Thirteen ideas, one per screen: a framework that outlasts the tips and tricks. The drama is in the watching, so take the silences.
Set the foundations, see how the tool actually works, learn the four competencies, then put them into practice.
AI Fluency is the ability to work with AI in ways that are effective, efficient, ethical, and safe.
The shift is treating AI as a collaborator you build a working relationship with, not a technology to study, and not a vending machine you query.
Key conceptThe exact definition: AI Fluency is the ability to collaborate with AI in ways that are effective, efficient, ethical, and safe, achieved through the 4Ds. → teacher notes.
From handing over a defined task, to thinking together, to setting AI up to act on your behalf.
These aren't rungs to climb; they're modes you pick per task. The four competencies apply to all three.
Four competencies that, together, make collaboration effective, efficient, ethical, and safe.
Delegation decides what's worth doing together; Description and Discernment loop in the middle; Diligence keeps it responsible. The rest of the course is these four.
A language model creates new content from patterns it learned; it isn't retrieving stored answers.
Three things had to arrive together: the transformer architecture, vast training data, and a leap in compute. Knowing it generates rather than retrieves is what makes its limits make sense.
Versatile across language and tasks, but bounded by what it knows and how far it can reason.
The limits aren't bugs to wait out; they're the reason you stay in the loop. Pair its breadth with your judgment, creativity, and oversight.
Before you prompt anything, work out which parts are yours, which are the AI's, and which are shared.
It takes both: your domain expertise and a real sense of the tool. The goal isn't to automate everything; it's the best split for this task.
Pick a real project, agree what good looks like, break it into tasks, then mark who does each.
The course carries one project through every competency. Naming success up front is what makes Discernment possible later: you can only judge against a target.
Three layers: the result you want, the way it should get there, and how it should work with you.
AI can't read your mind, and it's a partner, not a database. Process is the layer most people skip, and it's often as decisive as the goal itself.
Practical techniques that turn a vague ask into a clear brief.
Prompting is iterative, not one-shot; expect to refine. The shortcut most people miss: hand the AI your prompt and ask it to make it better.
The flip side of Description: evaluate the output, the approach, and the behaviour, the same three layers, seen from the other end.
Your expertise is what makes discernment sharp: you can only catch what you'd notice. Even the best models benefit from a human checking the work.
Real work happens in the loop between describing what you want and judging what you get.
This is where collaboration beats either side working alone. You stay the one who integrates judgment and owns the final call.
The responsibility layer: how you choose the tool, how open you are, and what you stand behind.
Different contexts expect different disclosure. A diligence statement names AI's role in the work and your responsibility for the output.
Key conceptDiligence is the responsibility D, in three parts: Creation (which AI, and how), Transparency (open about AI's role), Deployment (verify and vouch for what you share). → teacher notes.
Delegate, describe, discern, stay diligent, across automation, augmentation, and agency.
Treat AI as a thinking partner, hold to effective, efficient, ethical, and safe, and the framework keeps working as the models change.
Four competencies, three ways to work, one habit: practice.
tutto.one/praxis · daniel@tutto.one