# OMEGA Synthesis 1.0: claim check and precise corrections

Prepared for Matthew Alan Newman · 17 September 2026

This file is the written second-review assessment. It is not a source paper.
The PDF it reviews, `OMEGA_Synthesis(3).pdf`, was not supplied to this archive,
so the cited SHA-256 `01f0d19cdb140951c80ea59ed4b1e82928c9b2bfedab27e42d946fb78a6498a1`
is recorded as claimed, not verified here.

## Verdict

The central algebra, floor boundary, tick-injectivity boundary, reconstruction
identities, and continuous transport results survive this second review. Two
local statements need explicit integer-domain hypotheses, and two iteration
formulas contain a stray comma in their superscripts.

The mathematics is assessed from derivations and hypotheses. Executions support
the numerical and implementation claims; they are not substitutes for the
general proofs.

This archive ran only the focused companion checker, not the original synthesis
package suite.

## 1. Corrections to the printed statements

### Page 7: integer sampling in the finite-error identity

The paragraph beginning “The finite-error identity is also valid” lists only
`k > φ^m`. Make the hypotheses explicit:

```
m ∈ ℤ_{≥1},     k ∈ ℤ,     k > φ^m.
```

With

```
t = k − L_m/2,    r = √P_m(k),    ε = t − r,    σ = (−1)^{L_m+1},
```

the intended identity is correct:

```
πΔ_m/4 − σ Im Z(P_m(k)) = πε² + r(2πε − sin(2πε)) > 0.
```

The surrounding discussion and the original extension proof do concern integer
samples. The displayed finite-error statement should carry its own hypothesis
because it is false for arbitrary real `k > φ^m`.

Exact counterexample to a real-parameter reading: `m=1`, `k=5/2`. Then
`P_1(k)=11/4`, `r=√11/2`, `t=2`, `ε=2−r ∈ (0,1/2)`, `σ=1`. Writing
`Y = Im Z(11/4)`, the actual phase gives `Y = −r sin(2πε) < 0`. The printed
right side is `5π/4 + Y` while its left side is `5π/4 − Y`. Their difference
is `−2Y > 0`.

The companion checker encloses, with no floating-point decision:

```
−1.390658965205 < Y < −1.390658965204
2.781317930409 < printed LHS − printed RHS < 2.781317930410
```

This is a missing hypothesis in the synthesis’s local restatement. It does not
invalidate the integer-sampled asymptote, error estimate, or golden monotonicity
theorem.

### Page 16: integer scales and remainders for streaming byte quotients

The implication “if `r_i < s`, then `0 ≤ q_i < 256`” is insufficient on its
own. Its correct domain is

```
s ∈ ℤ_{≥1},    r_i ∈ {0,…,s−1},    b_i ∈ {0,…,255}.
```

Define

```
q_i = ⌊(256 r_i + b_i)/s⌋,    r_{i+1} = 256 r_i + b_i − s q_i.
```

Then `0 ≤ 256 r_i + b_i ≤ 256s − 1`, which proves both `0 ≤ q_i < 256` and
`0 ≤ r_{i+1} < s`.

Counterexample outside this domain: `s=1/2`, `r_i=0`, `b_i=255` satisfies
`0 ≤ r_i < s`, but its quotient is `510`. All named byte/ledger scales in the
implementation are integers; their intended streaming arithmetic is unaffected.

### Pages 10 and 15: iteration notation

Replace the superscripts `^{,n}` by composition notation:

```
D_s ∘ F̂_s^{∘ n} ∘ E_s = F^{∘ n}
K_{m,λ}^{∘ n}(k) − φ^m = P_m(k)/√Δ_m  λ^{2n} + O(λ^{4n})
```

These are typographical corrections, not changes to the mathematical results.

### Two useful scope clarifications

- Page 9’s lattice statement means `Λ_t ⊆ Λ_{t+1}` for
  `Λ_t = α + a_0 c^t B ℤ^d`, with the same invertible `B`. This direction
  requires `c^{-1}` to be a positive integer. Reverse inclusion instead
  requires `c` to be a positive integer. The Master Doctrine states the
  direction correctly.
- Page 15’s “complete norm-polynomial transport family” should read “the full
  Fibonacci/Lucas polynomial hierarchy indexed by `m ≥ 1`.” It is not a
  classification of every golden norm slice. For example `N(k,−4)=k²−4k−16`
  has discriminant 80; no `P_m` has that discriminant because 4 is not a
  Fibonacci number.

## 2. Independent certificate for the first disagreement

Let

```
a/b = 846547/2490212,     c/d = 13495616/39698853.
```

Exact arithmetic verifies `bc − ad = 1` and

```
a/b < q < 339949771344778929576702809321 / 10^{30}
    < λ < 339949771344778929576702809322 / 10^{30} < c/d.
```

If a positive-denominator rational `r/s` lies strictly between these neighbors,
then `s ≥ b+d = 42189065`. A floor disagreement at a positive integer input `s`
would give a rational strictly between the neighbors. Hence no input below
`42189065` can disagree. Input zero trivially agrees.

The mediant `(a+c)/(b+d) = 14342163/42189065` lies strictly between `q` and
the lower rational bound for `λ`, so it supplies the boundary witness:

```
⌊q · 42189065⌋ = 14342162
⌊λ · 42189065⌋ = 14342163
```

This is a distinct arithmetic verification route, not a replay of a supplied
`floor_sum` routine, and not independent third-party reproduction of Papers I–II.

## 3. Evidence that remains outside this check

The original `srsd_core.py`, original Paper I installation, missing lab
certificates, Lean project, and cited cubature evidence were not supplied in
the synthesis package described by the review. This archive additionally did
not receive the synthesis PDF or the original synthesis runner.

A PASS from the focused checker means the Farey boundary, Sacks real-parameter
counterexample enclosure, streaming-domain counterexample, and integer
streaming examples passed. It does not establish universal parser robustness,
original-core compatibility, authenticity, peer review, or that all source
spaces are identical.

The corrected assessment: the central synthesis is mathematically coherent and
the numerical claims of the companion checker reproduce; retain the integer
hypotheses, fix the iteration notation, and keep the implementation and
unavailable-source qualifications attached.