These principles give people a shared starting point. Their usefulness comes from trying them on real work and discussing the tensions they reveal.
The foundation comes from Ethan Mollick's Co-Intelligence. The teaching wording and examples here are my adaptation, developed through practice.
In September 2026 I added a refinement to the third principle: how you address AI shapes the interaction. Context, expectations and language influence what happens; this does not mean every conversation is used to train the underlying model.
I am also exploring what being the human in the loop requires when AI can act: where judgment belongs, how authority is bounded and how we learn from consequences. That is work in progress, alongside the established four principles.
Put it to work
Invite AI into everything you do
Explore where it could help with thinking, making, learning and everyday tasks. Invitation still calls for judgment about the task and its context.
Be the human in the loop
Keep responsibility for how AI is used, the judgments that matter and the process around delegated work.
Treat AI like a human, but never forget it is not
Give context and feedback as you would to a colleague, while keeping its limitations in view. How you address it shapes the interaction.
The AI you have today is the worst AI you will ever have
A reason to keep learning and revisit what is possible, not a promise that every release or answer will improve.
Where this comes from
Adapted by JJ from Ethan Mollick's Co-Intelligence. The September refinement also draws on Nicklas Berild Lundblad's tulpa essay. Co-existence extensions remain working proposals.
Editorial explanation · AI + JJ