Fractals as the prerequisite infrastructure — formalized, with the trap named

A short note accompanying fractal_prerequisite.py. It takes the phrase "fractals as the prerequisite infrastructure" seriously enough to separate the false reading from the true one, and builds the true one as a runnable check.

---

The trap

Read literally, "fractals are the prerequisite infrastructure — the thing everything else requires" is false, and this toolkit already has the machinery to say why. dependency_graph.py established the foundational prerequisite as the well-founded root: the base case that depends on nothing (math as root, the human as authority). A fractal is, by definition, the object with no base case — its point-descent never bottoms out, which is exactly what its non-integer dimension measures (fractal_recursion.py). So making the non-terminating object the foundation of everything would smuggle in the very infinite regress fixed_point_governor exists to refuse. Fractals cannot be the universal foundational prerequisite.

The true statement underneath

There is a precise, defensible sense in which self-similarity is a prerequisite — a conditional one:

Self-similar (fractal) structure is the prerequisite for scale-invariance: it is what a system needs when the same structure or rule must hold across a range of scales.

That is this toolkit's own design. The same non-self-approval, fail-closed, human-grounded pattern is meant to apply at the agent, the mesh, and the federation — one rule, every scale. That is a scale-invariance requirement, and self-similarity is what meets it. The Sierpiński worked case in the tool makes this literal: three copies of the rule per level, exponent α = log3/log2 = 1.585 — a fractal dimension recovered from the governance shape itself.

The discipline: bounded, or refused

Self-similarity is legitimate infrastructure only when bounded — grounded by a real inner cutoff (a smallest scale it must reach) and a real outer cutoff (a largest). Every real fractal is bounded: a coastline is self-similar from meters to hundreds of km, not below grain size nor above the planet; a vascular tree branches from aorta to capillary and then stops. The governor returns:

verdict	meaning
GROUNDED_SCALE_INVARIANT	a single power law (constant log-log slope) holds across a bounded band between real cutoffs — a satisfied, grounded prerequisite. Reports α (a fractal dimension).
SCALE_BREAK	within the bounded band, the slope departs at some scale — self-similarity breaks there; the prerequisite is not met across the whole band. Reports the break scale.
UNGROUNDED_SCALE_DEMAND	the requirement has no inner or no outer cutoff — an unbounded "hold at every scale, forever" demand. Refused fail-closed: the infinite regress in a fractal mask.

require_bounded is the fail-closed gate: it admits only the grounded case and raises on the other two, so "it must hold at every scale" cannot slip an ungrounded demand past as if it were rigor.

The detector is principled: self-similarity is a constant log-log slope, so a break is the first scale where the local slope departs from the self-similar reference — not a global-fit residual, which the post-break tail would contaminate (that was the first draft's bug, and it flagged the wrong scale).

Honest scope

It checks a declared scaling relation over a finite sample: a clean power-law fit confirms self-similarity across the scales you gave it, not for all conceivable scales. It says nothing about whether a system ought to be scale-invariant — only, given that it must be, whether the structure supplies bounded self-similarity or demands an ungrounded one. Fractals come out as a conditional prerequisite (for scale-invariant systems), never the universal foundational root — that role stays with the well-founded base case. Stdlib-only, deterministic, self-testing.