An AI course that asks you to start with zero AI
What you'll get
A simple framework for deciding which parts of creative work AI should support — and which parts you should protect.
Anthropic just released a free course called AI Fluency for Creative Work, and it starts in an unexpected place. It starts at zero AI and asks you to justify why the tool should come in at all.
For a company that makes its living selling access to a language model, that's an unusual opening. Which is exactly why it's worth a look.
A quick word on what it is
The course sits in Anthropic Academy and was built together with Rick Dakan (Ringling College) and Joseph Feller (University College Cork), the two academics behind what's called the AI Fluency Framework. It's aimed at writers, designers, illustrators, musicians and others who make things for a living. The foundation is four competencies, all starting with D: Delegation, Description, Discernment and Diligence. Delegate the right task, describe what you want, judge what comes back, and stand behind the end result.
So far it sounds like any other framework rhetoric. What's interesting sits in two lenses the course gives you before you're allowed to touch a tool.
The first is the Creative Value Lens: what the work is for, and what makes it yours. The second is the Production Lens: how it gets made, step by step. The point is to overlay them. Wherever your value as a creator sits, you keep AI away, or on a short leash. Wherever it's mostly about production, about getting something done, you can delegate more.
It's a simple idea, and the most useful one I've seen articulated about creative work and AI in a long time.
My point
My point is less about the course than about what it reveals. Fluency, in Anthropic's sense, is about knowing when to abstain. The skill is judgement. And judgement is what's missing from most of what's being produced right now.
Anyone who's worked in content for a while recognises the pattern. A draft from a model is usually competent and rarely wrong. It's also usually average, because it's built from the average. A language model gravitates towards the most probable next word, which by definition is the middle. For a creator, the middle is the most dangerous place to stand. Whatever is average becomes invisible.
That's why protecting what makes the work yours, as the course puts it, isn't a soft side note. It's strategy. Byron Sharp has spent years insisting that distinctive assets make a brand memorable, and the same logic applies to an individual creator's voice. Delegate away what you're known for, and you delegate away the reason anyone listens to you.
The course splits the work into an inner and an outer loop. The inner loop, Description and Discernment, is the fast dialogue: you ask, you judge, you ask again. The outer loop, Delegation and Diligence, covers the bigger decisions about what you hand off and how you stand behind it. Discernment is the hardest, and most people skip it. Telling the difference between "this is good" and "this is merely fluent and plausible" is a trained ability. You train it by reading more, seeing more and sharpening your sense of what good looks like, not by prompting smarter.
If you work in content, this is worth a couple of hours. It's free and gives you a vocabulary for decisions you're already making on gut feel. Take Framework & Foundations first, it's the prerequisite. But don't go in expecting shortcuts, because that's the entire point. The course makes you aware of where your speed is costing you the thing that's actually worth something.
Sources
- 01
More to read
