# SPDX-License-Identifier: Apache-2.0 # SZL Holdings — Open-Problem Bounty Board · Doctrine v11 id: conjecture-2-khipu-bft-safety title: "Khipu Byzantine Quorum Safety — Conjecture 2 (OPEN)" status: OPEN conjecture: 2 doctrine: v11 summary: > Unconditional Byzantine-fault-tolerant Khipu quorum safety (no split-brain under an equivocating organ) is Conjecture 2 — the genuine open BFT frontier. It is NOT a theorem under bare hypotheses. problem_statement: | The Khipu quorum protocol must guarantee agreement (no two honest quorums certify conflicting verdicts) in the presence of Byzantine organs. The open obligation is `ubuntu_quorum_safety` in `Lutar/Innovations/round12/Identity_Ayni_Quorum.lean`, stated unconditionally. A Byzantine organ can equivocate (sign two conflicting votes), so the unconditional statement cannot hold without the right charter and honesty hypotheses; that is exactly why it remains a conjecture. the_gap: | The conditional safety theorem is already proven (Wave 23, `khipu_quorum_safety_conditional`): under {n >= 3f+1, |faulty| <= f, |quorum| >= n-f, honest non-equivocation} two quorums cannot certify conflicting verdicts. The OPEN gap is whether the unconditional obligation can be discharged, or proven to require a strictly weaker hypothesis than honest non-equivocation — while never asserting the false unconditional-without-honesty statement. already_proven_do_not_reclaim: | Wave 23 `khipu_quorum_safety_conditional` and the Wave 13 `quorum_agreement_single_valued_vote` shadow are CONDITIONAL / non-Byzantine and already proven. This bounty is the UNCONDITIONAL frontier only; do not represent the conditional results as settling Conjecture 2. target: theorem_name: ubuntu_quorum_safety file: "Lutar/Innovations/round12/Identity_Ayni_Quorum.lean" repo: szl-holdings/lutar-lean acceptance_criteria: - id: lake-build-green check: "`lake build` is green on the pinned toolchain; lutar-lean lake-build.yml + lean.yml pass." - id: no-sorry check: "No `sorry` / `sorryAx` in the target obligation and its dependencies." - id: axiom-allowlist check: "`#print axioms ubuntu_quorum_safety` is a subset of [propext, Quot.sound, Classical.choice]; any added hypothesis must be a stated theorem premise, not a new global axiom." - id: no-new-axiom check: "No new `axiom` declarations and no `native_decide` trust escape hatches; the numbers drift gate (check_numbers_drift.py) stays green." - id: becomes-real check: "The proof is REAL: kernel-checked, zero `sorry`, in-policy axioms only." verification: arbiter: "lutar-lean CI (lake-build.yml gate + numbers drift, lean.yml kernel check) on a PR to main — automated, no bypass." must_become_real: true reward: amount: founder-set currency: USD note: > The monetary award is founder-set and published in the bounty repo's pinned issue. This board never invents a figure. extras: - "Lean co-author credit on the SZL Holdings thesis." - "Materially-advancing partial submissions are eligible for pro-rata recognition at founder discretion." submission: intake_repo: https://github.com/szl-holdings/lutar-lean pull_request: "Fork lutar-lean, discharge `ubuntu_quorum_safety`, open a PR to main. CI is the arbiter; green = eligible." references: - "Lutar/Wave23/QuorumSafety.lean" - "Lutar/Innovations/round12/Identity_Ayni_Quorum.lean" - "Lutar/Wave8/Byzantine.lean" - "../README.md (section: The Λ line — Conjecture 1; Byzantine BFT safety row)" honesty: > Unconditional Khipu BFT safety is Conjecture 2 — it is NOT a theorem. The unconditional-without-honesty statement is false (a Byzantine organ can equivocate). Only the conditional form is proven; never represent this OPEN conjecture as proved. A submission clears the bar only when the kernel verifies it (REAL).