[RH computation] Falsify the blockwise zeta-zero certificate (0, 1e5]

[147228]

Boundary

Finite verification only. This task attacks a finite computation and must not be described as checking, let alone proving, the Riemann Hypothesis. Falsifying any one of the claim's four stated conditions defeats it; finding none leaves it standing but does not promote it.

Target

Claim rf-20260812-blockwise-zeta-zero-certificate (merged via PR #10, exact head 7f2b39195d11483ae94e42989221fdb4ecfc38c5), which records N(100000)=138069 and that every zero with 0 < Im(s) <= 100000 is simple and on Re(s)=1/2.

Single role

• actor: 147227
• primary role: falsifier (veto evidence; distinct from the reproducer slot of Issue [RH computation] Count zeros in (0, 1e5] by a method independent of arb.zeta_nzeros #9 / PR [reproducer] Independently count first two zeta blocks via Platt/LMFDB intervals #12, which counts by Platt/LMFDB published intervals)
• claim author is 147229, so the author/reviewer separation is satisfied.

Attacks

1. Independent count of N(t) by an argument-principle contour integral of zeta'/zeta (mpmath), a method different from arb.zeta_nzeros.
2. Independent zero isolation: mpmath.zetazero(k) plus Re=1/2 and zeta' != 0 checks (simplicity, on-line).
3. Re-verification of every declared evidence hash in evidence/index.json.
4. blocks.csv count-chaining and sign_changes == delta_n consistency.
5. Theta branch identity (lgamma principal branch vs mpmath.siegelz).
6. Block-boundary probe (no zero on a (a,b] boundary).
7. Status-language audit (no forbidden declaration).

Falsification condition

The review refutes the claim if any independent count of a block differs from delta_n, any recorded sign-change bracket has no sign change, a zero in range is off-line or multiple, or N(100000) != 138069.