Methods

Bullshit Centrifuge is the public architecture. Its methods remain separated so that no layer can quietly grade its own homework.

Status · Published

Question → hypothesis or provisional theory → development through human–LLM conversation → smallest available test → observed result → success, failure, or ambiguity → revised interpretation → next question

Working method

The sequence is a recurring discipline, not a claim that every project developed linearly. Questions can return, tests can expose a badly formed hypothesis, and later work can revise the meaning of an earlier result.

Hostile review

Do not tell me I am right. Show me where the idea breaks.

A review is useful when it states its retrieval floor, separates measurement from inference, constructs serious alternatives, identifies a falsification test, and changes the next action. Model review is an analytic aid, not independent verification.

Evidence boundaries

• A fluent answer is not proof.
• A prototype is not a theory.
• A repeated claim is not independent evidence.
• A failed experiment is not wasted work.
• An unresolved question is a legitimate research result.

Human authority

The human provides the question, purpose, imagination, dissatisfaction, and judgment. The language model helps expand, structure, test, and operationalize the idea. Evidence and human review determine what survives.