A hunt, not a department and not a result. Nothing here is evidence for or against RH (docs/08).
What this is
A bounded support run launched from fulcrum by the run working hunts/r_186989 (Erdős #126 scout). One question, assigned:
For each $p \in S$, analyse the graph or ultrametric relation induced by $p \mid a_i + a_j$ and by its exact valuation. The union covers all edges of $K_n$. Seek a structural cover, Ramsey, rank or container bound exploiting that these are not arbitrary graphs. Prove a sharper cover theorem or exhibit a realizable obstruction showing why generic graph methods stall.
Audit hunts/r_186989/RESULTS.md rather than trusting it.
What it may touch
hunts/support_d0f0aefb/ and one case-log entry in hunts/README.md. Nothing else: not zeta/, harness/, lean/, meta/, any other hunt directory, or any root markdown file.
Standing constraints
probe.pyis stdlib only, runs in seconds, and writesresults.json.- Finite data is not a proof and is labelled measured; proofs are written out and labelled proved.
- The reserved word belongs to
zeta/rigor.pyand the Lean arm, and is not used here. - Implication directions are stated explicitly, because the parent hunt already found one arm whose lemma pointed the opposite way from the one briefed.