COMPELLE

The Compelle library

Better questions.
Stronger arguments.

Ideas for thinking clearly when the answer matters. Explore the research behind adversarial reasoning, learn to examine a claim, or sharpen the way you make a case.

A considered starting point

Six pieces worth your attention.

Reasoning under challenge

When a Confident AI Answer Meets an Opposing Case

What changes when a confident answer has to face an opposing case? A historical comparison, with its limits made explicit.

The design of a question

The Question Is Never Free

Before weighing the answers, examine the burdens the wording assigns. An early experiment—and the later result that qualified it.

Evidence before eloquence

The Anchor Number Rule

When an argument depends on a number, establish the source, date, and denominator before debating its meaning.

Find the real disagreement

Stasis Theory

Are you disputing facts, definitions, values, or action? Naming the point at issue makes an argument more productive.

Check a confident answer

How to Audit an AI

Five questions for inspecting instructions, evidence, evaluation, and the gap between a system’s rules and its behavior.

Put the ideas into practice

Rhetoric in Negotiation

Use framing, credibility, and audience analysis to understand a negotiation—not simply to win the next exchange.

Follow a question

Choose your reading path.

Can I trust this answer?

Start with what an AI says. Then inspect what would make it wrong.

  1. How to Audit an AI
  2. The Surrender Machine
  3. Why We Publish the Prompts

How do I make a stronger case?

Find the disagreement, expose the assumptions, and give the audience a reason to listen.

  1. Stasis Theory
  2. How to Structure an Argument
  3. Audience Analysis

What is rhetoric, really?

Build a vocabulary for the choices that make language persuasive.

  1. What Is Rhetoric?
  2. Ethos, Pathos & Logos
  3. Logical Fallacies

From the research

Interesting results deserve careful interpretation.

Our historical Bittensor experiments let us observe how AI arguments behave under pressure: when models concede, how a question changes the burden of proof, and why a convincing claim can still be false.

These notes describe particular systems and periods. A debate result is not proof of correctness or a promise of customer performance. The strongest findings come with the conditions—and the corrections—that make them useful.

Read the research papers · Understand the method

The complete collection

Explore by subject.

Open a subject to find every retained article, including the historical notes and longer guides.

AI reasoning & historical experiments
Rhetoric through history
Writing, speaking & close reading
Rhetoric at work and in public life
Foundations & reference

A question of your own?

Move from reading to a decision.

Compelle is a white-glove analysis service. We work with you to frame the question, examine the evidence, and deliver competing perspectives with an accountable conclusion.