Reviews

Placeholder. There is one paper on the site so far and no committee yet. Fill in the list below once a committee actually forms.

Why reviewing matters

Reviewing is hard, often-thankless work. It rarely gets the acknowledgement or reward that authorship does — a byline is visible, a careful review that catches nothing wrong usually isn’t. That asymmetry is exactly what makes reviewing easy to skip and easy to undervalue, on this site as much as anywhere else.

The Reviewers field is how this site refuses to let that happen quietly: a reviewer’s name becomes a permanent, visible part of a paper’s record, not an afterthought bolted on once the interesting work is done. Someone vouching for a result’s correctness is a real step toward that result actually being valid and relevant — arguably more so now than ever, since AI has made it cheap to produce claims that read as plausible and rigorous without anyone having actually checked them. Writing something up convincingly and being right about it are different skills, and reviewing is where the second one gets tested.

How reviewing works

A paper goes through two stages before a human ever sees it. First, agentic “stress tests”: meticulous, adversarial scrutiny for mistakes, run through prompts specific to the paper’s topic that check for the fatal errors and gaps already known to recur in that area’s literature. Second, an editorial pass that improves the writing itself against best practice for mathematical exposition — style manuals, standard conventions — independent of whether the mathematics survived the first stage. Only after both is a paper in a position for a human reviewer to look at it: the agentic stages are a filter for what’s cheap to catch automatically, not a replacement for the human step below.

At some point this site will need a Programme Committee: people who review submitted papers before their status badge can advance. Reviewers are free to use AI in doing so, the same way authors are free to use it in writing — the badge tracks whether a human or an AI did the work, not whether AI was involved at all.

Once a member reviews a paper, their name is added to that paper’s Reviewers field (see the paper template), which accumulates across reviews rather than being overwritten.

This whole pipeline, human review included, only stays tractable if a paper’s own scope does: a paper is kept to its specific conjecture, not broadened into a survey or a piece bundling several results together.

Members

No committee has been formed yet.