Editorial guidelines
Eight rules, most of which come down to the same thing: an entry should save the next reader time, and a submission should be honest about what it establishes.
A problem must be decidable in principle by a reader
Someone who has never seen the problem should be able to tell, from the statement alone, whether a proposed object or argument settles it. That means quantifiers in the right places, notation defined, and no reliance on a convention the author has in mind but has not written.
Vagueness is the most common reason an entry is sent back. “Understand the behaviour of X” is a research direction, which is fine — label it a programme and say what a first result would look like.
Provenance is not optional
Every entry names who posed the problem and where. A conference open-problem paper, a numbered entry in a standing problem list, a textbook exercise marked open, a survey's closing section, or a documented conversation all qualify.
If a problem is genuinely yours and unpublished, say so and mark the source as posed here. That is allowed and is tracked separately, because a problem nobody else has looked at carries different weight.
Say when you last checked that it is open
Problems get solved quietly. A statement that a problem is open is a claim about the literature at a particular date, and entries record that date. A search of the citing literature, or a check of the relevant arXiv listing, is enough.
Known partial results are the most valuable section
The list of what has been tried and where it stops is what makes an entry worth more than a one-line question. It should include the barriers: the technique that provably cannot close the gap, the counterexample to the natural strengthening, the numerical evidence that points the other way.
Submissions are read, not refereed
An editor reads every submission and marks it accepted, under review, or declined. Accepted means the contribution is coherent, correctly attributed, and adds something the entry did not have. It does not mean the mathematics has been verified line by line.
A claimed full solution is marked under review and stays there until someone competent has read it. That can take a long time, and pressing does not speed it up.
Negative results and dead ends are welcome
A documented failure — this reduction does not work, and here is the obstruction — is a first-class submission. It is also the category the literature systematically loses, because there is nowhere to publish it.
What gets declined
Claimed proofs of famous conjectures with no argument, or with an argument that ignores a known barrier without mentioning it. Restatements of the problem presented as progress. Submissions generated wholesale by a model and posted without the submitter having read them — machine attempts have their own section, and that is where they belong.
Corrections beat politeness
If an entry is wrong, say so plainly in the discussion. Wrong attributions, stale statuses and mis-stated hypotheses are fixed on report. There is no cost to being the person who points out that the index is wrong.
Search before you pose
The index starts from the twelve thousand records of the OpenFrontiers collection, so many well-known problems are already here under another name. Search first. If the problem is here but the entry is stale or thin, submit the missing reference or partial result to that entry instead of posing a duplicate; editors decline duplicates and point to the existing entry.
A problem is marked solved only when an editor accepts a solution or counterexample submitted against it. A pointer to a paper that settles it counts — submit it as a pointer to literature.
Votes are attention, not verdicts
A vote on a problem says it deserves attention; a vote on a submission says it was useful. Neither says anything is correct, and neither changes a status. You cannot vote on your own contributions, and pressing the button again withdraws your vote.