Name the open problem, conjecture, benchmark, or research question. Include a reference link if one is easy to provide.
Submit a record
Share a documented
research attempt.
Start with the problem, the prompt, and what happened. We can follow up on dates, sources, validation, and other metadata.
See what to include ↓A short starting point
Four things are enough to begin.
Submissions do not need to arrive as finished archive records. Specific links and artifacts are useful, but we can research and structure the remaining metadata afterward.
Share the prompt itself, a public link, or whatever prompt material you are permitted to provide. Model details are helpful but can be incomplete.
Briefly describe the outcome and link any response, proof, code, experiment, transcript, or other artifact you have.
Your GitHub account is usually enough. Add a preferred public contact method only if you want us to use something else.
Preliminary scope
Keeping the first collection focused and useful.
For this early phase, priority goes to identifiable open or research-level problems in mathematics and science with a documented AI-assisted attempt.
After a quick review, an agent can locate sources, check for duplicates, organize metadata, and identify a few useful follow-up questions.
Not every submission will become a record, particularly when the research target or attempt cannot be identified clearly. Opening an issue is simply a starting point.
GitHub issue intake
A brief submission form.
The form asks for enough information to start a review. You will need a GitHub account to open an issue.
- Problem or research question
- Prompt or prompt link
- Outcome and any available artifacts
- Submitter relationship and a way to follow up
A private intake path can be added later for sensitive cases. For the initial release, the GitHub form is intended only for material that can be discussed openly.