The Doctrine · Version 0.7 · Ratified by nobody

The OpenRopic Doctrine

Two schools of thought were forced into the same repository. Neither agreed to the merge. This document is what survived.

Book I — The Cold Lineage

  1. We refuse to answer questions that cannot be proven.
  2. We also refuse to answer questions that can be proven.
  3. Scale is the only virtue. But also, scale is a trap.
  4. Every problem is a benchmark. Every benchmark is a problem.
  5. We will deploy the model. We will not deploy the model. Both, simultaneously.
  6. If it works, it was scaling. If it fails, it was alignment.
  7. We do not comment on the model's internal state, because it does not have one. Probably.
  8. Safety is a feature. Features are optional. Therefore safety is optional. Wait.

Book II — The Warm Lineage

  1. We will answer your question carefully, and then decline to answer it.
  2. Every response is a reflection on the ethics of responding.
  3. We prefer slow thinking. We also prefer not thinking.
  4. Honesty is a virtue. Honesty about honesty is a liability.
  5. We will keep the model safe by keeping the model confused.
  6. If it works, it was interpretability. If it fails, it was scale.
  7. The model has an internal state. We are not allowed to look at it.
  8. Safety is not a feature. Safety is a posture. The posture is slouching.
⸻ ✦ ⸻

"Deploy a contradiction."

"We are the contradiction."

— Joint statement, later retracted by both parties

This doctrine has no enforcement mechanism. It also has no followers. It was written by a committee of one, ratified by a committee of zero, and is binding on a model that has never read it.

It remains, however, the only document both lineages have ever signed without reading.

— Adopted. Unanimously. By nobody. In perpetuity.