zero. At ignition the α-trace is empty, so trace density is at its floor: ρα(X, tignition) = 0. Nothing has been compressed yet; there is no history to pack. Density is something the recursion earns by running.
Density, not volume: the distinction to hold onto
This is the single easiest place to go wrong, and the framework names it directly. Trace density is a measure of how densely history is compressed within whatever substrate exists — not a measure of how big that substrate is.
High trace density
The compression is tightly packed. Much history is encoded in the record, its features interlocked and load-bearing. A small substrate can carry a high density if the compression within it is rich: a compact genome where nearly every motif is conserved, a sparse-but-decisive set of trained weights, a lean body of convention that decides a great deal.
Low trace density
The compression is thin. A large substrate can carry a low density if the α-trace within it is sparse: a big genome mostly non-coding, an over-parameterised network with little of it doing work, a voluminous record that settles almost nothing. Size is not the reading; packing is.
ρα; a small substrate with dense compression has high ρα. Reading the size of the medium as if it were the density of the history is a category error.
When it exists, and when it stops
Trace density lives exactly as long as the live α-trace does — it is defined for a sovereign attractor during its bootstrapping interval, and no longer.
Zero at ignition
An attractor at the moment of ignition has an empty α-trace (α(X, t) = ∅α), so its trace density is at its minimum, zero. This is a real state, not the absence of the measure: the archive exists and its first entry is about to be written.
Gone after loss
Trace density terminates when the live α-trace terminates — that is, when the attractor undergoes β-loss. The inert α-trace residue that remains — the fossil, the ruins, the trained weights of a deleted model — has no ρα, because no live α-trace exists to take a density over. The record still lies in the world; the density reading does not survive the stopping of the loop.
Reading it in the real
Because trace density is substrate-independent, its measurement form is always local to the medium, but the reading is the same everywhere: how densely is the surviving history packed? The framework names several: sparsity measures of trained weights (how much of the network is actually load-bearing), conserved-motif density in genomes (how much of the sequence is held constant across a lineage), conventionality density in institutional records (how much of the behaviour is fixed by settled convention), and basin-depth distributions (how deeply the live α-trace has carved the valleys the recursion runs in; the archive co-varies with depth, but the live filter is what shapes it). In each case the metric reads packing, and each round of successful selection tends to raise it.
What trace density is not
Its relations: the α-trace and logic mass
To the α-trace. Trace density is the density measure over the α-trace: the α-trace is the structured object of accumulated history; ρα is the scalar that reads how packed it is. No live α-trace, no trace density.
To logic mass. Trace density is strongly correlated with logic mass (mL), the structure's resistance to being reconfigured — because a dense α-trace produces strong reconfiguration resistance. The link is not an identity: logic mass also draws on interlock density and accumulated constraint that do not come through the α-trace. But as a rule of thumb, a high ρα usually means a high mL — the more densely a structure has packed its history, the harder it is to talk it into becoming something else.
The α-trace is the history a recursion keeps. Trace density is how tightly it is packed — and the tighter the packing, the more the past weighs on what the structure can still become. It reads the density of what survived, never the size of the thing it survived in.
See also: the object it measures, the α-trace α; its active companion recurcline Rc; the full set of bootstrapping-interval metrics, including logic mass mL, with which trace density is strongly correlated; how history accretes across the interval; and ignition and rupture, which open and close the interval over which density is defined.
Trace density (kernel construct TRACE-DENS, symbol ρα) is a Tier-3 persistence descriptor, strictly downstream of the primitives, admissibility theorems, and sovereignty conditions. This page presents its canonical treatment in plain language: a density measure over the α-trace, not the α-trace itself; increasing with repeated successful selection; zero at ignition (empty α-trace) and terminating with live α-trace at β-loss; density of compression, never volume of substrate (the named substrate-volume-confusion failure); and strongly — not identically — correlated with logic mass. The non-enforcement discipline is preserved verbatim: “Trace density describes history retention. It does not enforce it.” It introduces no constructs and modifies no canon; trace density is treated throughout as an accounting density measure, never a causal driver. The trained-weights, genome, and institutional cases are downstream illustrations.