For three years, I have worked independently in South Africa using OpenAI models as persistent engineering collaborators.
I began without a formal research team, institutional laboratory, sustained funding, or an advanced mathematical background. My starting point was basic: learning what prime numbers are, how Python works, what an algorithm is, and how computational experiments can be recorded and tested.
That work gradually developed into Art of Primality War, a long-term human–AI engineering programme built on ordinary consumer hardware.
The programme now includes:
- prime-search and candidate-generation systems;
- deterministic chained-search and prediction workflows;
- frozen-source and SHA-256 provenance records;
- replay and no-rescan controls;
- external validation gates;
- checkpointed CPU and GPU computational lanes;
- exact data-carrier experiments;
- curriculum, assessment, and evidence-generation systems;
- a versioned lifetime research archive containing telemetry, failures, corrections, outputs, and development history.
The important claim is not that every interpretation has already been proven.
The important fact is that a solo person, working with OpenAI models over three years, has produced a substantial body of executable software, persistent evidence, reproducible artifacts, test results, and real-world systems under severe resource constraints.
I have deliberately separated the work into three classes:
Demonstrated implementation
Executable scripts, deterministic artifacts, source hashes, checkpoints, reports, smoke tests, replay controls, external checks, and documented computational runs.
Provisional technical claims
The predictive value of certain search structures, efficiency claims requiring matched benchmarks, and the broader transferability of the systems.
Future hypotheses
Record-scale discoveries, larger scientific implications, commercial deployment, and the capability that may become possible with reliable access to frontier models and compute.
I am currently preparing a professional request to OpenAI for a structured technical-founder evaluation.
The request is specific:
Provide a defined pilot involving frontier-model or API access, compute credits, technical mentorship, and an opportunity to present the strongest results under controlled and reproducible review.
I am not asking the community to endorse extraordinary claims without inspecting the evidence.
I am asking a more practical question:
What is the correct pathway for a solo technical founder to place three years of auditable human–AI engineering work before the appropriate OpenAI startup, partnership, or technical-review team?
The project was developed while I was unemployed and under significant financial pressure. That circumstance is not presented as evidence of technical merit. It explains why subscription limits, interrupted access, and restricted compute materially affect development throughput.
The technical case must stand on its own.
I can provide, where appropriate:
- a one-page executive memorandum;
- a technical evidence brief;
- a claim-to-artifact index;
- frozen source files;
- SHA-256 records;
- runbooks and expected outputs;
- controlled demonstration material;
- a versioned research archive.
I would value precise guidance from community members who have successfully navigated OpenAI startup, partnership, research-access, or technical-evaluation channels.
I am particularly interested in answers to these questions:
- Which official OpenAI route is best suited to a solo founder with substantial existing technical artifacts?
- What evidence package is most likely to receive serious technical review?
- Should the initial submission focus on one reproducible system rather than the entire three-year programme?
- Are there founders, researchers, or community reviewers willing to examine a narrowly scoped demonstration before formal submission?
- What credibility failures should I avoid when presenting ambitious work that contains both verified implementation and still-unproven hypotheses?
This is not a request for unquestioning validation.
It is a request for the work to be evaluated at the level of its evidence.
Art of Primality War
Johannes Frederik Kruger
South Africa
From basic questions to executable systems. From isolated effort to auditable evidence. The next step is disciplined review.