Duality, Water, Flow, Beginning & End — four infrastructures, with honest scope

A combined note for duality_governor.py, water_infra.py, flow_conservation.py, and bounded_process.py. Each takes one of the recent words and builds the one precise, runnable thing underneath it — no more than is really there. "Math to reality" and the two fractal tools already existed and were not rebuilt.

---

Duality — duality_governor.py

The claim: duality is the shape every tool in this family already takes (proxy vs truth, words vs numbers, past vs future, recorded vs verified, black vs white raven). So it is not a new infrastructure — it is the precondition for any check: you can only test a claim against something the claim does not control.

What the tool adds is the failure nothing else catches — circular validation: a "ground truth" that is actually derived from the proxy it is meant to check. Verdicts: GROUNDED_DUALITY (two independent sources, check not a function of the claim — refutable), COLLAPSED_MONISM (one side only — uncheckable), CIRCULAR (check shares the claim's source), SUSPECTED_CIRCULAR (distinct sources on paper, but near-perfect correlation suggests derivation — a flag to investigate, since perfect agreement can be genuine). The demo catches an r = 1.0000 "audit recomputed from the KPI" as suspected-circular. The honest catch: near-perfect agreement is a reason to suspect, not relax.

Water — water_infra.py

Named honestly as the thinnest of the set: "water" is a metaphor, not a defined object like a fractal. Its one precise meaning: formless content takes the shape of its container — raw data has no governable meaning until a container (schema/type/unit) is declared, the same content reads as different values in different containers, and uncontained content cannot be governed. Verdicts: SHAPED, INCOMPATIBLE, FORMLESS. The demo shows the raw string "1/2" becoming 0.5 (ratio), a date (date container), or "1/2" (text) — the container, not the water, fixes the meaning. This is the parse-don't-validate discipline beneath words_vs_numbers, named and not inflated beyond it.

Flow — flow_conservation.py

Flow already lives as the telemetry/temporal stream; this isolates flow's other property — conservation. Flow through a chain of stages must balance: exit == entry − declared removal. The two silent failures are LEAK (less out — dropped/lost units) and FABRICATION (more out — duplicated/phantom units). This is mass-balance / Kirchhoff applied to data and process pipelines — literally how you audit an ETL chain for dropped or duplicated records. The demo catches a 60-unit leak at dedup and a 240-unit fabrication at a join, and also flags a disconnected chain (ENTRY_MISMATCH). The point of forcing a declared accounted-removal is that it stops a drop from hiding inside a vague "some filtering happens here."

Beginning & End — bounded_process.py

Beginning and end are duals, so they are one tool. A process is WELL_BOUNDED only if it has an independent beginning (a base case that does not presuppose its own output) and a real end (it halts within a declared bound). These are the inner and outer cutoffs that fractal_prerequisite.py requires to turn a regress into a grounded structure, and the same well-foundedness fixed_point_governor enforces. NO_BEGINNING = bootstrapping paradox (unbounded below); NO_END = non-terminating within the bound (unbounded above, the regress). Honest scope: halting is undecidable in general, so the tool does not prove termination — it runs to a declared bound (the honest stand-in for a termination proof) and refuses anything that outruns it.

---

The through-line

All four are deterministic, self-testing, stdlib-only, non-safety-critical. Together they show the recent words are the motifs the existing family is made of, now each pinned to one runnable check: duality is the precondition for checking, water is the container that gives content shape, flow is the conserved quantity that must balance, and beginning/end are the two cutoffs that bound a process into legitimacy — the same two cutoffs that ground a fractal.