PRH | Essay | 7.1 • The b-Closure of Mathematics

The b-Closure of Mathematics A Blur–First Program for Closing Theories Aleksandar Periˇsi´c November 2025 Abstract Blur means replacing sharp, brittle objects (hard cutoffs, point evaluations, oscillatory kernels, exact selectors) by positive, normalized approximate identities or by conservative closings that respect what we can and cannot access. I argue that blur is the missing organizing principle that lets us close mathematics in a pragmatic but rigorous sense: each theory can be completed up to an information budget, and statements are stabilized under blur before we attempt to unsmooth them. This yields a program—the b –closure of mathematics—that (i) stratifies theories by their information content and ergodicity, (ii) supplies a plan to close what is closable, and (iii) splits the unattainable into manageable descendants. The same logic unifies physics and math: physical laws start blurred; mathematics can adopt the same discipline without loss of rigor. Along the way I outline an “informational speed limit” analogy (a no–free–lunch bound that includes the prime/explicit–formula world) and a practical research/learning workflow where frustration drops because progress is measured at the blurred, stable level. Why blur, and what does it buy us? Blur is the act of replacing a sharp probe by a positive, normalized kernel or a conservative closing under which the statement of interest stabilizes. Convolution with a mollifier, Abel instead of Ces`aro summation, Poisson/Fej´er windows, or passing to an almost–everywhere selector are classical instances. In each case, a hard theorem reduces to a one–liner because positivity and dominated convergence do the heavy lifting; the real art lies in unsmoothing back to the sharp form when needed. Definition (Blur operators and b –stability).Fix a family ( Kσ ) σ>0 of positive, normalized kernels on a space X (e.g. Kσ≥ 0, RKσ = 1, Kσ→δ0 ). For f : X→R write ( Bσf )( x ) = Rf ( x−u ) Kσ ( u ) du . A property P( f ) is b –stable at x if there exists a scale regime σ↓ 0 such that P( Bσf ) holds uniformly for all sufficiently small σ ; we then write P( f ) holds in b –closure at x. In logic, we mirror this with conservative closings: we may work inside an enriched but conservative extension so long as the possibility to complete the argument survives all such closings; then a targeted existence claim holds already in the base theory (see §). The b–closure principle Principle (b–closure).To close a theory Tat an information budget I: C1. Calibrate a blur. Choose admissible kernels or closings such that key claims are b –stable. C2. Factor signal vs. envelope. Separate discrete/atomic features from the smooth envelope within the blurred picture. 1 C3. Quantify information. State explicitly what data length/precision is needed for a conclusion at the blurred level. C4. Unsmooth judiciously. Only remove the blur when the information budget actually supports it; otherwise, keep the blurred statement as the canonical theorem. Under b –closure, “existence” means existence with stable access: you can verify it in the blurred regime and you know what resources are required to pass to sharper forms. A taxonomy by growth and ergodicity Many theories stratify naturally by how information aggregates and by the presence of ergodic/mixing elements. A rough, deliberately coarse taxonomy: •Linear: responses add and blur distributes perfectly (Fourier/Abel tools trivialize). •Polynomial: interactions remain locally tame; blur still regularizes with bounded leakage. •Exponential: small parameter shifts amplify; blur must be bandwidth–aware. • Super–exponential/chaotic: strong sensitivity; only statistical/ergodic summaries are realistically b–stable. Each bucket occurs with/without ergodic elements. An ergodic element here simply means an intrinsic averaging mechanism that makes blurred probes representative of the whole. The practical import is informational: in some domains (e.g. primes) no amount of cleverness substitutes for data of size Nwith precision exceeding a threshold—not because we lack ideas, but because the object does not reveal certain statistics below that budget. In that case, b –closure records the best stable statements now and schedules the next targets when data budgets grow. Program: closing what can be closed Proposal (A b–closure workflow).For any target theory/problem: P1. Declare the probe. Fix admissible blur kernels (Gaussian, Paley–Wiener, Slepian/DPSS) and/or conservative closings. P2. State the blurred theorem. Prove the b –stable core that positivity and approximate identity guarantee; quantify leakage. P3. Separate channels. Project atomic (discrete) vs. smooth sectors after blur; treat each with its natural tools. P4. Budget the unsmoothing. Attach explicit data/precision thresholds (“how much is enough”). P5. If unattainable, split. Factor the object into sub–theories whose b –closures are separately attainable now. P6. Publish both the blurred theorem (stable, now) and the roadmap to its sharp form (future). 2 Physics and mathematics: the same discipline at different tolerances Every physical law is born blurred: Newton linearizes Einstein; Einstein blurs quantum fields at classical scales; Ohm and Amp`ere are first–order envelopes of richer electrodynamics; Tesla’s AC engineering exploits lossless limits as blurred ideals. The practical difference is that physics codifies its blur and tolerances; mathematics often pretends to have none. The b –closure stance is: keep exactness where it matters (logical consequence, definability), but adopt blur as the primary proving ground. Rigor is not diminished: statements are proved at the blurred level and only then selectively sharpened. A prime case study: explicit formula, bands, and an informational speed limit In the Mellin u = ln |v| picture, prime powers live on the prime band Ω = {klog p} and the explicit formula splits into an atomic comb on Ω, a smooth archimedean envelope, and endpoint residues. With band–limited or Gaussian blur, the two sectors decouple cleanly; detectors built from phase–scanned tests read off discrete vs. smooth mass with controllable leakage. Operationally: any off–line zero would inject an exponentially growing signature into the prime channel; under the Riemann hypothesis (RH), growth stays polynomial. This is an informational law: you cannot extract more from the primes than the calibrated blur and data budget allow. In this sense there’s a “speed limit”: no method can conjure signal that is not present at the blurred, budgeted scale. (The point is methodological; the details live in the blur/prime machinery.) Targeted choice without selectors: an axiom for existence under blur Blur has a logical avatar: sometimes we allow internal tools (like fragments of choice) during a conservative closing, and at the end all selectors vanish from the final statement. The Axiom of Blurred Choice (ABC) packages this: if after every conservative closing it remains possible to complete the argument using a fixed choice principle and obtain a choice function for a coded family, then already in the base theory the bare existence claim holds—no selector need be exhibited. This is exactly the practice of “use it inside, erase it at the end”, made formal and confined to the relevant existential. Research and learning without chronic frustration A blur–first workflow makes daily progress visible: •Ignore what you cannot know yet. Prove the blurred theorem you can prove now. •Collect what you do know. Calibrate detectors and budgets; record stable gains. •Read ahead. Explore the sharp picture to plan the unsmoothing steps. • Combine. Each pass tightens the blur or expands data until the sharp form is within reach. Frustration drops because the unit of progress is the stable blurred claim, not the all–or–nothing sharp finish. 3 What b–closure promises (and what it does not) This program does not magically solve everything. It does promise: 1. a uniform language for information budgets and detectors, 2. immediate, publishable blurred theorems (often dramatically simpler), 3. transparent roadmaps from blurred to sharp, 4. principled splits when an object is too stiff to close at once. Ambitious timelines (e.g., “4–5 years to a clarified landscape”) are best treated as programmatic: the b –closure makes the terrain legible and the targets concrete; the actual pace depends on budgets (data, computation, and proofs). Appendix A: Minimal blur calculus Definition (Admissible blur).A kernel Kσ is admissible if Kσ≥ 0, RKσ = 1, Kσ→δ0 , and either (i) Kσ is band–limited in Fourier with bandwidth Λ( σ ) ↑ ∞ , or (ii) Kσ is Schwartz with rapidly decaying tails. Remark (Blur ⇒ classical limits under mild regularity).If f has a classical limit L at x0 , then Bσf ( x0 ) →L . Conversely, if Bσf ( x0 ) →L and f has vanishing local oscillation at x0 (bounded variation/Lipschitz/monotone suffices), then the classical limit is L . This justifies treating b–stable laws as the canonical form until unsmoothing is funded. Appendix B: A disciplined split for any theory Given a target T: 1. declare admissible blur/closings and the observables, 2. prove the blurred theorem with explicit leakage, 3. project to discrete vs. smooth channels, 4. attach budgets to the unsmoothing, 5. if necessary, split T=T1⊕ · · · ⊕ Tmand iterate. Final word. The b –closure of mathematics is not a restriction but an emancipation: it legitimizes what we already do informally (smooth, linearize, pass to a.e., ignore null/meagre structure), makes it first–class, and turns it into a common program across number theory, dynamics, probability, analysis, and beyond. 4