1. Use the tool. Keep the judgment.
AI may accelerate analysis or creation, but the user remains responsible for truth, consequences and final decisions.
Principles
Intelligence without surrender.
A small set of principles for building and using AI systems that remain useful, accountable and human-directed.
2. Preserve provenance.
Outputs should remain connected to prompts, sources, models, decisions, versions and the artifacts they became.
3. Make uncertainty visible.
Confidence should not be performed. Systems must distinguish evidence, inference, approximation and absence.
4. Context belongs to the user.
The durable project record should be portable, inspectable and independent of one conversation or vendor.
5. Design for correction.
Versioning, rollback, superseded decisions and conflict handling are more honest than pretending the first output was inevitable.
6. Increase agency.
The best AI leaves the user more informed and capable, not merely more impressed.