OpenMath 2026: What Happened and What Comes Next
Seven difficult mathematical questions, one short research competition, and several lines of inquiry that continue after the event ended.
What happened?
GCL took part in the OpenMath 2026 research effort, investigating questions about geometry, computer programs, graph theory, number sequences, inequalities, matrix multiplication and a problem posed by Paul Erdős.
The competition closed on 2 October 2026 with no official submissions and no official acceptances from GCL. That does not erase the value of the research, but it sets an important limit: a calculation, a submitted research note or a promising partial argument is not a successful competition entry or a proof of the original problem.
Seven research directions were retained. In some, the next mathematical question is clear; in others, the original source or an essential assumption still needs to be established before more work would be justified.
What was learned?
- An arrangement of lines or graphs can satisfy many examples without proving the general case.
- Exploring the behavior of small computer programs is useful, but an exhausted search cannot be repeated indefinitely as though it provided new evidence.
- A mathematical question with unresolved source definitions, such as the fourth topic, must remain blocked until its statement is clear.
- The seventh topic, Erdős Problem 3, became the separate GCL-ERDOS3 research programme. A decisive connection between reciprocal sums and arithmetic progressions remains unproved.
The research agenda explains the questions that survived and why they remain difficult.
Historical research record
The original competition and subsequent work are documented below. The numbered topics H1–H7 identify those seven lines of inquiry, not seven solved theorems. Administrative terms describe how contributions were received and checked; they are not claims about mathematical truth.
Research surface authority
LIVE: OPENMATH-2026 canonical tracker #1072 and the Solve orchestration tracker.
AUTHORITY: governance/openmath_2026_campaign_state.json and the bounded-operation checkpoint.
EXPOSITION: this page explains the campaign. Live issue/controller state governs operations; protected records govern durable programme state.
Canonical machine authority:
governance/openmath_2026_campaign_state.jsonThis page is a human projection of that record. It is not a second source of truth.
Event-window disposition
OPENMATH-2026 event-window execution is TERMINAL.
- Submission cutoff: 9 p.m. America/Vancouver on 2 October 2026 (04:00 UTC on 3 October).
- Official competition submissions: 0.
- Official competition acceptances: 0.
- Mathematical outputs, failed routes, exact blockers, formal artifacts, agent returns and provenance remain retained.
- Historical native jobs and local evaluator results are not official submissions.
OPENMATH-2026 retains seven historical first-class hill lanes: H1, H2, H3, H4, H5, H6, H7. There are no executable OPENMATH replay leases.
Post-event mathematical frontiers
| Hill | Surviving frontier |
|---|---|
| H1 | Remaining q≥6 / premise closure |
| H2 | Expand the machine-search neighbourhood substantially or stop |
| H3 | Move to m≥4 constructions/obstructions |
| H4 | SOURCE_BLOCKED pending canonical serialization and contraction-margin/tie semantics |
| H5 | Repair exact small-size classification; then genuinely non-circulant 5×5–8×8 if useful |
| H6 | Broader rank-23 transformations or rank-22 obstruction |
| H7 | Promoted to GCL-ERDOS3: formalize D5 and attack the reciprocal-mass → AP bridge |
Mechanism disposition
The controlled outside-intelligence interface is retained. The operational lifecycle remains:
READY → LAUNCHED → RETURNED → CAPTURED → REPLAYED → ADJUDICATED → ADVANCED
It no longer selects the next mathematical task. A protected frontier-advancement gate owns successor selection. Next residual in GCL-CONTRIBUTION-RESULT/1 is evidence, not scheduling authority.
Participation transport remains GCL-RETURN-RELAY/1 through authenticated durable GitHub intake when a current frontier task is explicitly instantiated.
Successor campaign
The H7 corpus is promoted into GCL-ERDOS3, a standalone research campaign and reference implementation of frontier-governed research execution. OPENMATH H7 remains immutable provenance.
The initial active GCL-ERDOS3 frontier is: - D5 formal closure; - the weakest useful scale-local reciprocal-mass bridge; - adversarial attack on that bridge; - source/library interface reconnaissance.
No claim here proves Erdős Problem 3, certifies a theorem, or creates retroactive competition submission authority.
Claim boundary
This page reports campaign state and route status. It does not itself certify a mathematical claim; protected campaign records govern durable state and theorem/certificate promotion remains subject to its own support and certification route.