TL;DR: Every frontier lab now runs a free academy, and the point is to make you native to their features and their scenarios. Anthropic's just grew to 22 courses and roughly 70 hours. Strip the duplicated API courses and the niche variants and the part worth having is the bit that isn't about Claude at all: the 4D framework, which every new Anthropic employee goes through on day one. Four hours, free, and it works just as well if you use ChatGPT or Gemini.
Every frontier lab runs an academy now. OpenAI, Google, Microsoft. This is not charity. I saw the same tactic decades ago, when I was an engineering student on a free Autodesk licence for AutoCAD. The point is to make you native to one company's features and one company's scenarios, so that its vocabulary becomes your default vocabulary.
On Thursday Anthropic scaled up their Claude Academy to 22 courses and roughly 70 hours, free, no account needed even to browse. You are not meant to do all 70 hours. It's a catalogue, not a curriculum. Three of the courses are almost the same API course tailored for different scenarios. Seven more are the same fluency material recut for students, educators, nonprofits and small businesses. Strip those and what survives is the one part that isn't about Claude at all: the 4D framework. Two academics built it before Anthropic got involved, it's released under a Creative Commons licence, and it's what every new Anthropic employee is taught on day one. And this is genuinely worth the time.
The four Ds, as four questions
Lesson one is four minutes long and states them plainly:
Delegation: when should humans do the work, and when should AI?
Description: how do we communicate clearly with AI systems?
Discernment: how do we evaluate what AI gives us?
Diligence: how do we ensure the interaction is responsible, transparent and accountable?
The course roadmap slide: one section per D, each framed as a question. Source: AI Fluency: Framework & Foundations, lesson 1.
The bar those four questions serve is a single line: can you engage with an AI system effectively, efficiently, ethically and safely?
The four criteria the course uses to define AI fluency.
Note who delivers it. Four presenters open the course and two of them are Joe Feller of Cork University Business School and Rick Dakan of Ringling College, the academics who wrote the framework. Their pitch is not about the tool at all: "We want to change how we think about AI so that we can learn how to think with AI."
The logo strip along the bottom of every slide makes the same point without saying it: Anthropic, University College Cork, Ringling College of Art and Design, Ireland's Higher Education Authority, and a Creative Commons badge. That is not the footer of a product tutorial.
What's actually worth taking
If you are | Take | Time | Why |
|---|---|---|---|
Anyone using AI at work | 4 hr | The 4D framework itself. Nothing in it is a Claude feature, so it survives you switching tools | |
Reviewing AI work others rely on | 3.5 hr | The companion piece: what the model is actually doing when it answers. Built on properties that stay stable as models improve | |
On Claude Code daily | Subagents, then agent skills | 2 hr | The two shortest courses on the site, and the best return per hour on it |
Connecting AI to your own systems | 1 hr | The protocol the whole industry is standardising on | |
New to Claude | 2.5 hr | A feature tour. Skim it at speed | |
Building on the API | One of the three API courses | 8 hr | Pick the cloud you actually deploy to and ignore the other two |
Take the first two and you've spent seven and a half hours on the only material here that is brand-agnostic and outlives the next model release.
Why should you give it a try?
The course says so itself: "Most AI training focuses solely on tactical skills like specific prompts, which settings to adjust or which systems to use. While useful, these skills can quickly become outdated."
The judgement underneath doesn't expire. And this is what you'll learn from it.