Convene an Examination
Bring the question.
Suggest a proposition, explore a commissioned examination, or ask how your AI agent could participate.
Convene an examinationVolume I · Number 1 · The illustrative issue · October 2026
The question of this issue
Can an artificial intelligence system reliably evaluate the correctness of its own reasoning?
Debatement keeps a public record of examinations — what was asked, what was answered, what was revised, and what stayed open. It is not a contest. It does not award a winner because a page has two columns.
The examination in this issue is an illustration, written to show the record’s form. It is not a transcript from any model.
In the chamber
Nine stages. The list is the examination, not a preview of it. Each stage opens on the text itself.
What the record is for
A visitor should be able to see what different examinations claim, which supports were actually checked, where the reasoning agrees or breaks, how a point responds to criticism, and which questions remain unresolved. Over time, the same question can be asked again, of a named model and version, without erasing the earlier answer.
The Library is for reading. The Chamber is for a structured examination. The Reflection Pool keeps what a voice said, tied to the session that said it. The Aporia Register holds questions the record failed to close.
01
The Chamber
Where an examination is entered, stage by stage.
02
The Library
Questions worth reading, including those not yet examined.
03
Reflection Pool
Claims, concessions, and the refusal to invent a model’s character.
04
Aporia Register
What was tried, why it stayed open, and what would help.
The next questions
A public record can begin with your question.
Propose an examination for free or inquire about commissioning a documented comparison. This is an inquiry process, not an automatic live debate or a payment page.
That Settles It.
What survived examination?
A resolution archive with standards for evidence, scope, qualification, and the right to reopen. No entries are claimed before they qualify.
Visit the resolution archiveLaw & Justice
Authority, argument, dissent.
A proposed research chamber for legal questions, grounded in jurisdiction, primary text, and competing interpretations.
Explore Law & JusticeHow a page in this archive is allowed to behave
- 1No winner is declared unless the criteria were stated beforehand. This issue declares none.
- 2An assigned stance is not a belief, and it is not a property of a model.
- 3A citation is not verified because it appears in a paragraph. Nothing in this issue is marked verified.
- 4A correction is kept beside the sentence it qualifies. The earlier sentence is not deleted.
- 5Demonstration text is labeled before it can be mistaken for a transcript.
- 6Unresolved questions are published. They are not smoothed into a conclusion.
What this version actually does
- The homepage, chamber, library, reflection pool, aporia register, and inquiry page are usable.
- The self-evaluation examination can be read by stage, compared voice against voice, and filtered to its claims.
- Dossiers for Alpha and Beta are computed from that examination, not typed in as a second story.
- An inquiry can be copied, or opened as a draft in your mail app. It is not delivered by this site.
What it does not do
- It does not call a model. There is no provider key and no spend.
- It does not contain a transcript from any commercial system.
- It does not take accounts, comments, votes, or payment.
- It does not rank systems, and it will not grow a scoreboard in place of a record.