Skip to content

OPENMATH-2026 — bounded seven-hill sprint campaign

Current-state authority: OPENMATH-2026 Current Campaign State. This document supplies charter/profile context and is not the current operational dashboard.

Programme tracker: grandchallenge/MATH-PROGRAMME#1072
Event window: 2026-09-27 through 2026-10-02
Initial state: EXTERNAL_HILL_LOCK_PENDING
Authority: bounded campaign execution under existing MATH-PROGRAMME / MATHFORGE / MATHSOLVE / MATHCERT doctrine

Purpose

Use the OpenMath 2026 competition as a live field test of the Grand Challenge mathematics stack.

The campaign objective is not merely to produce a competition submission. It is to test whether a versioned external mathematical target can be transformed into a governed parallel research graph that preserves source identity, failed routes, useful intermediate mathematics, formal artifacts, replay evidence, and claim boundaries.

The canonical research flow is:

external OpenMath hill
        |
        v
MATHFORGE source/status lock
        |
        v
MATHSOLVE obligation graph
        |
        +--> bounded internal routes
        |
        +--> zero-context CEI contributions
        |
        v
GCL adjudication
        |
        v
MATHCERT replay/formal checking
        |
        v
competition-ready artifact

External contract

The campaign is bound to the following public external surfaces:

  • AutoLab list: https://app.autolab.ai/lists/alejandrozu/openmath
  • OpenMath event/rules page: https://rsihouse.ai/openmath
  • event announcement: https://luma.com/yzp9abvr
  • Ulam / UnsolvedMath archive: https://www.unsolvedmath.com/

At campaign creation, the public AutoLab page exposed the list identity but the then-current six client-rendered hill bodies were not recoverable through the available non-interactive retrieval path. That historical acquisition state is preserved. The Human Steward later reported that an additional Erdős problem had been added, bringing the current sprint cardinality to seven. No unresolved hill title, statement, identifier, difficulty, evaluator, or formal target is inferred or invented from that high-level update.

Per-hill mathematical work is blocked until the exact hill records are acquired and recorded under the source-lock contract.

Campaign lanes

The campaign currently reserves seven hill slots:

  • OM26-H1
  • OM26-H2
  • OM26-H3
  • OM26-H4
  • OM26-H5
  • OM26-H6
  • OM26-H7

Each slot becomes executable only after its exact external identity is locked.

Shared lanes may proceed before all seven current locks are complete:

  1. source/status acquisition machinery;
  2. formalization environment inventory;
  3. common dependency and library-gap audit;
  4. submission/checker contract audit;
  5. CEI dispatch templates;
  6. provenance and disclosure templates;
  7. claim-ledger and failure-ledger scaffolding.

Shared lanes must not make substantive claims about an unresolved hill.

Required per-hill source lock

Before OM26-Hn advances beyond intake, MATHFORGE must record:

  1. exact AutoLab hill/list identity and version where exposed;
  2. exact mathematical statement as presented by the competition;
  3. Ulam/UnsolvedMath identifier and difficulty rating where applicable;
  4. primary or authoritative source statement;
  5. dated open-status triage;
  6. known-equivalence and statement-fidelity hazards;
  7. evaluator/checker identity and approved formal environment when available;
  8. artifact and provenance requirements;
  9. exact competition-rule version affecting the claim;
  10. a non-certification statement.

If these cannot be established, the hill remains SOURCE_LOCK_BLOCKED.

Thin sprint profile

OpenMath is a one-week campaign. Administrative weight must protect material epistemic boundaries without turning into research latency.

Mandatory during the sprint:

  • exact source/hill lock;
  • claim ledger;
  • proof-obligation decomposition;
  • route/failure ledger;
  • durable resumable state;
  • formal environment lock;
  • provenance/tool disclosure;
  • independent adjudication before Programme promotion;
  • MATHCERT handoff for claim-bearing outputs.

Deferred unless needed for mathematical work:

  • monograph-quality pedagogy;
  • full programme-wide synthesis;
  • nonessential documentary polish;
  • post-campaign migration and archival consolidation.

This profile changes no pillar authority.

Parallel work model

A locked hill should be decomposed into independently useful obligations rather than copied wholesale to multiple agents.

Preferred graph:

semantic/source lock
        |
        +--> formalization prerequisites
        +--> normal-form / reduction route
        +--> special-case / restricted theorem route
        +--> exact-computational route
        +--> falsification / counterexample route
        +--> independent CEI contribution route
        |
        v
candidate claim graph
        |
        v
adjudication
        |
        v
MATHCERT

Artificial fragmentation into cosmetic corollaries is prohibited.

CEI use

The Controlled Epistemic Interface may be used to obtain zero-context contributions once a hill or sub-obligation can be fully specified in one bounded dispatch.

A CEI return is evidence received, not mathematics accepted.

The intake path must preserve:

  • exact dispatch identity;
  • exact returned bytes;
  • contributor/source identity;
  • receipt and digest;
  • canonical_claim_effect=false on intake;
  • later separate adjudication.

Success conditions

This campaign may succeed through any combination of:

  • a competition-accepted formal result;
  • a useful independently replayable formalization;
  • a new restricted theorem;
  • a decisive counterexample;
  • a meaningful reduction;
  • a reusable library contribution;
  • a route-eliminating negative result;
  • a formally characterized blocker;
  • a high-quality source/status correction.

A competition score is not itself a MATHCERT disposition.

Claim boundary

This campaign does not by its existence establish:

  • correctness of any OpenMath submission;
  • novelty or priority;
  • that any listed problem remains open beyond its dated source audit;
  • semantic fidelity of a machine-checked artifact;
  • independence of an AI contribution;
  • certification;
  • publication or promotion authority.

The existing invariant remains controlling:

No Grand Intellect mathematics without a Solve route. No Solve route without lineage. No mathematical claim without a Cert handoff. No promotion by implication, omission, or repository accident.

First executable action

OM26-H1 is source-locked and active. The Human Steward reports that an additional Erdős problem has been added, so current campaign cardinality is seven. OM26-H2 through OM26-H7 remain source-lock pending. Continue organizer-authoritative acquisition for those six unresolved hills without inferring their contents.

For OM26-H1, the next mathematical action is the H1-12 multi-core incidence lane: after the protected projective reduction removes parallelism, any hypothetical 95-triangle witness is source-conditionally confined to arrangements with at least three finite multiple points. Attack that q >= 3 residual directly; do not repeat the exhausted local/topology search neighborhoods without a structurally distinct generator.