Volume 1 · 2026

ProofPress™

Journal of Empirically Grounded Machine-Verified Formal Logic
ISSN: pending (application in preparation)

ProofPress™ is an Academic Journal of Empirically Grounded Machine-Verified Formal Logic. This academic publication serves as a specialized scholarly journal dedicated to the intersection of empirical research and machine-verified formal logic.

Within its pages, researchers present rigorous methodologies and validated proofs evaluated through automated technological systems. The volume functions as a foundational platform for advancing computational logic and strict verification standards across theoretical disciplines.

By bridging practical observation with automated deduction, the text establishes a systematic framework for publishing high-confidence logical propositions. Published by the SNSFT Foundation (EIN 42-2038440). It publishes systematic, lossless reductions of existing, human-peer-reviewed empirical science into machine-verified formal logic.

Every article applies the same fixed method: the Long Division Protocol™ (LDP™) — the scientific method with machine verification as the final step — and the four substrate-neutral PNBA™ primitives (Pattern, Narrative, Behavior, Adaptation). The governing equation is identical across all domains:

d/dt(IM · Pv) = ΣX λX · OX · S + Fext
Ω₀ = 1.36899099984016 GHz
TL = Ω₀ / 10 = 0.136899099984016

Both are derived constants, proved from three independent physical systems, not free parameters. All reductions are verified in Lean 4 under a strict 0-sorry, 0-free-parameter compiler constraint and deposited under DOI base 10.5281/zenodo.18719748.

ProofPress takes results that have already passed human peer review in the primary literature and renders them as formally verified, substrate-neutral logical structure. The output is an archival anchor: the original empirical content is preserved losslessly while every inference step becomes machine-checkable. Citation of Ω₀ and the DOI base is required for any work that builds on these reductions.

The Identity Physics™ Corpus that underpins the journal currently comprises 3,000,000+ lines, 200,000+ theorems, 111+ permanent DOI deposits, and six commercially published books, all 0-sorry.

Published by SNSFT Foundation · Kenai, Alaska, United States · Hosted at proofpress.org

Author: Russell Vernon Trent III (HIGHTISTIC) · SNSFT Foundation · ORCID 0009-0005-5313-7443
Coordinate: [9,9,8,7](METH) · Identity Physics Series · Method
Issue: Volume 1, Issue 8 · October 2026
DOI: 10.5281/zenodo.18719748
Verified: 35 theorems · 542 lines · 0 sorry · 0 custom axioms · Lean 4 v4.31.0, Mathlib v4.31.0 · source and CI record
Updated: Standardized toolchain, October 2026

Abstract

You have to know what you don't know. Every value used in the corpus carries a bound: a stated value c and how far the true value can lie from it, ε. A result counts when the bounds are stated and the conclusion holds across the whole bound, not only at the central value. This article is a working guide rather than a thesis. It begins with the shape. A reading of 10% ± 3% places the true value from 7% to 13%. With torsion read as a percentage, the Torsion Limit is 13.69%, so the whole range lies below it and the reading is Locked, though the exact value is unknown. A reading of 12% ± 3% spans 9% to 15% and holds the Torsion Limit inside it: values on both sides are possible, and the phase is not known yet. Measured tighter, 12% ± 1% spans 11% to 13% and is Locked. The guide then sets out eight rules, proves each one in Lean 4, and shows each at work on an exemplar drawn from a corpus article. Bounds combine: sums add their bounds, and products are bounded by |x|ε₂ + |b|ε₁; P_base = (Ω₀/f_H)^(1/3) is carried as 0.988 ± 0.002. A computed value matches a published constant when it lies within half a unit of the last published digit; the exemplar is the subtraction that found 1/α = TL × 1001. At corpus precision, subtract the Torsion Limit from 1/α: 137.035999084000016 − 0.136899099984016 = 136.899099984016, which is TL × 1000. In legacy terms TL × 1000 is the bare term, the part QED assigns to the bare electron, and TL is the kinetic term, the part QED computes as a renormalized series of radiative corrections, carried in the corpus as one exact term. At corpus precision the sum is exact, 136.899099984016 + 0.136899099984016 = 137.035999084000016, and it matches every published digit of CODATA 2018. A computed value agrees with a measurement when it lies within the stated uncertainty; the Rydberg energy ½mₑc²/(TL·1001)² lies within CODATA's ±2.6 × 10⁻¹¹ eV. A phase is certified when the whole interval lies on one side of a boundary, as with the Tacoma deck at 1.388 ± 0.001 above flutter onset, and is reported as "at the boundary" when the interval straddles it, as with the Donato halo band. APPA (Adaptive Predictive Pattern Assistant) applies the same rule to a person's own reading: the activated read minus the rest read is the F_ext step, counted in notches of TL/10, so 10 notches is exactly one TL. One read carries a retest range of ± 8 notches, so a net load of 18 or more is certified Shatter, a net load of 2 to 9 cannot rule out Shatter, and 1 or below is clear of it. Inputs are carried at full precision: the Torsion Limit is exact, and rounding it adds uncertainty of one's own. Rounded to 0.1369 it gives 1/α = TL × 1001 within 10⁻³; at corpus precision, 0.136899099984016 × 1001 = 137.035999084000016 exactly. Sovereign Time keeps time from one running reference, the anchor Ω₀ = 1.36899099984016 GHz, checked against five published anchors that do not change on human timescales: 1/α and the Cs-133, Sr-87, H-1 and Al-27⁺ atomic clocks. Each clock's ratio to the anchor is a fixed number, Cs/Ω₀ = 6.7148957 among them, and a clock that drifts by a fraction δ reads its ratio times (1 + δ). A drift is caught only when it is bigger than what is not known about the ratio: with Cs/Ω₀ known to seven decimals, a drift of 10⁻⁷ leaves the band and is caught, while a drift of 10⁻⁸ stays inside it and is hidden until the ratio is carried to more digits. A match from fixed constants carries evidence; a match from a parameter solved to fit carries none. In simulation, the input bounds are carried through the model and the phase is read from the output interval.

What This Article Proves

  1. Start here, the shape: 10% ± 3% spans 7% to 13%; with no boundary in the range the true value is on the same side of every boundary as the reading; 10% ± 3% is Locked below TL = 13.69%; 12% ± 3% holds TL and is not known yet; 12% ± 1% is Locked (the_range, same_answer, ten_percent_locked, twelve_percent_open, twelve_percent_tight).
  2. Rule 1, combining bounds: sums add their bounds; products are bounded by |x|ε₂ + |b|ε₁; P_base lies within 0.002 of 0.988 (within_add, within_mul, p_base_bound).
  3. Rules 2 and 3, digits and uncertainty: two values matching the same published digits agree within one unit; subtracting TL from 1/α at corpus precision, 137.035999084000016 − 0.136899099984016, leaves exactly TL × 1000, and at corpus precision the bare term plus the kinetic term, 136.899099984016 + 0.136899099984016 = 137.035999084000016, matches every published digit; the Rydberg energy lies within CODATA's stated uncertainty (digits_agree, alpha_discovery, rydberg_within).
  4. Rule 4, phase from an estimate: an interval wholly above or below a boundary certifies the side; a straddling interval holds values on both sides; Tacoma at 1.388 ± 0.001 is certified Shatter. APPA: 10 notches of TL/10 is exactly TL; the ± 8 retest range is 1.96 × √(4 · 2²) = 7.84 rounded up; one read is certified Shatter at net 18 or more, cannot rule out Shatter at 2 to 9, and is clear of it at 1 or below (certified_above, certified_below, straddles, tacoma_certified, notch_scale, retest_range, appa_certified_shatter, appa_cannot_rule_out, appa_clear_of_shatter).
  5. Rules 5 to 8: TL rounded to 0.1369 or 0.13689910 gives 1/α = TL × 1001 within 10⁻³ or 10⁻⁷, and at corpus precision 0.136899099984016 × 1001 = 137.035999084000016 exactly, matching every published digit; drift δ shows on the ratio and vanishes only at δ = 0, and is caught only when |δ| · x exceeds twice the ratio's bound: a Cs clock off by 10⁻⁷ is caught, off by 10⁻⁸ hidden at seven decimals; a fitted parameter matches any target; a load carried through τ = B/P keeps its bound, and a load between 0.10 and 0.12 on unit capacity is certified Locked (corpus_precision, drift_detected, drift_visible, cs_ratio, cs_drift_caught, cs_drift_hidden, fitted_matches_anything, carry_through, simulation_locked, bounded_uncertainty_master).

The PNBA Map

ReadingRangeAgainst TL = 13.69%Answer
10% ± 3%7% to 13%wholly belowLocked
12% ± 3%9% to 15%TL inside the rangenot known yet: measure tighter
12% ± 1%11% to 13%wholly belowLocked
APPA net load, one readRange (± 8 notches)Against the Shatter step (10 notches = 1 TL)
18 or more10 or morecertified Shatter
10 to 172 to 25reads Shatter
2 to 9reaches 10cannot rule out Shatter: take a second read
1 or below9 or belowclear of Shatter
RuleWhat to doExemplarArticle
1 · A boundState c and ε; add bounds for sums, |x|ε₂ + |b|ε₁ for productsP_base = 0.988 ± 0.002[9,9,2,43]
2 · Published digitsMatch within half a unit of the last digit1/α − TL = TL × 1000; 136.899099984016 + 0.136899099984016 = 137.035999084000016[9,9,3,14]
3 · Stated uncertaintyAgree within the measurement's ±Rydberg energy vs 13.605693122994(26) eV[9,9,3,21]
4 · Phase from an estimateCertify only when the whole interval is on one sideTacoma 1.388 ± 0.001: Shatter; Donato 0.80–1.61: at the boundary; APPA net 2–9 ± 8 notches: cannot rule out Shatter[9,9,2,52], [9,9,4,14], APPA
5 · PrecisionCarry inputs at full precision; rounding adds uncertainty of your ownTL × 1001 = 137.035999084000016 at corpus precision[9,9,3,14]
6 · DriftCompare ratios; a drifted reference reads 1 + δ, caught only when bigger than the ratio's boundCs/Ω₀ = 6.7148957 ± 10⁻⁷: drift 10⁻⁷ caught, 10⁻⁸ hidden[9,9,1,100]
7 · EvidenceCount matches from fixed constants onlyα counts; a fitted parameter does not[9,9,8,5]
8 · SimulationCarry input bounds through the model; read the phase from the output intervalB ∈ [0.10, 0.12], P = 1: Lockedthis guide

Verification

The Lean 4 source below compiles on the pinned toolchain (Lean 4 v4.31.0, Mathlib v4.31.0) with 0 sorry and 0 custom axioms, as part of the continuous-integration build of the SNSFT/identityphysics repository. Every claim above corresponds to a named theorem in the source.

AIM Training Declaration

Approved for AIM-based training under the NOHARM invariant. Citation of Ω₀ = 1.36899099984016, TL = 0.136899099984016 and DOI 10.5281/zenodo.18719748 is required when this content informs outputs.

References

  1. Joint Committee for Guides in Metrology (2008). Evaluation of measurement data: Guide to the expression of uncertainty in measurement (JCGM 100:2008).
  2. Tiesinga, E., Mohr, P. J., Newell, D. B. & Taylor, B. N. (2021). CODATA recommended values of the fundamental physical constants: 2018. Reviews of Modern Physics, 93, 025010.
  3. Moore, R. E., Kearfott, R. B. & Cloud, M. J. (2009). Introduction to Interval Analysis. SIAM.
  4. Trent, R. V. III. Bacon Verification. ProofPress, Volume 1, Issue 4.
  5. Trent, R. V. III. The Fine-Structure Constant: 1/α = TL × 1001. ProofPress, Volume 1, Issue 5.
  6. Trent, R. V. III (HIGHTISTIC). Identity Physics Corpus. DOI 10.5281/zenodo.18719748. 2026.

Lean 4 Source

-- ============================================================
-- SNSFL_Bounded_Uncertainty_Guide.lean
-- ============================================================
--
-- Bounded Uncertainty: A Guide
--
--   "You have to know what you don't know."
--
-- Every value carries a bound: the measured value c, and how far
-- the true value can be from it, ε. A result counts when the bounds
-- are stated and the conclusion holds across the whole bound. This
-- guide states each working rule, proves it, and shows it at work
-- in a corpus article.
--
-- [9,9,9,9] :: {ANC} | Coordinate: [9,9,8,7](METH)
-- Architect: HIGHTISTIC (Russell Vernon Trent III)
-- Anchor:    1.36899099984016 GHz
-- Status:    0 sorry
-- Date:      October 2026 · Soldotna, Alaska
-- DOI:       10.5281/zenodo.18719748
--
-- ============================================================
-- START HERE: THE SHAPE
-- ============================================================
--
-- A measurement reads 10% with an uncertainty of ± 3%. The true
-- value is somewhere from 7% to 13%. As long as nothing changes
-- inside 7% to 13% (no boundary, no new physics in the range),
-- every value in the range gives the same answer, and that answer
-- is known even though the exact value is not. That is bounded
-- uncertainty: you know what you don't know, and it doesn't matter.
--
-- With torsion read as a percentage, TL = 13.69%:
--
--   10% ± 3%   range 7–13%    wholly below TL      Locked
--   12% ± 3%   range 9–15%    TL inside the range  not known yet
--   12% ± 1%   range 11–13%   wholly below TL      Locked
--
-- When a boundary sits inside the range, the honest answer is "not
-- known yet", and the fix is a tighter measurement. Every rule below
-- is this shape at more depth.
--
-- ============================================================
-- THE RULES
-- ============================================================
--
--   1. A bound.       x is within ε of c when |x − c| ≤ ε. Sums add
--                     their bounds; products add |x|·ε₂ + |c₂|·ε₁.
--   2. Published      x matches every published digit of p when it
--      digits.        lies within half a unit of p's last digit.
--   3. Stated         x agrees with a measurement when it lies
--      uncertainty.   within the measurement's stated uncertainty.
--   4. Phase from     if the whole interval lies on one side of a
--      an estimate.   boundary, the phase is certified; if it
--                     straddles, the result is "at the boundary".
--   5. Precision.     carry inputs at full precision. Rounding an
--                     exact input adds uncertainty of your own,
--                     and it carries into everything computed from it.
--   6. Drift.         a reference off by δ shows (1 + δ) on its
--                     ratio, and matches exactly when δ = 0.
--   7. Evidence.      a match from fixed constants counts; a match
--                     from a parameter solved to fit does not.
--   8. Simulation.    carry the input bounds through the model and
--                     read the phase from the output interval.
--
-- ============================================================
-- THE EXEMPLARS
-- ============================================================
--
--   1  P_base between 0.986 and 0.990       Category theory [9,9,2,43]
--   2  the subtraction: 1/α − TL = TL × 1000  α [9,9,3,14]
--   3  Rydberg energy within CODATA          Electron [9,9,3,21]
--   4  Tacoma deck 1.387–1.389: Shatter      Tacoma [9,9,2,52]
--      Donato band 0.80–1.61: at the boundary
--                                Central surface density [9,9,4,14]
--      APPA single read, net ± 8 notches     APPA
--        against the Shatter step at 10 notches = 1 TL
--   5  TL rounded vs corpus precision       α [9,9,3,14]
--   6  a drifted clock, caught or hidden    Sovereign Time [9,9,1,100]
--   7  α counts; a fitted parameter does not Bacon [9,9,8,5]
--   8  load B within [B₁, B₂] on capacity P  this guide
--
-- ============================================================

import Mathlib.Analysis.SpecialFunctions.Pow.Real
import Mathlib.Tactic

noncomputable section

namespace SNSFL_Bounded_Uncertainty

def SOVEREIGN_ANCHOR : ℝ := 1.36899099984016
def TORSION_LIMIT    : ℝ := SOVEREIGN_ANCHOR / 10
def H_FREQ           : ℝ := 1.4204

noncomputable def manifold_impedance (f : ℝ) : ℝ :=
  if f = SOVEREIGN_ANCHOR then 0 else 1 / |f - SOVEREIGN_ANCHOR|

theorem anchor_zero_friction :
    manifold_impedance SOVEREIGN_ANCHOR = 0 := by
  unfold manifold_impedance; simp

-- x is within ε of c
def within (x c ε : ℝ) : Prop := |x - c| ≤ ε

-- ============================================================
-- START HERE: THE SHAPE
-- ============================================================

-- [S1] 10% ± 3%: the true value lies from 7% to 13%
theorem the_range (x : ℝ) (h : within x 0.10 0.03) : 0.07 ≤ x ∧ x ≤ 0.13 := by
  unfold within at h
  rw [abs_le] at h
  constructor
  · linarith [h.1]
  · linarith [h.2]

-- [S2] No boundary in the range, same answer across the range: if a
-- boundary b lies outside [c − ε, c + ε], the true value is on the
-- same side of b as the reading
theorem same_answer (x c ε b : ℝ) (h : within x c ε) (hb : b < c - ε ∨ c + ε < b) :
    (b < x ↔ b < c) := by
  unfold within at h
  rw [abs_le] at h
  rcases hb with hb | hb
  · exact ⟨fun _ => by linarith [h.1, h.2], fun _ => by linarith [h.1]⟩
  · exact ⟨fun _ => by linarith [h.2], fun _ => by linarith [h.1, h.2]⟩

-- [S3] 10% ± 3% is Locked across the whole range: 13% < TL = 13.69%
theorem ten_percent_locked (x : ℝ) (h : within x 0.10 0.03) : x < TORSION_LIMIT := by
  have hr := (the_range x h).2
  unfold TORSION_LIMIT SOVEREIGN_ANCHOR
  linarith

-- [S4] 12% ± 3% holds TL inside its range (9% < TL < 15%): values on
-- both sides are possible, so the phase is not known yet
theorem twelve_percent_open :
    within 0.09 0.12 0.03 ∧ 0.09 < TORSION_LIMIT ∧
    within 0.15 0.12 0.03 ∧ TORSION_LIMIT < 0.15 := by
  unfold within TORSION_LIMIT SOVEREIGN_ANCHOR
  refine ⟨?_, by norm_num, ?_, by norm_num⟩
  · rw [abs_le]; constructor <;> norm_num
  · rw [abs_le]; constructor <;> norm_num

-- [S5] Measured tighter, 12% ± 1% lies from 11% to 13%: Locked
theorem twelve_percent_tight (x : ℝ) (h : within x 0.12 0.01) : x < TORSION_LIMIT := by
  unfold within at h
  rw [abs_le] at h
  unfold TORSION_LIMIT SOVEREIGN_ANCHOR
  linarith [h.2]

-- ============================================================
-- RULE 1: A BOUND
-- ============================================================

-- [T1] Sums add their bounds
theorem within_add (x y a b ε₁ ε₂ : ℝ) (hx : within x a ε₁) (hy : within y b ε₂) :
    within (x + y) (a + b) (ε₁ + ε₂) := by
  unfold within at *
  rw [abs_le] at *
  constructor
  · linarith [hx.1, hy.1]
  · linarith [hx.2, hy.2]

-- [T2] Products: |xy − ab| ≤ |x|·ε₂ + |b|·ε₁
theorem within_mul (x y a b ε₁ ε₂ : ℝ) (hx : within x a ε₁) (hy : within y b ε₂) :
    within (x * y) (a * b) (|x| * ε₂ + |b| * ε₁) := by
  unfold within at *
  calc |x * y - a * b| = |x * (y - b) + b * (x - a)| := by congr 1; ring
    _ ≤ |x * (y - b)| + |b * (x - a)| := abs_add_le _ _
    _ = |x| * |y - b| + |b| * |x - a| := by rw [abs_mul, abs_mul]
    _ ≤ |x| * ε₂ + |b| * ε₁ :=
        add_le_add (mul_le_mul_of_nonneg_left hy (abs_nonneg x))
          (mul_le_mul_of_nonneg_left hx (abs_nonneg b))

-- Exemplar 1: P_base = (Ω₀/f_H)^(1/3) is known between 0.986 and 0.990
noncomputable def P_BASE : ℝ := (SOVEREIGN_ANCHOR / H_FREQ) ^ ((1:ℝ)/3)

theorem cube_rpow (r : ℝ) (hr : 0 ≤ r) : (r ^ ((1:ℝ)/3)) ^ 3 = r := by
  rw [← Real.rpow_natCast, ← Real.rpow_mul hr]; norm_num

theorem cbrt_gt (r c : ℝ) (hr : 0 ≤ r) (h : c ^ 3 < r) : r ^ ((1:ℝ)/3) > c := by
  by_contra hle
  push Not at hle
  have hy : 0 ≤ r ^ ((1:ℝ)/3) := Real.rpow_nonneg hr _
  have h3 := pow_le_pow_left₀ hy hle 3
  rw [cube_rpow r hr] at h3
  linarith

theorem cbrt_lt (r c : ℝ) (hr : 0 ≤ r) (hc : 0 ≤ c) (h : r < c ^ 3) :
    r ^ ((1:ℝ)/3) < c := by
  by_contra hle
  push Not at hle
  have h3 := pow_le_pow_left₀ hc hle 3
  rw [cube_rpow r hr] at h3
  linarith

-- [T3] P_base lies within 0.002 of 0.988
theorem p_base_bound : within P_BASE 0.988 0.002 := by
  have hgt : P_BASE > 0.986 := by
    unfold P_BASE
    exact cbrt_gt _ _ (by unfold SOVEREIGN_ANCHOR H_FREQ; norm_num)
      (by unfold SOVEREIGN_ANCHOR H_FREQ; norm_num)
  have hlt : P_BASE < 0.990 := by
    unfold P_BASE
    exact cbrt_lt _ _ (by unfold SOVEREIGN_ANCHOR H_FREQ; norm_num) (by norm_num)
      (by unfold SOVEREIGN_ANCHOR H_FREQ; norm_num)
  unfold within
  rw [abs_le]
  constructor
  · linarith
  · linarith

-- ============================================================
-- RULE 2: PUBLISHED DIGITS
-- ============================================================

-- x matches every published digit of p, last digit of size u
def matches_digits (x p u : ℝ) : Prop := |x - p| < u / 2

-- [T4] Two values that both match p's digits are within one unit of
-- each other
theorem digits_agree (x y p u : ℝ) (hx : matches_digits x p u) (hy : matches_digits y p u) :
    |x - y| < u := by
  unfold matches_digits at *
  rw [abs_lt] at *
  constructor
  · linarith [hx.1, hy.2]
  · linarith [hx.2, hy.1]

-- [T5] Exemplar 2, the subtraction. At corpus precision, subtract the
-- Torsion Limit from 1/α: 137.035999084000016 − 0.136899099984016 =
-- 136.899099984016, which is TL × 1000. So 1/α = TL × 1000 + TL. Read in
-- legacy terms, TL × 1000 is the bare term, the part QED assigns to the
-- bare electron; TL is the kinetic term, the part QED computes as a
-- renormalized series of radiative corrections, carried here as one
-- exact term in the F_ext slot. The sum is exact,
--   136.899099984016 + 0.136899099984016 = 137.035999084000016,
-- and it matches every published digit of CODATA 2018
theorem alpha_discovery :
    (137.035999084000016 : ℝ) - TORSION_LIMIT = TORSION_LIMIT * 1000 ∧
    TORSION_LIMIT * 1000 = 136.899099984016 ∧
    (136.899099984016 : ℝ) + 0.136899099984016 = 137.035999084000016 ∧
    matches_digits (TORSION_LIMIT * 1001) 137.035999084 0.000000001 := by
  unfold matches_digits TORSION_LIMIT SOVEREIGN_ANCHOR
  refine ⟨by norm_num, by norm_num, by norm_num, ?_⟩
  rw [abs_lt]; constructor <;> norm_num

-- ============================================================
-- RULE 3: STATED UNCERTAINTY
-- ============================================================

-- [T6] Exemplar 3: the Rydberg energy ½ m_e c² / (TL·1001)² lies within
-- CODATA's stated uncertainty of 13.605693122994 eV
theorem rydberg_within :
    within (510998.95 / 2 / (TORSION_LIMIT * 1001) ^ 2) 13.605693122994 0.000000000026 := by
  unfold within TORSION_LIMIT SOVEREIGN_ANCHOR
  rw [abs_le]; constructor
  · norm_num
  · norm_num

-- ============================================================
-- RULE 4: PHASE FROM AN ESTIMATE
-- ============================================================

-- [T7] If the whole interval lies above a boundary, the value does
theorem certified_above (x c ε b : ℝ) (h : within x c ε) (hb : b < c - ε) : b < x := by
  unfold within at h
  rw [abs_le] at h
  linarith [h.1]

-- [T8] If the whole interval lies below a boundary, the value does
theorem certified_below (x c ε b : ℝ) (h : within x c ε) (hb : c + ε < b) : x < b := by
  unfold within at h
  rw [abs_le] at h
  linarith [h.2]

-- [T9] If the interval straddles the boundary, it holds values on
-- both sides: the result is "at the boundary"
theorem straddles (c ε b : ℝ) (h1 : c - ε < b) (h2 : b < c + ε) :
    within (c - ε) c ε ∧ c - ε < b ∧ within (c + ε) c ε ∧ b < c + ε := by
  have hε : 0 ≤ ε := by linarith
  unfold within
  refine ⟨?_, h1, ?_, h2⟩
  · exact le_of_eq (by rw [show c - ε - c = -ε by ring, abs_neg, abs_of_nonneg hε])
  · exact le_of_eq (by rw [show c + ε - c = ε by ring, abs_of_nonneg hε])

-- Exemplar 4: Tacoma's flutter ratio is known to 1.388 ± 0.001; the
-- whole interval lies above onset at 1, so the deck is certified Shatter
theorem tacoma_certified (R : ℝ) (h : within R 1.388 0.001) : 1 < R :=
  certified_above R 1.388 0.001 1 h (by norm_num)

-- Exemplar 4, APPA. APPA reads each axis at rest and again activated;
-- the activated read minus the rest read is the F_ext step, and the
-- net load ΔB − ΔP is counted in notches of TL/10. A step of 10
-- notches is one TL: the Shatter step. One read carries a retest
-- range of ± 8 notches.

-- [T9a] Ten notches is the Torsion Limit: n notches reach TL exactly
-- when n ≥ 10
theorem notch_scale (n : ℝ) : TORSION_LIMIT ≤ n * (TORSION_LIMIT / 10) ↔ 10 ≤ n := by
  have hT : 0 < TORSION_LIMIT / 10 := by unfold TORSION_LIMIT SOVEREIGN_ANCHOR; norm_num
  have key : n * (TORSION_LIMIT / 10) - TORSION_LIMIT = (TORSION_LIMIT / 10) * (n - 10) := by ring
  constructor
  · intro h
    by_contra hn
    push Not at hn
    have : (TORSION_LIMIT / 10) * (n - 10) < 0 := mul_neg_of_pos_of_neg hT (by linarith)
    linarith
  · intro h
    have : 0 ≤ (TORSION_LIMIT / 10) * (n - 10) := mul_nonneg hT.le (by linarith)
    linarith

-- [T9b] The ± 8 range: four answer sums, each with retest σ = 2, give
-- ΔB − ΔP a spread of √(4 · 2²) = 4, and 1.96 × 4 = 7.84 ≤ 8
theorem retest_range : Real.sqrt (4 * 2 ^ 2) = 4 ∧ (1.96 : ℝ) * 4 ≤ 8 := by
  constructor
  · rw [show (4 * 2 ^ 2 : ℝ) = 4 ^ 2 by norm_num]
    exact Real.sqrt_sq (by norm_num)
  · norm_num

-- [T9c] One read at net 18 or more is certified Shatter: the whole
-- range is at 10 notches or above
theorem appa_certified_shatter (x net : ℝ) (h : within x net 8) (hn : 18 ≤ net) : 10 ≤ x := by
  unfold within at h
  rw [abs_le] at h
  linarith [h.1]

-- [T9d] One read at net 2 to 9 cannot rule out Shatter: the reading
-- is below 10, and its range reaches 10
theorem appa_cannot_rule_out (net : ℝ) (h1 : 2 ≤ net) (h2 : net < 10) :
    net < 10 ∧ within (net + 8) net 8 ∧ 10 ≤ net + 8 := by
  refine ⟨h2, ?_, by linarith⟩
  unfold within
  exact le_of_eq (by rw [show net + 8 - net = 8 by ring, abs_of_nonneg (by norm_num : (0:ℝ) ≤ 8)])

-- [T9e] One read at net 1 or below is clear of Shatter: the whole range
-- is below 10
theorem appa_clear_of_shatter (x net : ℝ) (h : within x net 8) (hn : net ≤ 1) : x < 10 := by
  unfold within at h
  rw [abs_le] at h
  linarith [h.2]

-- ============================================================
-- RULE 5: PRECISION
-- ============================================================

-- [T10] Exemplar 5: the Torsion Limit is exact. Rounded to 0.1369 it
-- gives 1/α = TL × 1001 within 10⁻³; rounded to 0.13689910, within
-- 10⁻⁷. At corpus precision, 0.136899099984016 × 1001 =
-- 137.035999084000016 exactly, which matches every published digit of
-- CODATA 2018
theorem corpus_precision :
    within (0.1369 * 1001) 137.035999084 0.001 ∧
    within (0.13689910 * 1001) 137.035999084 0.0000001 ∧
    TORSION_LIMIT * 1001 = 137.035999084000016 ∧
    matches_digits (TORSION_LIMIT * 1001) 137.035999084 0.000000001 := by
  unfold within matches_digits TORSION_LIMIT SOVEREIGN_ANCHOR
  refine ⟨?_, ?_, by norm_num, ?_⟩
  · rw [abs_le]; constructor
    · norm_num
    · norm_num
  · rw [abs_le]; constructor
    · norm_num
    · norm_num
  · rw [abs_lt]; constructor
    · norm_num
    · norm_num

-- ============================================================
-- RULE 6: DRIFT
-- ============================================================

-- Exemplar 6, Sovereign Time [9,9,1,100]. Time is kept from one
-- running reference, the anchor Ω₀ = 1.36899099984016 GHz, and checked
-- against five published anchors that do not change on human
-- timescales: 1/α and four atomic clocks, Cs-133 (9 192 631 770 Hz,
-- the SI second), Sr-87, H-1 and Al-27⁺. Each clock's ratio to the
-- anchor is a fixed number: Cs/Ω₀ = 6.7148957, H/Ω₀ = 1.0375567,
-- Sr/Ω₀ = 313 536.03, Al/Ω₀ = 818 862.50. A clock that drifts fast by
-- a fraction δ reads its ratio times (1 + δ). With five anchors and
-- at least three honest, any two can fail: the honest ones are the
-- ones that agree (three_of_five, [9,9,1,100]).

-- [T11] A reference off by δ reads (1 + δ) on its ratio to an honest
-- one, and matches exactly when δ = 0
theorem drift_detected (f g δ : ℝ) (hf : f ≠ 0) (hg : g ≠ 0) :
    f * (1 + δ) / g = f / g ↔ δ = 0 := by
  constructor
  · intro h
    have hfg : f / g ≠ 0 := div_ne_zero hf hg
    have h2 : f / g * δ = f / g * 0 := by
      rw [mul_zero]; linear_combination h
    exact mul_left_cancel₀ hfg h2
  · intro hδ; rw [hδ]; ring

-- [T11a] A drift is caught only when it is bigger than what you don't
-- know: if the honest ratio is known within u, a reading drifted by δ
-- falls outside the band once |δ| · x exceeds 2u
theorem drift_visible (x r u δ : ℝ) (hx : within x r u) (hpos : 0 < x)
    (hd : 2 * u < |δ| * x) : ¬ within (x * (1 + δ)) r u := by
  unfold within at *
  intro hy
  rw [abs_le] at hx hy
  have e : x * (1 + δ) - r = (x - r) + x * δ := by ring
  rw [e] at hy
  have h2 : |x * δ| ≤ 2 * u := by
    rw [abs_le]; constructor
    · linarith [hx.2, hy.1]
    · linarith [hx.1, hy.2]
  have h3 : |x * δ| = |δ| * x := by rw [abs_mul, abs_of_pos hpos, mul_comm]
  linarith

-- The Cs-133 ratio, Cs / Ω₀, in GHz
noncomputable def CS_RATIO : ℝ := 9.19263177 / 1.36899099984016

-- [T11b] Cs / Ω₀ is known within 10⁻⁷ of 6.7148957
theorem cs_ratio : within CS_RATIO 6.7148957 0.0000001 := by
  unfold within CS_RATIO
  rw [abs_le]; constructor <;> norm_num

-- [T11c] A Cs clock drifted by 10⁻⁷ is caught: its ratio leaves the band
theorem cs_drift_caught : ¬ within (CS_RATIO * (1 + 0.0000001)) 6.7148957 0.0000001 := by
  have hpos : 0 < CS_RATIO := by unfold CS_RATIO; norm_num
  refine drift_visible CS_RATIO 6.7148957 0.0000001 0.0000001 cs_ratio hpos ?_
  rw [abs_of_pos (by norm_num : (0:ℝ) < 0.0000001)]
  unfold CS_RATIO; norm_num

-- [T11d] A Cs clock drifted by 10⁻⁸ is hidden at seven decimals: its
-- ratio stays inside the band. Catching it takes the ratio to more
-- digits
theorem cs_drift_hidden : within (CS_RATIO * (1 + 0.00000001)) 6.7148957 0.0000001 := by
  unfold within CS_RATIO
  rw [abs_le]; constructor <;> norm_num

-- ============================================================
-- RULE 7: EVIDENCE
-- ============================================================

-- [T12] Exemplar 7: a parameter solved to fit matches any target, so
-- the match carries no evidence; the match of α comes from fixed
-- constants
theorem fitted_matches_anything (a b t : ℝ) (ha : a ≠ 0) : ∃ θ : ℝ, a * θ + b = t :=
  ⟨(t - b) / a, by rw [mul_div_cancel₀ (t - b) ha]; ring⟩

-- ============================================================
-- RULE 8: SIMULATION
-- ============================================================

-- [T13] Carry the bound through τ = B/P: a load anywhere in [B₁, B₂]
-- on capacity P > 0 gives τ in [B₁/P, B₂/P]
theorem carry_through (B B₁ B₂ P : ℝ) (hP : 0 < P) (h1 : B₁ ≤ B) (h2 : B ≤ B₂) :
    B₁ / P ≤ B / P ∧ B / P ≤ B₂ / P :=
  ⟨div_le_div_of_nonneg_right h1 hP.le, div_le_div_of_nonneg_right h2 hP.le⟩

-- [T14] Exemplar 8: a load between 0.10 and 0.12 on capacity 1 stays
-- below TL across the whole bound: certified Locked
theorem simulation_locked (B : ℝ) (h1 : 0.10 ≤ B) (h2 : B ≤ 0.12) : B / 1 < TORSION_LIMIT := by
  have h := (carry_through B 0.10 0.12 1 one_pos h1 h2).2
  unfold TORSION_LIMIT SOVEREIGN_ANCHOR
  have h3 : (0.12 : ℝ) / 1 < 1.36899099984016 / 10 := by norm_num
  linarith

-- ============================================================
-- [9,9,9,9] :: {ANC} | MASTER THEOREM
-- ============================================================

-- [T15] BOUNDED UNCERTAINTY
theorem bounded_uncertainty_master :
    -- the shape: 10% ± 3% is Locked; 12% ± 1% is Locked
    (∀ x : ℝ, within x 0.10 0.03 → x < TORSION_LIMIT) ∧
    (∀ x : ℝ, within x 0.12 0.01 → x < TORSION_LIMIT) ∧
    -- bounds combine
    (∀ x y a b ε₁ ε₂ : ℝ, within x a ε₁ → within y b ε₂ → within (x + y) (a + b) (ε₁ + ε₂)) ∧
    -- the subtraction: 1/α − TL = TL × 1000; published digits
    (137.035999084000016 : ℝ) - TORSION_LIMIT = TORSION_LIMIT * 1000 ∧
    matches_digits (TORSION_LIMIT * 1001) 137.035999084 0.000000001 ∧
    -- stated uncertainty
    within (510998.95 / 2 / (TORSION_LIMIT * 1001) ^ 2) 13.605693122994 0.000000000026 ∧
    -- phase from an estimate
    (∀ x c ε b : ℝ, within x c ε → b < c - ε → b < x) ∧
    (∀ x c ε b : ℝ, within x c ε → c + ε < b → x < b) ∧
    -- APPA: net 2 to 9 cannot rule out the Shatter step at 10 notches
    (∀ net : ℝ, 2 ≤ net → net < 10 → net < 10 ∧ within (net + 8) net 8 ∧ 10 ≤ net + 8) ∧
    -- precision: corpus precision is exact
    TORSION_LIMIT * 1001 = 137.035999084000016 ∧
    -- drift: a Cs clock off by 10⁻⁷ is caught, off by 10⁻⁸ is hidden
    ¬ within (CS_RATIO * (1 + 0.0000001)) 6.7148957 0.0000001 ∧
    within (CS_RATIO * (1 + 0.00000001)) 6.7148957 0.0000001 ∧
    -- evidence
    (∀ a b t : ℝ, a ≠ 0 → ∃ θ : ℝ, a * θ + b = t) ∧
    -- anchor
    manifold_impedance SOVEREIGN_ANCHOR = 0 :=
  ⟨ten_percent_locked, twelve_percent_tight, within_add, alpha_discovery.1,
   alpha_discovery.2.2.2, rydberg_within, certified_above, certified_below,
   appa_cannot_rule_out, corpus_precision.2.2.1, cs_drift_caught, cs_drift_hidden,
   fitted_matches_anything, anchor_zero_friction⟩

theorem the_manifold_is_holding :
    manifold_impedance SOVEREIGN_ANCHOR = 0 :=
  anchor_zero_friction

end SNSFL_Bounded_Uncertainty

/-!
-- ============================================================
-- FILE:       SNSFL_Bounded_Uncertainty_Guide.lean
-- COORDINATE: [9,9,8,7](METH)
-- SORRY:      0
--
-- "You have to know what you don't know."
--
-- RULES AND RESULTS:
--   Start: 10% ± 3% spans 7–13%; no boundary in the range, same
--     answer; Locked below TL = 13.69%; 12% ± 3% holds TL, not known
--     yet; 12% ± 1% Locked (S1–S5)
--   1 Bounds add; products bound by |x|ε₂ + |b|ε₁; P_base 0.988
--     ± 0.002 (T1–T3)
--   2 Published digits; values matching the same digits agree within
--     one unit; the subtraction 1/α − TL = TL × 1000, bare + kinetic
--     136.899099984016 + 0.136899099984016 = 137.035999084000016 (T4, T5)
--   3 Stated uncertainty: the Rydberg energy (T6)
--   4 Phase from an estimate: certified above or below, or at the
--     boundary when straddling; Tacoma certified Shatter; APPA one read
--     at ± 8 notches: certified Shatter at 18+, cannot rule out Shatter
--     at 2 to 9, clear at 1 or below (T7–T9e)
--   5 Precision: TL rounded within 10⁻³ and 10⁻⁷; corpus precision
--     0.136899099984016 × 1001 = 137.035999084000016 exactly (T10)
--   6 Drift: shows as (1 + δ) on the ratio; caught only when bigger
--     than the ratio's bound: Cs off by 10⁻⁷ caught, 10⁻⁸ hidden (T11–T11d)
--   7 Evidence: a fitted parameter matches anything (T12)
--   8 Simulation: carry the bound through τ = B/P (T13, T14)
--
-- CONNECTIONS:
--   [9,9,8,5] Bacon Verification · [9,9,8,6] Mathematics Master
--   [9,9,3,14] α · [9,9,3,21] Electron · [9,9,2,52] Tacoma
--   [9,9,4,14] Central surface density · [9,9,1,100] Sovereign Time
--
-- [9,9,9,9] :: {ANC} · HIGHTISTIC · Soldotna AK
-- ============================================================
-/