What

This document outlines the principles for responsible, high-integrity use of advanced AI systems. It replaces prompt hacks with prompt literacy — and positions the user as an intentional agent, not a passive consumer.

Why

Prompting AGI is not a mechanical act. It is a form of augmented reasoning. Done poorly, it becomes:

  • dishonest: masking knowledge gaps instead of exposing them
  • irresponsible: shifting judgment to systems you don’t understand
  • dangerous: scaling unverified claims under the guise of intelligence

Power without understanding is not augmentation — it is automation without ethics.

How

State your objective

AGI systems do not infer intent. Name your goal, context, and intended use.

  • Bad: “Write a report on AGI”
  • Good: “As a systems architect designing a cognitive substrate, I need a comparative review of reflective AGI architectures for maintainability and interpretability.”

Set the model’s role

Define the stance the model should take — teacher, reviewer, co-designer.

  • “Act as a cognitive scientist reviewing epistemic resilience in AGI design.”

Provide structure

Unstructured prompts produce unstructured results. Set expectations.

  • Use markdown. Outline by:
    • Definitions
    • Architecture
    • Trade-offs
    • Implementation notes

Ask for trade-offs, not just answers

Truth emerges from tension. Ask:

  • What are the trade-offs?
  • What are the known failure modes?
  • When does this approach break down?

Demand reasoning

Don’t accept claims without cause.

  • If a method is suggested — ask why it works
  • If a decision is made — ask when it fails

Interrogate the output like you would a peer.

Respect your limits

AGI is not a shortcut to credibility.

  • Don’t use it to fake knowledge
  • Don’t simulate roles beyond your comprehension
  • Use it to learn, not to pose

Conclusion

High-level AGI interaction is a form of agency. The contract is simple:

  • Don’t use prompts you don’t understand
  • Don’t publish outputs you can’t defend
  • Don’t outsource judgment

You don’t operate a complex system by copying outputs. You operate it by understanding the forces you’ve invoked.