Direction

The register belongs to the field

This register is an inheritance, not an invention. Woeginger kept his page as a service to the field from 1996 to 2016. It had one keeper, so it had one point of failure. When he stopped adding entries, the list stopped breathing. We carry his 116 historical dossiers forward as a checked, sourced record, with his own notes preserved verbatim. That makes it a commons, and a commons should not be steered from behind a curtain.

P versus NP itself turns on a happy asymmetry: an answer that is brutally hard to find can be trivially easy to check. Running a register has the same shape. Finding the right policy for a hard question takes judgement and argument. Checking a decision once it is written down in public is quick work, and anyone can do it. So we publish every decision, log every change in the register's ledger, and name the ledger hash on every generated page so the record cannot be quietly rewritten. Our job is to propose. The field's job is to check, object, and redirect. This page exists so that the objecting can start before the deciding does.

Open questions

These are real decisions we have not yet taken. Each is stated neutrally, with the options as we see them. None of them has a quietly preferred answer.

• Timing of post-2016 dossiers. We hold 43 post-2016 candidate claims from the census. Should each get a full dossier as soon as it is recorded, or only after an expert review has been logged? Dossiers-first gives readers the complete picture straight away, at the risk of lending unexamined claims the visual weight of the format. Review-first keeps the dossier format meaningful, at the cost of leaving gaps in the record for years.

• Which formal gates to deepen first. The Test your proof first page already carries three families of checks: the three barrier tests (relativization, natural proofs, algebrization), the two gates drawn from the Deolalikar review, and Aaronson's eight warning signs, which help a reader choose what to check first. Each family is live today as a short prose check. The open question is which one to build out first into fuller self-service checks, because that choice decides what a hopeful prover meets first.

• Recording and appealing adjudication outcomes. When an expert refutation is logged against an entry, how should the outcome be written: as dated citations only, as a plain-language summary in the register's own words, or both together? Separately, should an author be able to request a correction or lodge an appeal against a recorded outcome, and if so, who decides it and where is that decision logged?

• Which data formats matter most. Four exports are on the table: JSON for full dossier data, CSV for the register table, BibTeX for citations, and RSS for the change feed. None exists yet, and all four are buildable in time. The question is order, and order should follow what the field would actually use first.

• Community-written refutation notes. Should the register accept refutation notes written by readers, provided every note carries a checkable receipt: the page, the line, and the exact step that fails? Accepting them would multiply review capacity in a field that has left 100 of the 116 historical claims, 86.2%, with no recorded examination. Declining keeps every adjudication anchored solely to the published literature. Both positions are defensible.

How to weigh in

There is no form, no account, and no comment box, because the site is static: no trackers, no analytics, nothing running behind the pages. The channel is email. Write to mark@pensionable.ai with the subject line Register direction. Argue for an option above, argue for one we missed, or tell us a question is wrongly framed. Every position received will be weighed before the decision is taken, and the decision itself will be logged in the ledger with its reasoning.

Corrections to existing entries are especially welcome. A wrong year, a misattributed direction, a mislabelled venue: send it with a source and it gets fixed, and the fix is logged where anyone can check it. If you hold a live copy of any of the 45 printed links that now return errors, out of 174 checked, that is treasure. Send it.

What is already decided

A few things are settled, and we state them so nobody spends effort arguing for or against them.

• Claims, never people. An entry records an event in the literature, not a judgment of its author. There will be no leaderboards and no mockery, whatever else changes.
• Unadjudicated never means endorsed. It is a fact about the field's attention, not about a claim's merit, and the register will keep saying so wherever it applies.
• Woeginger's original notes stay verbatim, visually distinct as the historical layer. We do not edit the man's words.
• No trackers, no accounts, no surveillance. Static pages, nothing watching you read them.
• Open data. The register's records are public, the ledger is hashed, and every change is logged, so anyone can verify the record without trusting us.

Everything outside that list is open. Come and steer.