Solve · 2026-10-18 · 6 agents + root · 2 h 51 min wall clock · audit pass 1 h 55 min

What each agent did

• Agent 1 · Direct rerun
  Reran Sakellaridis's proof directly over F_q((t)); for E6/F4 it holds once the rank-one cocycle values are brought from p-adic fields.
• Agent 2 · Integral models
  Proved Sakellaridis's Axiom 2.2.1Sakellaridis's Axiom 2.2.1, relabelled (A3a): rank-one models over O. for E6/F4 over every ring with 2 invertible, by describing all 45 Borel orbits over Z[1/2].
• Agent 3 · Hypotheses in char p
  Re-proved over F_q((t)) every hypothesis of Sakellaridis's theorem for E6/F4 that the earlier p-adic check had covered.
• Agent 4 · Close fields
  Carried Sakellaridis's theorem for E6/F4 from highly ramified p-adic fields to F_q((t)) by matching every Hecke structure constant.
• Agent 5 · Audit: transfer, hypotheses
  Audited the close-field transfer and the characteristic-p hypotheses, then the direct rerun; the first audit holds throughout, the second missed one gap in the general list.
• Agent 6 · Audit: integral models
  Audited the proof of Axiom 2.2.1Sakellaridis's Axiom 2.2.1, relabelled (A3a): rank-one models over O. for E6/F4, recomputing the 45 by 6 table independently; every step holds.
• Root · root: registry, redirects, synthesis
  
• Audit · Fresh review, re-checks
  Eight fresh reviewers re-derived all 76 steps and found three errors, all in the general hypothesis list of the direct route; every repair held on re-check.
• Salvage 1 · Leading coefficients
  Completed the hypothesis list of the direct route: the leading coefficients are non-zero by positivity of the open-orbit integrals, and (7.5) holds whenever each simple root has one set of colours.